ABSTRACT
Title of dissertation: THE PYROCONVECTIVE PATHWAY FOR
STRATOSPHERIC WATER VAPOR AND
AEROSOL
George P. Kablick III, Doctor of Philosophy,
2019
Dissertation directed by: Professor Zhanqing Li
Department of Atmospheric & Oceanic Science
Earth System Science Interdisciplinary Center
A detailed analysis of pyrocumulonimbus (pyroCb) cases is presented that
explores their convective dynamics, stratospheric plume characteristics and down-
stream radiative effects. Satellite observations in conjunction with ground station
data and radiative transfer models are used to quantify the impact that pyroCbs
have on localized stratospheric aerosol and water vapor. The initial meteorologi-
cal and fire conditions are explored using a cloud- and aerosol-resolving model to
determine the dominant mechanics driving the convection and their effects on micro-
physics. Results show that intense sensible heat fluxes are the dominant convection
trigger over a wildfire in an unstable atmosphere. Direct observations by cloud
profiling radar of the active convective stage of a pyroCb are analyzed for the first
time, and comparisons with non-pyro meteorological deep convection in the same
vicinity and season show that the pyroCb has an extreme delay in the growth of
precipitation-sized cloud droplets to altitudes above the homogeneous freezing level.
Stratospheric aerosol plume morphology is analyzed for several cases, and
an empirical heat accumulation efficiency model is developed to describe observed
radiatively-induced self-lofting in the stratosphere. The model results suggest py-
roCb aerosol plumes are ? 30% efficient at converting shortwave radiative heating
into sensible heating, thereby driving buoyant uplift once injected into the strato-
sphere. PyroCbs directly inject H2O vapor into the stratosphere, which is shown
to be significantly large for two separate cases. The cloud-resolving model confirms
a previous hypothesis that uniquely small ice particle microphysics can enhance
stratospheric H2O in detrained convective anvils. Satellite retrieval evidence sug-
gests plume water vapor anomalies are a result of inefficient removal of small ice par-
ticles within the detrained pyroCb anvil. Model-injected total water?represented
as the sum of all ice and absolute humidity?shows at least 30% of H2O survives the
convective detrainment stage, and diminishes within the evolving plume over the
observation period when using satellite observations of H2O as a benchmark. In the
plumes presented herein, pyroCb H2O anomalies are as large as 4?3 ppmv above the
background in the lower stratosphere. Detailed line-by-line radiative transfer sim-
ulations suggest that these anomalies produce an instantaneous longwave radiative
forcing up to +1.0 W m?2 at the tropopause.
THE PYROCONVECTIVE PATHWAY FOR STRATOSPHERIC
WATER VAPOR AND AEROSOL
by
George P. Kablick III
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2019
Advisory Committee:
Professor Zhanqing Li, Chair
Michael Fromm, Ph.D.
Professor Russell Dickerson
Professor Ross Salawitch
Professor George Hurtt
?c Copyright by
George P. Kablick III
2019
Dedication
For my parents and grandparents.
ii
Acknowledgments
I?d like to thank my research advisor Zhanqing Li for providing wisdom, sup-
port, and the opportunity to pursue my Ph.D. research. I?m grateful for the en-
couragement. I?m also deeply grateful to Mike Fromm at the Naval Research Lab
who has been a great friend and mentor. I?d also like to acknowledge Professors
Russ Dickerson, Ross Salawitch and George Hurtt for serving on my dissertation
committee.
There are numerous additional people who have helped me throughout this
research, either through providing material support or beneficial discussion. They
include, Seoung Soo Lee, Dave Peterson, Gerald Nedoluha, Alyn Lambert, Michael
Schwartz, Steve Miller, Hugh Pumphrey, Pete Caffrey, Andrew Jongeward, and
many others. Thank you all.
Finally, I would like to thank my wife Meredith for her unending patience,
fortitude and encouragement, and also my children Lily and Peter for their comic
relief.
iii
Statement of Originality
This dissertation contains original research from which these major conclusions
can be drawn:
1. Cloud radar profiles through active an active pyroCb show precipitation for-
mation and droplet growth are suppressed in supercooled air until reaching
the homogeneous freezing level.
2. PyroCb plumes injected into the stratosphere enhance local water vapor mix-
ing ratios, and unique ice microphysics within detrained anvils probably con-
tribute to this enhancement.
3. Empirical results using convective cloud modeling tests show there needs to
be ? 30% survival of total water in the stratosphere from ice and vapor to
match observations.
4. Radiative-induced diabatic self-lofting is described with a heat accumulation
model based on the first law of thermodynamics, and an empirical efficiency
parameter to account for conversion of solar heating to buoyant uplift through
temperature increases. The efficiency for three different cases is found to be
? 30%.
5. Global water vapor enhancements in the stratosphere are quantified for an ex-
treme pyroCb event, and tropopause radiative forcing from this enhancement
is found to be +1.0 W m?2.
iv
All of these conclusions are based on research that has been conducted by
myself, either as the lead author on three papers or a co-author on two additional
papers:
Kablick, G., M. Fromm, S. Miller, P. Partain, D. Peterson, S. Lee, Y. Zhang,
A. Lambert, Z. Li (2018), The Great Slave Lake pyroCb of 5 August 2014:
observations, simulations, comparisons with regular convection, and impact
on UTLS water vapor, J. Geophys. Res., 123, (21), 12?332. (Chapter 2, 3)
Fromm, M., R. McRae, J. Sharples, and G. Kablick (2012), Pyrocumulonimbus
Pair in Wollemi and Blue Mountains National Parks, 22 November 2006, Aust.
Meteorol. Ocean. J., (62), 117?126. (Chapter 2)
Kablick, G., M. Fromm, Z. Li (2019), Self-Lofting of Biomass Burning Plumes
within the Stratosphere, to be submitted to Mon. Wea. Rev.. (Chapter 4)
Peterson, D., J. Campbell, E. Hyer, M. Fromm, G. Kablick, J. Cossuth, M. De-
Land, (2018), Wildfire-driven thunderstorms cause a volcano-like stratospheric
injection of smoke, npj Clim. Atmos. Sci., (1:30) (Chapter 4)
Kablick, G., M. Fromm, M. Schwartz, H. Pumphrey, G. Nedoluha, Z. Li (2019),
Radiative Effects of an Unprecedented Stratospheric Intrusion of Water Vapor
from Pyroconvection, to be submitted to J. Climate. (Chapter 5)
v
Table of Contents
Dedication ii
Acknowledgements iii
Statement of Originality iv
List of Tables viii
List of Figures ix
List of Abbreviations xvii
1 Introduction 1
1.1 UTLS Water Vapor and Climate . . . . . . . . . . . . . . . . . . . . 3
1.2 Tropopause-Involved UTLS H2O Trends . . . . . . . . . . . . . . . . 8
1.2.1 Deep Convection and Ice Microphysics . . . . . . . . . . . . . 11
1.3 Pyrocumulonimbus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 PyroCb Events 21
2.1 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.1.1 Great Slave Lake 2014 . . . . . . . . . . . . . . . . . . . . . . 22
2.1.2 Pacific Northwest Event 2017 . . . . . . . . . . . . . . . . . . 26
2.2 Properties of Active PyroCb . . . . . . . . . . . . . . . . . . . . . . . 30
2.2.1 Weather and Fire Conditions . . . . . . . . . . . . . . . . . . 36
2.2.2 PyroCb and Cb Comparisons . . . . . . . . . . . . . . . . . . 45
2.3 Detrained Stratospheric Plume Properties . . . . . . . . . . . . . . . 52
3 Cloud-Aerosol Modeling of Active PyroCb Convection 61
3.1 Model Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
vi
4 Diabatic-Lofting of Injected PyroCb Plumes 70
4.1 Stratospheric Injection Mass . . . . . . . . . . . . . . . . . . . . . . . 72
4.2 Plume Rise Observations and Simulations . . . . . . . . . . . . . . . 80
4.2.1 Satellite Observations . . . . . . . . . . . . . . . . . . . . . . . 80
4.2.2 Modeling Plume Rise . . . . . . . . . . . . . . . . . . . . . . . 87
5 Radiative Effects of Residual PyroCb-Injected UTLS Water Vapor 100
5.1 Observed Water Vapor Enhancements . . . . . . . . . . . . . . . . . . 101
5.2 Line-by-Line Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2.1 Fixed Anomaly Results . . . . . . . . . . . . . . . . . . . . . . 110
5.2.2 Lagrangian Anomaly Results . . . . . . . . . . . . . . . . . . . 116
5.3 The Big Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6 Summary of Work, Conclusions, and a Future Direction for PyroCb Impact
Studies 124
6.1 PyroCb Convection and Cloud Modeling . . . . . . . . . . . . . . . . 124
6.2 Stratospheric PyroCb Aerosol . . . . . . . . . . . . . . . . . . . . . . 127
6.3 Water Vapor Forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.4 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Bibliography 132
vii
List of Tables
2.1 June?August 2014 deep convective core cases found using the modi-
fied CloudSat identification algorithm described in the text within a
rectangular region between 50?65?N and 103?120?W, including the
GSL pyroCb and Alberta Cb. . . . . . . . . . . . . . . . . . . . . . . 49
4.1 2017 PNE pyroCb plume potential temperature observations (?) and
model-predicted (??) using various efficiency parameters (?). Also
shown are CALIOP stratospheric AOD and SBDART radiative heat-
ing rate anomalies (?Q) in K day?1. n?Q is the maximum tempera-
ture change assuming a steady state over n days between observations
(K; only shown where n > 1, otherwise ?Q = n?Q). Last row gives
total change in ? and each ??. ?? values in bold are closest match to
plume ?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.2 As in Table 4.1, but for the 2014 GSL pyroCb plume. . . . . . . . . . 95
4.3 As in Table 4.1, but for the 2009 Black Saturday pyroCb plume. . . . 97
4.4 PyroCb events with significant observational evidence of diabatic
plume rise that have been published in peer-reviewed literature. In-
cluded are three additional cases not analyzed herein (all values of z
in km and ? in K). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
viii
List of Figures
1.1 Perturbations and radiative transfer through AFGL midlatitude at-
mosphere. Model profiles in each panel show (a) temperature (black)
and H2O (blue), where the solid line is the standard mixing ratio and
dashed/dotted lines are the +1 or ?1 ppmv perturbations; (b) air
density (black) and the percentage WVMR perturbation (blue); and
(c) longwave fluxes (upward; red, downward; blue, net; black) and
cooling rates (green). The eight lines in (d) are the flux and cooling
rate differences between the +1 or ?1 ppmv perturbations on the
standard model results shown in (c). . . . . . . . . . . . . . . . . . . 7
1.2 Tropospheric phenomenon known to interact with the UTLS. . . . . . 17
2.1 Visible imagery of the GSL pyroCb over the course of three hours on
5 August as seen by Aqua MODIS (a, e), Terra MODIS (b, d) and
NPP VIIRS (c, f). Yellow star is location of the Buffalo Junction
surface site, across which the active pyroCb moved. Red line in (e)
is track of CALIOP and CloudSat (see Figure 2.6). Photographs in
(g?h) taken by Mike Smith of Yukon Wildland Fire Management, and
are used with permission. . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2 The PNE pyroCb event at approximately 20:00 local time on 12 Au-
gust 2017 as viewed by AVHRR. (a) BT11 and (b) BT3 showing the
typical pyroCb signature cold cloud tops in the thermal channel and
warm cloud tops in the near infrared (discussed in Section 2.2. Shown
here are four pyroCb anvils originating from three fires discussed in
the text. Fire locations are displayed as red pixels in (b). Images
created by Scott Bachmeier and released to public domain (CIMSS
pyroCb blog). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 (a?c) VIIRS images of the Pacific Northwest during the three days
centered on the PNE (red dots are fire locations). Rows (d?f) are
OMPS UVAI images on the days that the plume moved northwest-
ward around a high pressure Omega-block pattern, ?gathered? in
northern Canada, and subsequently moved southeastward. OMPS
images created by Colin Seftor and used with permission. . . . . . . . 29
ix
2.4 As in Figure 2.1c and 2.1f, but for VIIRS 3.7 ?m (a, b) and 10.8 ?m
(c) BT of the GSL pyroCb. All values are in K. Images (b) and (c)
are ? 100 minutes after a) on 5 August. Fire 14WB-025 is denoted
with yellow arrow in (a) and (b). Note the darker color of the pyroCb
anvil (yellow circle) in (b) as compared with the lighter cirrus clouds
to the northwest (upper left of images) as the convection becomes
detached from the fire, indicating an extremely small and narrow ice
particle size distribution. . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Aqua MODIS BT11 of the GSL pyroCb (a) and the concurrent Al-
berta Cb to the south (b) at 20:20 UTC. The red lines show the nadir
ground track of CALIOP and CloudSat. Note the ?cold-U? shape of
the pyroCb outflow anvil with warmer temperatures corresponding to
the central overshooting top. The Cb shows no such structure even
though the minimum brightness temperatures are similar: ?62 ?C
and ?64 ?C for pyroCb and Cb, respectively. . . . . . . . . . . . . . . 32
2.6 CALIOP 532 nm backscatter (top row) and CloudSat 94 GHz radar
reflectivity (bottom row) at the 20:20 UTC intersection of the active
GSL pyroCb (a,c) and Alberta Cb (b,d) cores (see red line in Fig-
ure 2.1e, 2.5a for pyroCb and Figure 2.5b for Cb). CALIOP shows
that the GSL pyroCb has a highly attenuating anvil cloud with over-
shooting top near 14 km (a), and CloudSat indicates strong echoes
in the mid-levels, but weak echoes near the surface, and no echoes
between 13?14 km (c). The yellow dashed line indicates the CALIOP
cloud top in both panels. Conversely for the Alberta Cb, CALIOP
(b) and CloudSat (d) have similar cloud top heights, and the radar
reflectivity shows a complex vertical structure of with large reflectiv-
ity near the cloud top and extending down to the surface. Also shown
are the 0 ?C and ?38 ?C isotherms from ERA-I reanalysis at both
locations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.7 Synoptic-scale conditions from NARR showing winds and geopoten-
tial heights at 300 hPa (a), relative humidity and winds at 850 hPa
(b), and surface frontal analysis (c). ??? symbols denote location of
GSL pyroCb. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.8 Map of fire perimeters (yellow with black outlines) and ground ob-
servation sites (black circles) maintained by the Government of the
Northwest Territories. Fire 14WB-025 spawned the GSL pyroCb.
Data from three of the observation sites shown in Figure 2.9. Red
dot marks location of photographs in Figure 2.1g,h. . . . . . . . . . . 38
2.9 Time series of temperature and relative humidity (a?c; black and
blue, respectively), wind speed (d?f) and both instantaneous and 24-
hour accumulated rainfall (g?h; orange and blue, respectively). Red
vertical line on time series plots denotes time of pyroCb. The pyroCb
moved directly over the Buffalo Junction Station (a,d,g). Ground-site
data courtesy of Franco Nogarin of the Government of the Northwest
Territories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
x
2.10 36-hour time-series of normalized hourly GOES-15 FRP (red) and the
mean estimated fire kinetic temperature (black) centered on the time
of pyroCb (green line). Note the large fire temperatures and peak in
FRP just before initiation. Gray shading is local nighttime. Gaps in
FRP are due to invalid retrievals by the WF ABBA algorithm likely
due to weak fire output in the local evening/morning. . . . . . . . . . 40
2.11 Interpolated 18:00 UTC sounding for the GSL pyroCb. Red hor-
izontal lines show CloudSat and CALIOP observed cloud bound-
aries. Atmospheric temperature between ? 900?1000 hPa is shown
adjusted by using average ground station Ts (red dashed line), pro-
ducing CAPE=1985 J kg?1 and CIN=1160 J kg?1 (blue and gray
shading, respectively). Thick black horizontal lines show LCL and
EL also computed with Ts. Green line is mixed-layer LCL (lowest 50
hPa) computed with Ts adjustment. CIN is limited to pressures be-
tween EL and the observed overshoot altitude in lower stratosphere.
Dark yellow bar between 3.7?4.2 km is convective cloud base height
range from ?Full Simulation? modeling experiment (Chapter 3). . . . 43
2.12 (Left) CloudSat reflectivity and CALIOP ??532 through the Wollemi
and Grose Valley pyroCb columns. (right) Trajectories initialized at
the high OMI UVAI values through the ground track of these two
nadir-looking instruments (black line) with the MODIS fire hot spots
(orange outlines). This figure adapted from Figures 3 and 6 of Fromm
et al. (2012). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.13 CloudSat horizontal-mean reflectivity profiles plotted as a function
of height above the LCL through deep convection during the period
1 June?31 August 2014 near the GSL pyroCb. Shown in (a) are the
GSL pyroCb (blue), Alberta Cb (red) and the additional 15 mete-
orological DCC profiles identified. Blue and red shading represent
the standard deviation about the mean for GSL pyroCb and Alberta
Cb. Horizontal lines in a) show FL (dot-dashed), HFL (dashed), EL
(solid) and CloudSat cloud top (dotted) for both the pyroCb and
Cb. Green line is the radar reflectivity from ARW Full Simulation
pyroCb model run. (b) Black line shows the linear regression slope
of mean reflectivity profiles between these thermodynamic levels and
gray shading shows the standard deviation for all cases. Individual
regressions of the GSL pyroCb and Alberta Cb cases are shown in
blue and red, respectively. . . . . . . . . . . . . . . . . . . . . . . . . 48
2.14 5-hourly GOES-15 IR brightness temperature imagery starting at
20:30 UTC on 08/05 comparing the anvil lifecycle of the GSL py-
roCb (blue arrows) and Alberta Cb (red arrows). . . . . . . . . . . . 53
xi
2.15 (Top) Map of HYSPLIT trajectory showing intersections with rele-
vant A-Train orbits. Segments are labeled A?I, and the date, segment
time stamp and corresponding HYSPLIT hour are given in the inset
table. (A?I) CALIOP VFM curtains showing the ?stratospheric fea-
tures? in black with MLS water vapor anomalies overlaid as colored
rectangles (% relative to local climatology). Black dashed line is 2.5
PVU dynamical tropopause and green contours are potential temper-
ature (K) from ERA-I reanalysis. Abscissae of A?I are in degrees of
latitude, and orange vertical lines bound the curtain profiles used in
Figures 2.17, 2.18. Corresponding ??532 curtains are shown in Fig-
ure 2.16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.16 As in Figure 2.15, but showing the ??532 instead of ?stratospheric
features? identified using the VFM product. Note the altitude range
is extended compared with 2.15 to show more of the backscattering
differences between strongly attenuating clouds in the troposphere
with the stratospheric pyroCb plume. . . . . . . . . . . . . . . . . . . 56
2.17 MLS water vapor profiles through the pyroCb plume (red lines) obser-
vations shown in Figure 2.15. Blue lines are the profiles of background
WVMR from the 2005?2014 JJA gridded climatology that fall within
the region bounded by orange lines in panels A?I in Figure 2.13, blue
shading shows the standard deviation of these background profiles.
The gray dashed lines delineate the potential temperature boundaries
of the plume over the course of its observable lifetime. . . . . . . . . . 58
2.18 Mean values of MLS WVMR anomalies and IWC (black and blue
lines, respectively) and median CALIOP ?532 (red lines) over the two-
week observable lifetime of the GSL pyroCb plume. Each MLS point
in this time series represents a per-profile vertical mean between 360?
400 K for measurements encompassing the plume (dashed lines in
Figure 2.14), and each CALIOP point is the median value for the
entire ?stratospheric feature? between 360?400 K. . . . . . . . . . . . 59
3.1 ARW Full Simulation ice particle mixing ratio within the cirrus out-
flow at the equilibrium level (? 13 km) during the mature stage of
pyroconvection (approximately 3 hours after convection began). Red
circle denotes fire hot-spot location used to initialize the simulation.
Compare with VIIRS image in Figure 2.1f. . . . . . . . . . . . . . . . 65
3.2 Profiles from four pyroCb model simulations showing the averaged
vertical distributions of the a) spatiotemporally-averaged updraft mass
fluxes, and b) ice mass densities. See text for descriptions of the four
simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
xii
3.3 Profiles of WVMR spatiotemporally averaged for each simulation for
all cloud regions within the domain (solid lines), and surviving wa-
ter post-detrainment (dashed lines), estimated here to be 30%. The
latter includes vapor contributions from ice sublimation and existing
absolute humidity. The MLS observation during the active pyroCb?
when no WVMR enhancement is yet present?is shown as a gray line
as a reference to the background profile. . . . . . . . . . . . . . . . . 68
4.1 The first reliable CALIOP ??532 intersection with PNE stratospheric
plume on 14 August (?1.5 days after injection) at 11:30 UTC. Over-
laid are MLS observations of CO. Maximum CO mixing ratio obser-
vation is > 670 ppmv (white ??? near 70.5 ?N). Columns of light
vertical shading labeled 1 and 2 are locations used to assess total
aerosol extinction for comparison with OMPS UVAI (see Figures 4.2
and 4.3). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2 Profiles of scattering ratio (left) and derived non-molecular extinction
(aerosol particulates, right) from the PNE plume intersection in Fig-
ure 4.1. (a, b) correspond with column 1 and (c, d) with column 2.
Vertical integration of plume extinction for tropopause height >11.5
km yields AOD ' 3.5 and ' 1.5, respectively. . . . . . . . . . . . . . 77
4.3 Empirical results of mass estimation of the PNE plume shown in Fig-
ures 4.1 and 4.2 showing the a) along-track AOD derivation and b)
simulated UVAI using the 330 nm and 390 nm channels. Red lines
show a) the value of ?total optical attenuation? of the downward-
looking vertical CALIOP instrument (note a majority of the plume
between 66-73 ?N has fully attenuated the lidar) and b) stratospheric
UVAI threshold. There is good agreement with the plume?s simu-
lated UVAI ? 15 and the stratospheric component of the CALIOP
observation in 4.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4 CALIOP/MLS observation of the PNE plume on 3 September 2017
at 21:00 UTC (? 3 weeks post-injection). Location of CO maximum
(? 170 ppbv) and large ??532 values show plume has risen at least 10
km above the injection height (white dashed line). . . . . . . . . . . . 81
xiii
4.5 Hand-drawn trajectories of three sections of the 2017 PNE plume
tracked using the dual MLS H2O/CO threshold. Panel (a) shows the
start date (12 September) when the plume could be sufficiently iden-
tified as three independently moving sections. The ?1 Month? tra-
jectory between the 12 August pyroCb date and the plume locations
on 12 September is an approximation based on the plume tracking
method described in the text. Panels (b?d) show termination dates
of each plume section?s observability. Colored dots are the GPH of
the MLS plume observation (bottom colorbar). Inset in panel (d)
focuses on the area over the mid-Atlantic where the Square plume
completed a circumnavigation; locations are marked when observa-
tions occurred between 29 September and 2 October (bearing east
toward west) and 11?25 November (bearing from southeast toward
northwest, then abruptly shifting toward northeast). . . . . . . . . . . 83
4.6 Potential temperatures of all northern hemisphere MLS thresholded
observations during the first three months of the PNE plume evolu-
tion. A constant H2O threshold of 7 ppmv is applied with diminishing
CO thresholds as the plume ages. The vertical dashed line marks the
point at which the plume separates over central Asia, and the three
dashed lines emanating from this time show the different ascent rates
of three tracked plume sections. . . . . . . . . . . . . . . . . . . . . . 85
4.7 Illustration of pyroCb smoke plume evolution from onset through de-
cay. Left: available observations are used to define the injection-?
at a specific location and time. Middle: the NP-phase is when the
earliest-possible 3D observations of the plume can be characterized,
and coincides with strong diabatic-lofting potential due to high con-
centrations of absorbing aerosol. Right: the EP-phase is the period
during which advection and dispersion are reducing the concentra-
tion of aerosol to the point of non-observability, and diabatic-lofting
potential is rapidly diminishing. . . . . . . . . . . . . . . . . . . . . . 86
4.8 (a) Observed and modeled diabatic rise of the PNE Square plume
(see red trajectory in Figure 4.5). Blue ??? show the potential tem-
perature of the CALIOP-observed plume centroid with error bars
denoting the upper and lower backscatter boundaries, augmented
with MLS plume observations of colocated HCN/CO maxima (green
dots). Simulation of radiatively-forced diabatic lofting for various ?
are shown over observations. Prince George radar echotop-based in-
jection ? = 370 K is noted. (b) The stratospheric AOD derived from
CALIOP extinction. . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.9 As in Figure 4.8, but for the GSL pyroCb plume. NP phase is shorter
than the 2017 PNE case before stratospheric ascent is observed. Also,
this plume was only trackable for approximately two weeks, and had
a much smaller total ??. Injected plume ? = 357 K, averaged over
depth of CALIOP observation is noted. . . . . . . . . . . . . . . . . . 94
xiv
4.10 As in Figure 4.8, but for the 2009 Black Saturday pyroCb plume.
Green ??? in this case are MLS observations colocated HCN/CO
maxima to fill CALIOP gap (light blue vertical bar between mid-
February to mid-March). Blue ??? are CALIOP observations, and
black ??? is surface lidar observation from Sa?o Jose? dos Campos,
Brazil. Melbourne surface radar echotop-based ? = 370 K is noted. . 96
5.1 August 2017 histograms of all Aura MLS H2O observations in the
Northern Hemisphere lower stratosphere using coincident CO thresh-
olds (a) ? 20 ppbv, (b) ? 30 ppbv, (c) ? 40 ppbv, (d) ? 70 ppbv.
Count densities shown for each 1 km layer between GPH = 15-25 km
for observations constrained by 380 ? ? ? 600 K. Note logarithmic
scale. Observations to the left of 7 ppmv threshold (dashed line) were
ignored for plume identification used in Chapter 4. . . . . . . . . . . 102
5.2 As in Figure 5.1 but for September 2017. . . . . . . . . . . . . . . . . 104
5.3 As in Figure 5.1 but for October 2017. . . . . . . . . . . . . . . . . . 105
5.4 As in Figure 5.1 but for November 2017. . . . . . . . . . . . . . . . . 106
5.5 (left column) Aura MLS WVMR anomalies (red contours) and fixed
local climatologies (black contours) within the three 2017 PNE py-
roCb plumes after the 12 September separation. All contours in
ppmv. (right column) Longwave LBLRTM heating rate anomalies
throughout the troposphere and lower stratosphere computed from
these observations. Square plume (top row), Triangle plume (mid-
dle row) and Circle plume (bottom row). Vertical gray bars indicate
where MLS missed observations. . . . . . . . . . . . . . . . . . . . . . 111
5.6 Fixed Anomaly H2O longwave radiative heating (left column) and
radiative heating anomalies (right column) for the individual dates
of the three plume sections. Note the significant cooling within and
above the plume and mild warming below the plume to the tropopause.114
5.7 TOA (left) and tropopause-level (right) longwave RF using the H2O
Fixed Anomalies from Figure 5.5. Note the Square plume is confined
to the tropics for most of its observable duration, and has a negative
RF at TOA (a decrease in OLR), whereas the Triangle and Circle
plumes are in the mid-latitudes, and have a positive RF at TOA. In
all cases, there is a positive RF at the tropopause. . . . . . . . . . . . 116
5.8 As in Figure 5.5, but vapor concentration anomalies and radiative
heat flux differences are calculated using the Aura MLS Lagrangian
H2O climatology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.9 As in Figure 5.6, but for the Lagrangian climatology. (a) Note the
Square plume has three general QLW profiles, and (c) the Triangle
has two as these plumes move to different latitude bands through
summer and winter months. . . . . . . . . . . . . . . . . . . . . . . . 118
5.10 As in Figure 5.7, but for the Lagrangian climatology. . . . . . . . . . 120
xv
5.11 The global Aura MLS H2O/CO data record with constant thresholds.
All potential temperature observations are shown for the a) northern
hemisphere and b) southern hemisphere where MLS observes H2O? 7
ppmv and CO ? 70 ppbv. Highlighted are the 2017 PNE plume in the
northern hemisphere and the lack of 2009 Black Saturday signal in the
southern hemisphere. Note the annual signals between 380 < ? < 400
K are from the Monsoon and those between 550 < ? < 600 K are
from each hemispheres polar vortex. . . . . . . . . . . . . . . . . . . . 122
xvi
List of Abbreviations
a.m.s.l. above mean sea level
AVHRR Advanced Very High Resolution Radiometer
BT brightness temperature
BTD brightness temperature difference
CALIOP Cloud-Aerosol Lidar with Orthogonal Polarization
CALIPSO Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation
CAPE convectively available potential every
Cb cumulonimbus
CCL convective condensation level
CIN convective inhibition
CPR Cloud Profiling Radar (CloudSat)
DCC deep convective clouds
ECMWF European Centre for Medium-range Weather Forecasts
EL equilibrium level
FL freezing level
FPH frost point hygrometer
GPH geopotential height
GSL Great Slave Lake
HFL homogeneous freezing level
HYSPLIT Hybrid Single-Particle Lagrangian Integrated Trajectory model
IWC ice water content
LCL lifting condensation level
MDT mountain daylight time
MLS Microwave Limb Sounder
MODIS MODerate Imaging Spectroradiometer
NARR North American Regional Reanalysis
NASA National Aeronautics and Space Agency
NOAA National Oceanic and Atmospheric Administration
NRL Naval Research Lab
PBL planetary boundary layer
ppmv parts per million (per unit volume)
ppbv parts per billion (per unit volume)
PSD particle size distribution
PVU potential vorticity units
pyroCb pyrocumulonimbus
pyroCu pyrocumulus
TTL tropical tropopause layer
UTC coordinated universal time
UTLS upper troposphere/lower stratosphere
VMR (water) vapor mixing ratio
xvii
w.r.t. with respect to
? potential temperature
re effective radius
xviii
Chapter 1: Introduction
The tropopause is a well-defined boundary between the troposphere and the
stratosphere in the midlatitudes. This boundary is delineated in several ways, but
the traditional meteorological definition is the ?cold-point? temperature in a vertical
sounding, defined by the World Meteorological Organization as the lowest altitude
at which the lapse rate (?) becomes consistently ? ? 2 ?C km?1 for a vertical
depth of 2 km (World Meteorological Organization, 1992). The air above and below
the tropopause is warmer than the boundary itself, producing a very stable region
that acts to inhibit transport across it. This phenomenon is easily observed when
an intense cumulonimbus (Cb) grows in height to a level of neutral buoyancy, also
called an equilibrium level (EL), where the temperature of a saturated convective
parcel equilibrates with the environmental temperature. Upon reaching an EL, a
majority of the convective column is no longer buoyant, and clouds spread out to
form a cirrus anvil. The tropopause represents one such boundary. Particularly
intense updrafts in Cb can penetrate through this level and form an overshooting
top, and these relatively rare events are a direct intrusion of tropospheric air and
cloud matter into the lower stratosphere (LS).
On the other hand, in the tropical latitudes the tropopause interface is now
1
thought of as a transitional layer instead of a hard boundary (Fueglistaler et al.,
2009). Traditional parcel theory still adequately explains convective events in this
region, but transport through the tropical tropopause layer (TTL) is thought to
be more common than troposphere/stratosphere exchange in the midlatitudes be-
cause convection can be more sustained and intense. However, there is uncertainty
in the dominant mechanisms of cross-boundary processes in both the tropics and
midlatitudes.
Even in the tropics where the boundary is less well-defined, the isentropic
surface (constant potential temperature) ? = 380 K is considered a highly stable
?overworld? that receives minimal direct influence on short time scales from the
troposphere (Holton et al., 1995). The TTL and sources of transport through it
have been (and are continuing to be) studied intensely as climate science increasingly
acknowledges the large role the stratosphere has on surface and tropopause energy
budgets, but aside from volcanic events, comparatively little attention has been given
to midlatitude phenomenon that can directly influence the stratosphere (Fromm
et al., 2010).
Over the past two decades, there has been increasing recognition that py-
rocumulonimbus (pyroCb) events are a common occurrence in the midlatitude sum-
mer, and they routinely penetrate the tropopause with a plume comprised of ice,
smoke and biomass burning gases (Fromm and Servranckx , 2003; Fromm et al.,
2005, 2008a). The stratospheric aerosol impact of these events may be important
on annual time scales when considering several events occur throughout a fire sea-
son, but singular pyroCb events can have major direct injections on the scale of
2
volcanoes (Peterson et al., 2018). In addition to the effects on the stratospheric
aerosol burden, recent work has shown that direct injection of ice during pyroCbs
may play a role in determining down-stream water vapor mixing ratio (WVMR) in
the stable layers of the LS (Kablick et al., 2018).
The overall impact that these events have on climate via direct stratospheric in-
jection is unknown, but may be important. The irregularity of major volcanic erup-
tions leads to an ex post facto determination of their overall importance on strato-
spheric processes, but pyroCbs are now recognized as a continuing seasonal occur-
rence with a basic understanding of the conditions necessary to generate them (Pe-
terson et al., 2017a,b). Therefore, it is worth studying this phenomenon in the
context of climate as they may be prognosticable in the future.
In order to do such a climate study, the nature of pyroCb dynamics first
needs to be established through convective structure, microphysics and detrainment
properties, in addition to the individual radiative effects water vapor and aerosol
have once injected into the stratosphere. This study starts this work. The rest of
this chapter provides motivation with background information, and then details the
objectives.
1.1 UTLS Water Vapor and Climate
Long-term WVMR records from different measurement sources disagree in
both the magnitude and the sign of any trend that may exist in the upper tropo-
sphere and lower stratosphere (UTLS). For instance, Hurst et al. (2011) analyzed
3
a 30-year record from balloon-borne frost point hygrometer (FPH) measurements
at the National Oceanic and Atmospheric Administration (NOAA) site in Boul-
der, Colorado, and concluded that there has been a +1.0 ? 0.2 ppmv (27 ? 6%)
increase since the start of the measurement period. Whereas Hegglin et al. (2014)
constructed a merged satellite record using bias-corrected data from several space-
borne instruments and a chemistry-climate model, and concluded both that the LS
has experienced a negative WVMR trend of ?0.27 ? 0.18 ppmv (2? uncertainty)
and that the Boulder balloon record is not representative of the global trend.
Lossow et al. (2018) presented an updated analysis on the Boulder dataset,
and showed an altitude-dependent trend ranging between +0.1 and +0.45 ppmv
decade?1, but this result once again does not agree with zonally-averaged trends
at the same latitude from the merged satellite data during the same time period.
However prior to these studies, a different methodology was used by (Solomon et al.,
2010) to construct a merged data set with the same satellite retrievals, and found an
increase in WVMR over this time period, but with strong decadal-scale oscillations.
The causes of these decadal oscillations, as well as the trends and the discrepancies
between averaging a continental point-measurement at Boulder and satellite zonal
averages remain a source of debate. What remains clear, however, is the importance
of lower stratospheric water on climate.
In their study, Solomon et al. (2010) explained that the semi-decadal decline
in stratospheric WVMR reported between 2001?2005 was significant enough to mit-
igate the rate of surface warming from greenhouse gases by ? 25%. That result
is based on full spectrum line-by-line radiative forcing calculations using the com-
4
bined balloon and satellite observations. However, there is disagreement over the
magnitude of tropopause-level radiative forcing (RF ) caused by stratospheric water
vapor anomalies. Estimates of the adjusted climate response depend on the source
of data and the type of radiative transfer model (RTM) used. Current estimates of
RF range from +0.24 W m?2 for a ? 1 ppmv increase above the tropopause from
1980 to 2000 (Solomon et al., 2010) to +0.12 W m?2 for a ? 0.7 ppmv increase
from 1979 to 1997 (Oinas et al., 2001). For comparison, the RF of anthropogenic
CO2 increases since 1750 is +1.66?0.17 W m?2 (Forster et al., 2007), and the total
contribution of H2O to the overall clear sky energy budget is ? 71 W m?2 (Kiehl
and Trenberth, 1997). Despite the difference in the magnitude of the estimates,
and without considering the secondary effects on stratospheric chemistry, there is
generally a good understanding that an increase in H2O above the tropopause will
cause warming in the troposphere and cooling in the stratosphere given the potency
of the H2O absorption spectrum.
Water vapor climate forcing can be shown from an instantaneous perspective
(i.e., ignoring feedbacks and chemistry changes) using a model atmosphere with
perturbed WVMR profiles, and computing RF . One-dimensional radiative forcing
is defined in terms of the upward and downward flux (F and F, respectively) as
the inverse of the change in net flux (Fnet) from a control state:
RF = ??Fnet, (1.1)
where
?Fnet = (F ? F)anomaly ? (F ? F)control. (1.2)
5
Figure 1.1 shows the forcing experienced by imposing a ?1 ppmv water vapor
anomaly on the Air Force Geophysical Laboratory (AFGL) standard midlatitude at-
mosphere (?control?). This figure shows the WVMR change in terms of an absolute
amount (a) and a vertically-resolved percentage (b). Note that the percent change
in Figure 1.1b is only shown for an increase. Temperature (a) and air density pro-
files (b) are held constant, and a 1-dimensional longwave (wavenumber ?? between
10?3000 cm?1) band model is used to compute the radiative flux and cooling rate
for both the unperturbed and the perturbed WVMR profiles, which are applied
uniformly throughout the total depth of the atmosphere.
This figure shows that increasing WVMR in the upper troposphere and lower
stratosphere increases F at the tropopause while silghtly decreasing F (blue and
red dashed lines in Figure 1.1d, respectively). The combined effect of this is to create
a positive longwave RF . Note the black dashed line is ?Fnet, so that negative values
in the profile represent positive RF . The cooling rate is the flux divergence in a
layer, so negative values of cooling rates?and in this case cooling rate differences?
indicate warming with respect to the control atmosphere.
This simple example illustrates the importance of understanding short- and
long-term changes in LS H2O because this increase produces an instantaneous
tropopause RF > 0. H2O is also very important in stratospheric ozone budgets, and
water vapor feedback has been estimated to be 0.3 W m?2 for a +1 ?C temperature
anomaly in the mid-troposphere (Dessler et al., 2013), so the issue extends beyond
direct radiative forcing.
6
a) b)
c) d)
Figure 1.1: Perturbations and radiative transfer through AFGL midlatitude atmo-
sphere. (Continued on following page)
7
Figure 1.1: (continued) Model profiles in each panel show (a) temperature (black)
and H2O (blue), where the solid line is the standard mixing ratio and dashed/dotted
lines are the +1 or?1 ppmv perturbations; (b) air density (black) and the percentage
WVMR perturbation (blue); and (c) longwave fluxes (upward; red, downward; blue,
net; black) and cooling rates (green). The eight lines in (d) are the flux and cooling
rate differences between the +1 or ?1 ppmv perturbations on the standard model
results shown in (c).
1.2 Tropopause-Involved UTLS H2O Trends
An interesting approach taken by Hurst et al. (2011) was to statistically sep-
arate the midlatitude stratospheric water vapor observations series into distinct
epochs based on statistical growth/decay trends. Doing this, they found significant
growth in the trends of three out of four total epochs, with a significant decline in
one epoch and an strong altitude gradient among the trends for all epochs (see their
Figure 1). For the earliest epoch (1980?1990), they showed stronger trends near the
tropopause?an indication that the majority of the increase was from tropospheric
sources. The third epoch of their analysis, corresponding to the 2001?2005 decline
noted by Solomon et al. (2010), was the only period with a negative trend at all
altitudes.
The Hurst et al. (2011) study also presented average monthly profiles that
showed an obvious increase in mixing ratio above 19 km, which they attributed
8
to photochemical production involving methane (see their Figure 3). The seasonal
variation of the midlatitudes was also apparent below this altitude, with larger mix-
ing ratios in the boreal summer months. This seasonal variability could be caused
by two possibilities: 1) during boreal summer, there can be an increase in moisture
transport from the tropical tropopause layer (TTL) from stronger isentropic mixing
across the tropopause boundary with the weakening of the Hadley circulation and
the sub-tropical jet (Fueglistaler et al., 2009), or 2) there is an increase in over-
shooting convection in the midlatitudes, contributing to larger mixing ratios from
detrained convective parcels.
That conclusion is based on the current understanding that the only significant
sources of stratospheric H2O are photochemical methane oxidation in the upper
stratosphere (Bates and Nicolet , 1950) and meridional transport of comparatively
moisture-rich air that has been lifted through the tropical tropopause (Kley et al.,
2000). However, if all stratospheric methane is oxidized (creating approximately
two H2O molecules per molecule of CH4), it would still not create enough water to
explain the observed increase over their 30-year Boulder balloon record.
It is possible that the dominant source of LS H2O could be upwelling through
the TTL. Tropical influences on midlatitude fluctuations in LS moisture are thought
to be controlled by transient moisture uplift through the TTL where the WVMR
of lifted air masses is subject to the ambient temperature of the surrounding air,
and then those air masses are subsequently transported poleward (Sherwood and
Dessler , 2001; Fueglistaler et al., 2009; Wang and Dessler , 2012). In the tropics?
the upward branch of the Hadley Circulation?integration of these transient events
9
results in a positive moisture flux into the stratosphere from below. The amount
of water vapor that eventually crosses this vertical boundary is thought to be a
function of the coldest temperature the air mass encounters as it is lofted. Ther-
modynamic considerations dictate that as a parcel is cooled, the saturation vapor
pressure decreases and the vapor in the parcel is lowered to equilibrium by conden-
sation to the ice phase (Pruppacher and Klett , 1997), Thus, the coldest temperature
the parcel experiences would determine its water vapor mixing ratio upon entry into
the LS. Typically this coldest temperature occurs at the tropopause, and colder tem-
peratures inhibit moisture flux by ?freeze drying? the air as it transported (Jensen
et al., 2001). The implied possibility in (Hurst et al., 2011) is that this source of
midlatitude LS H2O may be increasing.
However, it is also possible?and even likely?that the FPH instruments used
in the Boulder record have experienced inconsistencies in both manufacturing and
calibration over the 30-year duration used in the study, and that these inconsis-
tencies could be artificially causing a positive WVMR trend over this time period.
This possibility is bolstered by the fact that the Boulder balloon record disagrees
with various merged satellite records (Hegglin et al., 2014). However, the vertical
differences in rates of WVMR increase may have a physical source.
The cold-point temperature in the TTL is known to have a strong seasonal
fluctuation, so there is a natural cycle to the moisture flux across the tropical
tropopause. In fact, most of the LS H2O trends discussed in Hegglin et al. (2014)
are shown with the chemistry-climate model to be a result tropopause temperature
fluctuations. The annual mean temperature at the tropical tropopause is ? 195 K
10
with a ? 8 K peak-to-peak annual fluctuation (Fueglistaler et al., 2009). The cold-
est temperatures occur during boreal winter and the warmest temperatures during
boreal summer.
The dominant cause of these temperature fluctuations is thought to be a re-
sult of the seasonal migration of the Intertropical Convergence Zone (ITCZ), which
increases (decreases) the radiative cooling as the latitude of maximum convection
moves closer to (farther away from) the equator during the boreal winter (summer).
Also, other cycles such as the El-Nin?o Southern Oscillation (ENSO) and the quasi-
biennial oscillation (QBO) in the stratosphere affect the interannual variability of
tropopause temperature (Ramaswamy et al., 2001). In summary, there are strong
natural cycles that can contribute to UTLS moisture fluctuations on shorter time
scales. One lesser-explored aspect, however, is how ice microphysics may play a role,
and to what extent any changes to this role would be natural. The main source of
ice in the UTLS is deep convection, so could changing deep convection frequency or
ice cloud properties influence LS H2O?
1.2.1 Deep Convection and Ice Microphysics
A major conclusion of the Hurst et al. (2011) study was that the combination
of methane-produced H2O in the upper stratosphere and tropical changes do not
adequately explain the vertical gradients of moisture growth or their magnitudes
during the four epochs between 1980 and 2010. Intuitively, there would be a need
for at least one other important mechanism that increases midlatitude stratospheric
11
water vapor with stronger contributions at higher altitudes, assuming the positive
trends in the Boulder balloon record are not instrument/retrieval-related. Those
authors postulate that a combination of dynamical and thermodynamical factors are
involved, which would contribute to changes in both the creation and redistribution
of H2O.
Wang and Dessler (2012) showed the seasonality in LS moisture from 2005?
2010 and the vertical phase shift in the UTLS mixing ratio at four isentropic levels
as observed by the Microwave Limb Sounder (MLS) instrument on-board the Aura
satellite (? = 360, 370, 380 and 400 K corresponding to altitudes of approximately
15, 16, 17 and 18 km, respectively). That study confirmed previous conclusions from
legacy H2O-monitoring instruments that there is an upward propagation to higher
altitudes, and it is not instantaneous; there is an ? 4 month phase delay in the
peak value between the lowest level and the highest level 3 km above. This delay is
referred to colloquially as the ?atmospheric tape recorder? (Mote et al., 1996).
In addition, Wang and Dessler (2012) also showed that there is a strong sea-
sonal dependence on the maximum amount of LS WVMR independent of the tape
recorder phenomenon, which is driven by the QBO?s effect on tropopause tempera-
ture. They argued that the outlying events that produce maximum values (such as
intense overshooting convection) are well-correlated with the seasonal-mean values.
They suggested that this process indicates intense overshooting convection with its
associated ice clouds and large WVMR does play a significant role in moistening
the LS, at least in the tropics. However, for the sake of completeness it must be
stated that tropical overshooting convection is relatively rare, and the major mech-
12
anism determining H2O mixing ratios entering the stratosphere is still considered to
be the traditional Brewer model of slow tropical tropospheric ascent driven by the
meridional radiative budget imbalance (Brewer , 1949).
Therefore, the majority of stratospheric moisture budget studies are concerned
with changes to UT mixing ratios, not overshooting convection. Depending on the
ambient saturation vapor pressure at these levels where ice is detrained, there may
be a net hydration or dehydration of these air masses due to the removal/survival
processes of ice crystals (Randel and Jensen, 2013). It seems counterintuitive, but an
upper tropospheric air layer already above saturation would cause detrained crystals
to grow in size to a point at which they would fall out (gravitational settling), thus
drying the air. The obverse is also true where subsaturated air would ?remove?
the crystals by sublimation in situ, thereby moistening the air. A recent long-term
modeling- and observation-based study reaffirmed this process (Schoeberl et al.,
2018).
As Sherwood (2002a) suggested, anomalies in stratospheric moisture trends
may be caused/augmented by changes to the microphysical properties of ice clouds
by anomalous aerosol concentrations affecting convective properties. This effect
would not be confined to the tropics because, for instance, aerosol-invigorated con-
vection in the midlatitudes could result in more overshooting anvils. This would
constitute an additional, independent source of moisture in the LS not subject ei-
ther the supersaturation of UT air or the slow ascent across the tropical tropopause.
13
1.3 Pyrocumulonimbus
As the previous sections in this chapter have illustrated, there is ample mo-
tivation for understanding the processes involved with stratospheric penetration,
and many of the known phenomenon that do so are the focus of many research
projects. To that end, this project focuses on one that is under-studied: pyroCbs.
These midlatitude events commonly inject plumes into the UTLS (Damoah et al.,
2006; Fromm et al., 2010; Peterson et al., 2017a). They typically form over large
wildland fires in the mid- to high-latitudes on an annual basis, and produce a plume
comprised of a mixture of biomass burning aerosols, gases, and cloud ice particles
that otherwise would not be present in the UTLS in such abundance (Fromm and
Servranckx , 2003).
Since their discovery, scientific understanding of these events has increased:
a conceptual model has been put forth regarding the meteorological and burning
conditions necessary for pyroCb development (Peterson et al., 2017b); the droplet
and ice particle size distributions (PSD) have been shown to possess a very small
effective radius (Rosenfeld et al., 2007); and the outflow anvils are known to have a
distinct increase in lifetime compared with meteorological cumulonimbus forming in
the same synoptic conditions (Lindsey and Fromm, 2008). However, there are still
several aspects of the pyroCb phenomenon that require a better understanding, such
as the internal dynamics of the updraft column, the relative importance of aerosol-
cloud microphysical interactions, the stratospheric impact of injected smoke, ice,
and water vapor.
14
Both observations and modeling studies of regular (non-pyro-) convection have
shown that small ice particles result from larger quantities of cloud condensation nu-
clei (CCN) in the boundary layer (Khain et al., 2005), and that anvil size is generally
larger (Fan et al., 2013; Yan et al., 2014). In these situations CCN abundance re-
duces the average droplet size in the updraft columns, and a so-called ?invigoration
effect? may result (Rosenfeld et al., 2008; Li et al., 2011). A reduction in droplet
size limits collisions and coalescence, thereby maintaining a greater amount of con-
densed water that will increase latent heating above the freezing level and enhance
the buoyancy of rising air (Rosenfeld and Woodley , 2000; Khain et al., 2005).
For pyroCb events, idealized model simulations have also shown that a large
CCN concentration can significantly enhance latent heat release (Reutter et al.,
2014), but the role of this aerosol effect on updraft velocity was shown to be small
when compared with buoyancy enhancements from extreme sensible heat fluxes over
a fire (Luderer et al., 2006). An additional consideration behind the large number
density of small ice particles in convective clouds is the strength of the updraft itself.
The supersaturations are larger in such conditions, leading to rapid generation of
ice crystals above the freezing level (Rosenfeld et al., 2007). Regardless of the
dominant influence?whether it is a CCN effect on the PSD or the surface heat
flux?it is certain that pyroconvection has intense updrafts and clouds with very
small droplets and ice crystals.
A corollary to diminishing ice crystal size is that the inefficient removal of these
particles has the potential to increase absolute humidity within an aging plume. Ice
particles that comprise convective anvils undergo various removal processes as they
15
age, including sublimation and sedimentation. Sublimation occurs when conditions
become sub-saturated with respect to ice as the anvil mixes with ambient air. In
a study based on convective anvil penetration of the tropical tropopause, Sherwood
(2002b) hypothesized that mixing drier air into ice-rich cirrus containing small par-
ticles would reduce the saturation ratio to the point that ice sublimation would
increase downstream absolute humidity as compared with cirrus with larger parti-
cles.
In a previously-mentioned follow up paper, Sherwood (2002a) relied on geo-
stationary satellites to infer there was a detectable anti-correlation between tropical
UTLS absolute humidity and convective anvil ice particle size. In other words,
smaller effective radii in convective cirrus near the tropical tropopause preceded an
increase in water vapor concentrations at higher altitudes, to where tropical upper
tropospheric air slowly migrates (Fueglistaler et al., 2009). In addition to this grad-
ual upward lifting, overshooting convection can directly deposit water into the LS
as discussed in the previous section. Although locally significant, this mechanism
is thought to be less important in the overall budget than the transient uplift of
moisture in UT air (Dessler , 2002). The overall implication of these mechanisms
is that the rate of moistening/drying in the TTL plays a direct role in determining
the LS water vapor budget.
However, cross-tropopause transport in the mid-to-high-latitudes is compar-
atively rare, and is typically only a result of direct injection (Holton et al., 1995;
Bedka, 2011). Figure 1.2 illustrates the meteorological- and volcanic-source events
known to impact the UTLS. Number 6 on the figure represents pyroCb that occur
16
UTLS Pathways
1. Brewer-Dobson circulation 5. Volcano
2. Quasi-isentropic exchange 6. pyroCb
3. Monsoon 7. xCb
1 4. Baroclinic cyclone
25
Stratospheric Stratosphere
20 Plume
5 1
1
15 Lower Stratosphere 420 K
TTL UTLS 2
Plume 380 K
Lowermost Stratosphere
10
5 Tropopause
L Upper Troposphere
3
5
4 7 6 Free
Troposphere
0
Tropics Subtropics Extratropics Polar
Figure 1.2: Tropospheric phenomenon known to interact with the UTLS.
poleward of the subtropics, where the tropopause is typically between 10-12 km
a.m.s.l. As demonstrated in this figure, Cb have weak overall LS penetration, and
the potential for direct injection of air masses into the stable regions above ? ' 380
K is very small compared with pyroCb.
Another microphysical process to consider is sedimentation, which could re-
duce the impact a sublimating ice cloud would have on absolute humidity by remov-
ing a potential vapor source from the parcel. In this process ice precipitates to lower
altitudes after the updrafts maintaining the suspended particles are weakened or as
particles aggregate into sizes large enough to gravitationally overcome the updraft
forces.
In their study on tropical convective anvils, Jensen et al. (2009) noted that
17
Altitude (km)
both the sublimation and growth rates of ice crystals are a function of the saturation
ratio and a deposition coefficient (see their Figure 11). They used in situ aircraft
observations during the Tropical Composition, Cloud, and Climate Coupling (TC4)
campaign to argue that small ice crystals do not persist longer than a few hours; they
either grow through deposition or they completely sublimate in subsaturated air;
their results were consistent with previous work on other types of small-ice-crystal
clouds such as wave clouds. In the case of a pyroCb, an anvil would need to have an
ice saturation ratio very close to 1.0 in order for the crystals to neither sublimate
nor grow too quickly to be removed by sedimentation. Lindsey and Fromm (2008)
used infrared observations to show pyroCb anvils persisted between ? 18?30 hours
longer than that of nearby meteorological Cb anvils post-detrainment, which would
mean the saturation ratio must be favorable to ice for a longer period of time.
In addition to aerosols and ice, pyroCb are known to inject a large amount
of combustion-generated gases into the stratosphere (Pumphrey et al., 2011). One
of these is water vapor, which has the distinct role of contributing to hydrometeor
and latent heat processes throughout the depth of the convective cloud (Potter ,
2005). Meteorological deep convection over North America and the Asian Monsoon
is known to directly inject locally-substantial amounts of water vapor into the lower-
most stratosphere, but the satellite-based studies have shown the net effect of those
injections to be small (Schwartz et al., 2013). PyroCb, however, are unusually
intense and their cloud properties are distinct from Cb, so an important question is
whether ice microphysics plays a role in the amount of water vapor injected into the
LS during these events since changes to ice PSD may affect changes in sublimated
18
vapor down-stream. For this to be true, ice crystals would have to sublimate in the
UTLS instead of being removed through sedimentation. The best way to address this
question would be from in situ aircraft measurements through active and detrained
pyroCb anvils. Unfortunately such an encounter has not been documented, so this
study relies on remote sensing instruments and a cloud-resolving model.
Lastly, the net aerosol loading from pyroCb events is unknown, but is possibly
a significant source of ?background? UTLS aerosol optical depth. Stratospheric
aerosol sources were once thought to be either volcanic in origin or chemically-
produced locally with tropospheric sulfate precursors (Brasseur and Solomon, 1986).
Smoke (black and brown carbon) are the main aerosol components of pyroCb plumes,
and there is a strong need to understand the lifetime and evolution that these plumes
experience once injected.
1.4 Objectives
The objectives of this study include the following:
1. Establish the convective lifecycle of a pyroCb with an emphasis on i) convective
triggers, ii) aerosol influences on cloud microphysics, iii) meteorological con-
ditions, and iv) stratospheric detrainment. Additional analysis is to compare
pyroCb dynamics with regular convection. (Chapter 2)
2. Simulate pyroCb convective dynamics and detrained plume properties with an
aerosol-cloud-resolving model, performing sensitivity tests on i) aerosol load-
ing, ii) meteorological precursors and iii) fire-generated sensible heat flux. Es-
19
timate these individual contributions to stratospherically-injected total water
content, including ice cloud mass and absolute humidity. (Chapter 3)
3. Develop a radiative transfer framework to simulate the stratospheric pyroCb
aerosol plume morphology, and make an estimate of total mass injected for
an extreme case. Validate the simulated results with observations. Examine
additional cases from literature within the same radiative framework to test
its robustness. (Chapter 4)
4. Set benchmarks for the water vapor component of pyroCb plume radiative
forcing based on observed WVMR anomalies using a highly detailed absorption
spectrum and a line-by-line radiative transfer model, and test the sensitivity
of this forcing for two different climatologies used to compute the WVMR
anomalies. (Chapter 5)
20
Chapter 2: PyroCb Events
The stratospheric effects of pyroCb injections may be better understood when
analyzing the nature of the convective events themselves. Meteorological and fire-
related dynamics contribute to both the initiation and the maintenance of convective
growth, and it is possible the intensity of this relationship is directly related to the
net stratospheric impact of the injected plumes. For example, recent studies have
shown that pyroCb initiation can be triggered by either dynamical uplift from an
approaching front (Peterson et al., 2018) or by a sudden increase in fire energy
output (Kablick et al., 2018). Both of these types of triggers?although different
mechanisms?are simply a perturbation to the atmospheric stability above the fire
column.
This chapter provides an analytical overview of the meteorological conditions
and the stratospheric impacts of two well-documented case studies, and augments
this analysis with additional evidence from supplementary published cases. It fo-
cuses heavily on a pyroCb that formed south of Great Slave Lake (GSL) in the
Northwest Territories of Canada in 2014, but introduces an additional pyroconvec-
tion case from the Pacific Northwest in 2017 used in later chapters.
The former pyroCb (hereafter ?GSL pyroCb?) is detailed herein because it
21
had very good observational coverage from both satellites and ground sites. Data
collected from both these sources are substantial enough to provide detailed insight
into its convective lifecycle, and allow for comparisons with regular (non-pyro) Cbs.
Additionally, an aerosol- and cloud-resolving convective model is used to parse the
important differences between surface heat fluxes and latent heating effects of ex-
treme aerosol concentrations on the convection and subsequent plume properties,
including the effect on detrained stratospheric water vapor.
The latter pyroCb, or rather pyroCb cluster (hereafter Pacific Northwest Event
?PNE?) created the largest-ever observed stratospheric pyroCb plume. It was ob-
servable using several satellite instruments for at least five months post-UTLS in-
jection. It is introduced in this chapter, but the bulk of analysis on the PNE case
is contained in Chapters 4 and 5.
2.1 Case Studies
2.1.1 Great Slave Lake 2014
The GSL pyroCb began at approximately 18:45 UTC on 5 August 2014 over
fire 14WB-025 (number designation used by Parks Canada/Wood Buffalo National
Park; 60.2?N, 115.5?W). The fire had been burning for approximately six days after
a lightning strike on 30 July. There were many fires burning in and around the Wood
Buffalo National Forest and GSL regions at the time, some of which had already
produced pyroCbs prior to 5 August, but the GSL pyroCb had fortuitous satellite
coverage in addition to several ground stations recording observations nearby during
22
the active stages of convection. Additionally, the detrained plume was observable
by instruments in NASA?s A-Train (Stephens et al., 2002; Winker et al., 2009) for
over two weeks after the event. As such it provides a useful case for an examination
of a complete pyroCb lifecycle, including any water vapor impact of UTLS-injected
ice.
Over the course of three hours on 5 August multiple polar orbiting satellites
captured the growth and development of the GSL pyroCb prior to, during, and after
the active stages of convection. Figure 2.1 shows True-Color images by the MODer-
ate Imaging Spectroradiometer (MODIS) from both the Aqua (a, e) and Terra (b,
d) platforms and similar images by the Visible Imaging Infrared Radiometer Suite
(VIIRS) on-board Suomi NPP (c, f). A combination of the high northern latitude
summertime conditions and the unusually early local time for pyroconvection (12:45
MDT) allowed each of the these imagers to capture multiple daytime passes.
Figure 2.1a-f show the GSL fire near the center of the images, and document
the pyroCb from the early stages of cumulus congestus (a?c) through the mature
convective column (d, e) and the detraining anvil stage (f). At approximately 20:20
UTC (14:20 MDT), Aqua MODIS made its second daytime pass (e), which occurred
directly over the pyroCb column. The red line in Figure 2.1e shows the nadir ground
track of the active A-Train instruments discussed later in this section: the CloudSat
Profiling Radar (CPR, Stephens et al., 2002) and the Cloud-Aerosol Lidar with
Orthogonal Polarization (CALIOP, Winker et al., 2009). Apparent brightness of
smoke and haze between each image are due to the increase/decrease in sunlight
scattering into the sensor as the viewing angle changes.
23
g) 19:51 UTC h) 20:28 UTC
Figure 2.1: Visible imagery of the GSL pyroCb over the course of three hours on 5
August as seen by Aqua MODIS (a, e), Terra MODIS (b, d) and NPP VIIRS (c, f).
Yellow star is location of the Buffalo Junction surface site, across which the active
pyroCb moved. Red line in (e) is track of CALIOP and CloudSat (see Figure 2.6).
Photographs in (g?h) taken by Mike Smith of Yukon Wildland Fire Management,
and are used with permission.
24
Figure 2.1g-h are photographs taken by Mike Smith of Yukon Wildland Fire
Management of the pyroCb approximately one-half hour apart during the ?mature?
stage of active pyroconvection (location where these photographs were taken is de-
noted as red circle on Figure 2.8 near Sandy Lake Road). Note the appearance of
mammatus like clouds in Figure 2.1h when the anvil had moved overhead of the
photographer?s location.
In the days leading up to the GSL pyroCb, there was a large amount of smoke
being generated in the region, and it prevented many of the perimeter flights used to
accurately estimate daily fire progression [personal communication with Northwest
Territory Fire Resource personnel ]. However, perimeter data derived from MODIS
Level 2 hot-spots indicate the fire consumed about 65400 ha in total during the six
day period between ignition and the pyroCb event. The fire just to the east of center
in Figure 2.1 at ? 114?W (fire 14WB-028; hereafter ?Eastern Fire?) also produced
a pyroCb. Cumulus congestus is also seen forming and dissipating over the Eastern
Fire throughout the six snapshots in Figure 2.1, but it does not develop a pyroCb
until after the anvil from the GSL pyroCb passes nearby, at which time it became
apparent in geostationary imagery (discussed further in Section 2.2.2).
The intense convection from the Eastern Fire did not produce a mature a
pyroCb until approximately 22:30 UTC, two hours after the GSL Fire. It is likely
that the anvil from the Eastern Fire pyroCb interacted with the GSL pyroCb anvil
blowoff to produce a larger UTLS plume than would have existed otherwise as
indicated by the two-tiered appearance of the early morning visible-band ?day after?
plume at 11:30 UTC on 6 August (see Supporting Information Movie S2 in Kablick
25
et al. (2018)).
It is unknown why the two pyroCbs, both within the same region favorable
to convection (discussed in Section 2.2.1), developed over two hours apart. Any
number of factors such as fuel types and microscale meteorology could play a role in
different rates of development for two pyroCb in the same region, but an in depth
examination of the sensitivity to these variables is beyond the scope of this work.
However, Section 2.2.1 analyzes the meteorological state and discusses the possible
triggers for convection. Chapter 3 uses this information as initial conditions in
several model simulations to show the surface heat flux is likely the most important
factor in determining pyroCb generation.
2.1.2 Pacific Northwest Event 2017
Up until 12 August 2017, it was widely acknowledged by experts in the field
that the largest pyroCb event (in terms of a stratospheric plume) was the ?Black Sat-
urday? event from 8 February 2009 in Victoria, Australia (Siddaway and Petelina,
2011; Pumphrey et al., 2011). That event has been?and continues to be?widely
studied both for the surface fire and weather conditions (Dowdy et al., 2017) and
the long-lasting UTLS impact (Field et al., 2016). Then in the summer of 2017,
four separate clusters of fires burning in the Pacific Northwest generated five total
pyroCbs within the span of 5 hours on the afternoon of 12 August and generated the
largest ever observed pyroCb plume (Peterson et al., 2018). The fires were burn-
ing along the Cascades in Washington State and on the eastern side of the Coast
26
126 124 122 120 118 126 124 122 120 118
54 54
52 52
a) b)
Figure 2.2: The PNE pyroCb event at approximately 20:00 local time on 12 August
2017 as viewed by AVHRR. (a) BT11 and (b) BT3 showing the typical pyroCb
signature cold cloud tops in the thermal channel and warm cloud tops in the near
infrared (discussed in Section 2.2. Shown here are four pyroCb anvils originating
from three fires discussed in the text. Fire locations are displayed as red pixels in
(b). Images created by Scott Bachmeier and released to public domain (CIMSS
pyroCb blog).
Mountain range in western British Columbia in advance of a frontal low pressure
system moving in from the Pacific Ocean. Each pyroCb was active and intense for
several hours, and the plumes of these fires were entrained into a synoptic jet that
advected them to the northern coast of Canada over the next two days.
Figure 2.2 shows images from the polar-orbiting Advanced Very High Resolu-
tion Radiometer (AVHRR) of the mature stage of this pyroconvection swarm. The
left panel (a) shows the thermal infrared brightness temperature at 10.8 ?m (BT11),
and panel (b) shows 3.9 ?m (BT3). Red dots in panel (b) are the fire hot spots.
27
This convection occurred in an area that was very favorable to uplift with a very
hot, dry and windy surface and an elevated moisture layer between 600?700 hPa
associated with the advancing cold front.
Figure 2.3a?c shows VIIRS visible imagery over 11?13 August with fire hot
spots overlaid as red dots. Panel (c) clearly shows the frontal clouds draped from
southeast to northwest with a heavy concentration of smoke towards the top of the
image. Rows (d?f) show the progression of the plume in terms of its Ultra-Violet
Aerosol Index (UVAI) from the Ozone Mapping and Profiler Suite (OMPS) on-board
Suomi NPP. The UVAI is a two channel algorithm that qualitatively detects both the
concentration and altitude of absorbing aerosol plumes. Larger values correspond
to both higher-altitude- and densely-concentrated-plumes, and this event generated
values as large as 55 (unitless).
The progression over 12?16 August appears to show that the individual pyroCb
plumes became ?concentrated? as they moved around an Omega-block-patterned
high pressure system that was situated over Alberta and Saskatchewan. Chapter 4
discusses of the early days of this plume concentration and possible consequences
for its UTLS evolution. It also provides more information on the UVAI metric and
its use in determining aerosol concentration. Chapter 5 examines this plume from
a water vapor perspective, and estimates the UTLS H2O component to radiative
forcing.
28
British
Columbia British British
Columbia Columbia
a) 11 Aug Washington b) 12 Aug Washington c) 13 Aug Washington
d) 14 Aug: max. AI = 49
e) 15 Aug: max. AI = 50
f) 16 Aug: max. AI = 55
Figure 2.3: (a?c) VIIRS images of the Pacific Northwest during the three days
centered on the PNE (red dots are fire locations). Rows (d?f) are OMPS UVAI
images on the days that the plume moved northwestward around a high pressure
Omega-block pattern, ?gathered? in northern Canada, and subsequently moved
southeastward. OMPS images created by Colin Seftor and used with permission.
29
2.2 Properties of Active PyroCb
This section utilizes the GSL pyroCb case to examine the similarities and
differences between pyroCb and meteorological Cb cloud properties. Figure 2.4
contains VIIRS BT3 during the times corresponding to Figure 2.1c,f. The images
are constructed so the fire location hot spots identified by this channel are displayed
as red-orange-yellow colors with the GSL Fire denoted by the yellow arrow, and
the cooler BT3 values corresponding to terrestrial and cloud temperatures are in
gray-scale.
One common feature observed with pyroCbs in this part of the BT spectrum
is a ?warm? anvil (Lindsey and Fromm, 2008; Fromm et al., 2010). Lindsey et al.
(2006) explained there is a strong inverse relationship between daytime BT3 and
ice particle size for diameters <80 ?m at a constant solar zenith angle since the
solar reflectance component increases dramatically with decreasing effective radius
(re). In the terrestrial-only infrared part of the spectrum (e.g., channels such as
11 and 12 ?m), there is an insignificant contribution from solar reflectance, and for
optically thick clouds at a given emitting temperature there is little difference in
behavior of these BTs with particle size. The largest influence in these channels is
the temperature of the cloud itself. A visual comparison of the pyroCb anvil BT3
in Figure 2.4b (yellow circle) to that of the cirrus clouds in the upper left corner of
the images highlight the effect of particle size. Similar effects between the thermal-
and the near-IR BT channels are seen for the PNE case in Figure 2.2.
However, the optical thickness of a cloud needs to be accounted for when
30
a) 19:35 UTC b) 21:15 UTC c) 21:15 UTC
250 260 270 280 290 300 310 320 330 340 350 360 195 210 225 240 255 270 285 300
Figure 2.4: As in Figure 2.1c and 2.1f, but for VIIRS 3.7 ?m (a, b) and 10.8 ?m
(c) BT of the GSL pyroCb. All values are in K. Images (b) and (c) are ? 100
minutes after a) on 5 August. Fire 14WB-025 is denoted with yellow arrow in
(a) and (b). Note the darker color of the pyroCb anvil (yellow circle) in (b) as
compared with the lighter cirrus clouds to the northwest (upper left of images) as
the convection becomes detached from the fire, indicating an extremely small and
narrow ice particle size distribution.
assuming daytime BT3 ?warmness? indicates a small re. Brightness temperatures
from pixels containing semi-transparent clouds, for example, can have a significant
contribution of radiance from emission below the cloud layer, making them appear
warmer than opaque clouds at the same altitude. Figure 2.5a shows the BT11 from
Aqua MODIS at 20:20 UTC (corresponding with Figure 2.1c). Minimum BT11 for
the pyroCb top is ?62 ?C, and the anvil has a distinct ?cold-U? feature typical of
intense, overshooting thunderstorms (Setva?k et al., 2010).
One commonly used cloud optical opacity test is the brightness temperature
31
a) GSL pyroCb b) Alberta Cb
Figure 2.5: Aqua MODIS BT11 of the GSL pyroCb (a) and the concurrent Alberta
Cb to the south (b) at 20:20 UTC. The red lines show the nadir ground track of
CALIOP and CloudSat. Note the ?cold-U? shape of the pyroCb outflow anvil with
warmer temperatures corresponding to the central overshooting top. The Cb shows
no such structure even though the minimum brightness temperatures are similar:
?62 ?C and ?64 ?C for pyroCb and Cb, respectively.
difference between two thermal infrared channels in the atmospheric window, in
this case MODIS channels 32 (12 ?m) and 31 (11 ?m), denoted herein as BTD12?11.
Small absolute values (|BTD ?12?11|< 3.0 C) indicate optically thick clouds. This
test was validated by Peterson et al. (2017a) for use with pyroCb anvils, and the
result of that test on the GSL pyroCb anvil yields a BTD12?11 spread between ?0.1
to +1.0 ?C; well within the ?optically thick? range. This result gives confidence that
the warm daytime BT3 values indicate an abundance of very small ice particles in
the anvil.
32
A few minutes earlier during the same daytime Aqua MODIS pass, the A-
Train observed a concurrent meteorological Cb in central Alberta (Figure 2.5b;
hereafter ?Alberta Cb?) centered at 53.5?N. The Alberta Cb underwent convective
development at approximately the same time (? 18:45 UTC) as the GSL pyroCb as
seen in GOES-West (GOES-15) imagery, and had a minimum BT = ?64 ?11 C with
|BTD ?12?11| < 1.0 C, and a more uniform appearance to the BT field (no ?cold-U?).
CloudSat and CALIOP also passed directly over a deep convective portion of the
Alberta Cb, as shown by a red line.
Figure 2.6 contains the ?curtains? through both the GSL pyroCb and the Al-
berta Cb from these sensors. Level 1 CALIOP 532 nm total attenuated backscatter
(??532) is shown above the corresponding CloudSat CPR 94 GHz radar reflectiv-
ity from the 2B-GEOPROF product, and overlaid on the images is a dashed line
that represents the uppermost cloud observation by CALIOP for each case. The
lidar detects the pyroCb overshooting cloud top altitude to be approximately 14
km a.m.s.l., whereas the radar does not observe this overshoot. It has a maximum
cloud height detection of 13 km (Figure 2.6a, c). CALIOP observes large ??532 and
full attenuation within a very narrow vertical depth (< 1 km), whereas the CPR
has radar reflectivities beneath the noise threshold of ?28 dBZe at these uppermost
cloud top levels.
This discrepancy is not present with the Cb case, where both instruments
detect a cloud near 12 km (Figure 2.6b, d). The sensitivities of these instruments to
ice particle size have been documented in many places (Miller and Stephens , 2001;
Stephens et al., 2002; Austin et al., 2009; Delanoe? and Hogan, 2010) and is thought
33
a) GSL pyroCb b) Alberta  Cb
c) GSL pyroCb d) Alberta Cb
Figure 2.6: CALIOP 532 nm backscatter (top row) and CloudSat 94 GHz radar
reflectivity (bottom row) at the 20:20 UTC intersection of the active GSL pyroCb
(a,c) and Alberta Cb (b,d) cores (see red line in Figure 2.1e, 2.5a for pyroCb and
Figure 2.5b for Cb). CALIOP shows that the GSL pyroCb has a highly attenuating
anvil cloud with overshooting top near 14 km (a), and CloudSat indicates strong
echoes in the mid-levels, but weak echoes near the surface, and no echoes between
13?14 km (c). The yellow dashed line indicates the CALIOP cloud top in both
panels. Conversely for the Alberta Cb, CALIOP (b) and CloudSat (d) have similar
cloud top heights, and the radar reflectivity shows a complex vertical structure of
with large reflectivity near the cloud top and extending down to the surface. Also
shown are the 0 ?C and ?38 ?C isotherms from ERA-I reanalysis at both locations.
to be the cause of the cloud top discrepancy in the GSL pyroCb. The mid-visible
CALIOP is most sensitive to the second moment of a PSD?proportional to the
total cross-sectional area of all particles contributing to scattering?whereas the W-
Band CPR is most sensitive to the sixth moment?proportional to the total volume
34
of scatterers. Therefore the GSL pyroCb anvil must be comprised of a large number
of very small scatterers as compared with the Alberta Cb. Using this reasoning in
combination with the warm BT3 and cold BT11 (Figures 2.4 and 2.5, respectively) it
is concluded that the GSL pyroCb generated an extremely narrow ice PSD with very
small re. Note that these two instruments orbit in a formation designed to observe
the same scene?nominally flying within 15 seconds of each other?so differences in
cloud morphology between observations is minimal and would not account for the
apparent cloud top discrepancy.
An interesting feature of note in Figure 2.6c is the large value of radar reflec-
tivity associated with apparently dry smoke just to the northwest of the pyroCb.
This smoke is visible below ? 4 km between 60.7?60.8?N (compare with visible
MODIS image in Figure 2.1e), and has values approaching ?5 dBZe. Smoke by
itself typically does not produce detectable CloudSat radar reflectivities, but being
close to the source, it is likely that this smoke contains chaff or other relatively large
biomass-burning debris that has not yet fallen out (Fromm et al., 2012). Plotted
on top of each panel in Figure 2.6 are the isotherms at T = 0 and ?38 ?C from
the CloudSat ECMWF-AUX product?the freezing level (FL) and homogeneous
freezing level (HFL), respectively.
These thermodynamic levels are known to be important boundaries for latent
heat processes in convection, so they are denoted on these curtains for later reference.
In Section 2.2.2 these thermodynamic boundaries are used along with the lifting
condensation level (LCL) and equilibrium level (EL) to discuss the microphysical
interpretation of the CPR reflectivity profiles of the GSL pyroCb and Alberta Cb.
35
21:00 UTC 300 hPa wind barbs and contours 21:00 UTC 850 hPa wind barbs (kt) and  
(kt) and geopotential height contours (gpm) relative humidity contours (%)
1016
72N 72N
100
66N 66N 80 1016
1008
60N 60N 60 1012
40
54N 54N
20
48N a) 48N b) c)
140W 130W 120W 110W 129W 120W 111W 102W
Figure 2.7: Synoptic-scale conditions from NARR showing winds and geopotential
heights at 300 hPa (a), relative humidity and winds at 850 hPa (b), and surface
frontal analysis (c). ??? symbols denote location of GSL pyroCb.
Also included is a statistical analysis of additional CPR intercepts through deep
convective clouds (DCC) from the same region and season. Prior to that discussion,
however, it is useful to establish the meteorology and fire precursors.
2.2.1 Weather and Fire Conditions
Figure 2.7 show the weather conditions in place prior to the initiation of con-
vection. Winds and geopotential heights at 300 hPa from the North American
Regional Reanalysis (NARR, Mesinger et al., 2006) indicate the pyroCb formed at
the southern edge of a jet entrance-region with an approaching shortwave trough
aligned with the Pacific Northwest coast (Figure 2.7a). These conditions are known
to be favorable for updraft formation and convection (Uccellini and Johnson, 1979).
There was also an approaching increase in tropospheric moisture around 850 hPa
(Figure 2.7b).
36
PyroCb modeling has shown mid-tropospheric moisture can be entrained into
the convective column and contribute to instability aloft (Trentmann et al., 2006),
and observational statistics of all intense pyroCbs from 2013 within North America
seem to confirm this (Peterson et al., 2017b). Additionally, a strong low-level jet
was also in place at 850 hPa which could have increased entrainment rates, helping
to sustain the convection once it began (see wind barbs in Figure 2.7b). Ground
observation sites operated by the government of the Northwest Territories recorded
surface temperatures ? 32 ?C, relative humidities ? 20% and sustained wind speeds
between 15-20 k h?1 (see Figures 2.8 and 2.9). These are the classic hot, dry, and
windy conditions associated with intense wildland fires.
The traditional fire-weather indicators, the Haines Indices (Haines , 1988), were
high-risk at the time of the GSL pyroCb with values of 6, 6, and 3, for the low-,
mid-, and high-elevation variants, respectively. A stationary surface boundary was
generating convection to the north/northwest around this time (Figure 2.7c), but
did not appear to reach far enough south to directly impact the pyroCb location as
evidenced by the lack of independent (non-pyro) cloud formation in the immediate
vicinity of the fire (see Figure 2.1). Therefore, with favorable conditions, but without
the dynamical trigger in place to initiate the convection, it is concluded that the
fire itself was the trigger, and provided the necessary energy to initiate convection.
This conclusion is supported by the behavior of GOES-West fire radiative
power (FRP) and kinetic fire temperature just before the GSL pyroCb began. A
36-hour time series of these values is shown in Figure 2.10, centered around the
time of pyroconvection (green line). Within this timeframe, the temperature of the
37
Wbnpqd2
Figure 2.8: Map of fire perimeters (yellow with black outlines) and ground observa-
tion sites (black circles) maintained by the Government of the Northwest Territories.
Fire 14WB-025 spawned the GSL pyroCb. Data from three of the observation sites
shown in Figure 2.9. Red dot marks location of photographs in Figure 2.1g,h.
fire (black dots) oscillates between 500-700 K, but the pyroCb only forms after the
strong increase in FRP (red line) just before to 18:00 UTC (12:00 MDT). The FRP
increased from < 500 MW to > 6000 MW in less than 2.5 hours prior to the pyroCb,
and fire temperatures increased by ? 150 K over the same time. This preceding
rapid increase in FRP is consistent with the conceptual model put forth in Peterson
et al. (2017b). On this day the fire spread quickly, burning at an average rate of 2725
ha h?1 based on daily MODIS fire perimeter data, which is an order of magnitude
38
a) b) c)
d) e) f)
g) h) i)
rainfall 
24 accum. rainfall
Figure 2.9: Time series of temperature and relative humidity (a?c; black and blue,
respectively), wind speed (d?f) and both instantaneous and 24-hour accumulated
rainfall (g?h; orange and blue, respectively). Red vertical line on time series plots
denotes time of pyroCb. The pyroCb moved directly over the Buffalo Junction
Station (a,d,g). Ground-site data courtesy of Franco Nogarin of the Government of
the Northwest Territories.
larger than all other days on which the fire burned. The surface heat fluxes generated
by such a rapid increase in FRP would certainly trigger the initial upward motion
of surface and planetary boundary layer (PBL) parcels in an unstable atmosphere.
Figure 2.2.1 shows an interpolated sounding using the 5 August 12:00 UTC and
39
Figure 2.10: 36-hour time-series of normalized hourly GOES-15 FRP (red) and
the mean estimated fire kinetic temperature (black) centered on the time of pyroCb
(green line). Note the large fire temperatures and peak in FRP just before initiation.
Gray shading is local nighttime. Gaps in FRP are due to invalid retrievals by the
WF ABBA algorithm likely due to weak fire output in the local evening/morning.
6 August 00:00 UTC soundings from the nearby Ft. Smith observation station. To
construct this sounding, a temporal linear interpolation was used at each pressure to
estimate the T and dewpoint temperature (Td) at 18:00 UTC. Beneath the remaining
PBL inversion, an atmospheric temperature adjustment is estimated at this time
by connecting the top of the inversion with the mean surface temperature (Ts =
35 ?C) measured at nearby ground stations (red dashed line below 1 km). This
adjustment gives the sounding a general appearance of the ?inverted-V? type, which
was also shown in Peterson et al. (2017b) to be a typical precursor to intense pyroCb
development.
The convectively available potential energy (CAPE) of the sounding without
40
the Ts adjustment is approximately 250 J kg
?1 with an LCL of 1.8 km (not shown).
This value increases to 1985 J kg?1 (blue shading in Figure 2.2.1) with an LCL of
2.5 km when using the Ts adjustment. One noteworthy feature of this sounding is
the moist layer of air around 450 hPa associated with the advancing trough. The
dewpoint depression at this pressure (? 4.5 ?C) is the lowest in the free troposphere.
Entrainment of dry air into growing convection at this pressure would typically
reduce buoyancy by cooling the air through evaporation (Simpson, 1980), but it
is possible that the higher relative humidity associated with this layer would limit
the drying effects of entrainment, and therefore be less inhibitive to vertical cloud
development.
For pyroCb cases, it could be argued that the convective condensation level
(CCL), rather than the LCL, is more appropriate to use as an estimate for cloud
formation because the CCL represents the level at which a surface parcel would
saturate from lifting due to heating, rather than forced dynamical uplift. However,
in the case of the adjusted sounding, the CCL is found to be at an altitude ? 2.6 km,
which is near enough to the LCL for the difference to be ignored. This occurs because
the adjusted PBL temperatures closely follow a dry adiabatic profile. At the time
the pyroCb began, Ts had almost reached the convective temperature (Tc ' 36 ?C)
needed to trigger free convection above the condensation level. All these arguments
have been made without considering an increase to the PBL air temperature from
fire-enhanced surface heat fluxes, which should be quite large as indicated by FRP.
Thus, although dynamical triggers were likely on the cusp of interacting with the
GSL fire, and Ts was approaching Tc in nearby non-fire areas, it is safe to conclude
41
that the fire triggered the convection.
Additional features shown in Figure 2.2.1 are the cloud boundaries seen by
CALIOP and CloudSat. The overshooting top peaks at ? 14 km and the cloud
base (according to radar reflectivity) is near 4 km. The large difference between the
LCL and the cloud base is most likely attributed to the large difference between the
adjusted sounding and the actual T and Td from surface heat flux contributions pre-
viously mentioned. Indeed, Lareau and Clements (2016) used ground-based mobile
doppler lidar to show cloud bases over fires are typically much higher than LCLs
determined from nearby soundings. The GSL pyroCb retrieved kinetic fire temper-
ature was > 300 ?C, which leaves little doubt that the lower-most PBL lapse rate
should be super-adiabatic (the CAPE under this condition would be on the order
of 15000 J kg?1). The equilibrium level (EL) of the adjusted sounding at 12.4 km
is representative of the tropopause.
The gray shading in Figure 2.2.1 represents the inhibiting energy (CIN) needed
to overshoot into the lower stratosphere to 14 km (CALIOP cloud top). The gray
area is encompassed by this cloud top, and both the parcel temperature and sounding
temperature, which have a vertex at the EL. This CIN value is 1160 J kg?1), and?
using parcel theory?would necessitate an updraft velocity on the order of 50 m s?1
at the EL for convection to reach the observed overshoot altitude. Unfortunately,
the GSL pyroCb did not occur within range of a operational doppler radar, so
no confirmation can be made, but this updraft velocity is consistent with intense
overshoots from regular convection that have been modeled to inject water vapor
into the lower-most stratosphere (Wang , 2003).
42
CALIOP cloud top (~14 km)
EL (~12.4 km)
Full Simulation 
cloud base range
CloudSat cloud base (~4 km)
Ts adj. mixed-layer LCL (~3.3 km)
LCL with Ts adjustment (~2.5 km)
TS = 34?
Figure 2.11: Interpolated 18:00 UTC sounding for the GSL pyroCb. (Continued on
following page)
43
Figure 2.11: (continued) Red horizontal lines show CloudSat and CALIOP observed
cloud boundaries. Atmospheric temperature between ? 900?1000 hPa is shown ad-
justed by using average ground station Ts (red dashed line), producing CAPE=1985
J kg?1 and CIN=1160 J kg?1 (blue and gray shading, respectively). Thick black
horizontal lines show LCL and EL also computed with Ts. Green line is mixed-layer
LCL (lowest 50 hPa) computed with Ts adjustment. CIN is limited to pressures
between EL and the observed overshoot altitude in lower stratosphere. Dark yellow
bar between 3.7?4.2 km is convective cloud base height range from ?Full Simulation?
modeling experiment (Chapter 3).
An interesting feature of the GSL pyroCb is the apparent lack of precipitation
during its most intense stage. The Buffalo Junction ground station (denoted by the
star in Figure 2.1) was operating nominally at the time when the pyroCb advected
over its location, and recorded no precipitation even as the wind speed peaked above
18 km h?1 and surface temperature dipped by approximately 4 ?C, presumably in
response to downdrafts or temporary cloud cover (see Figure 2.9). Despite attempts
to estimate precipitation amounts with CloudSat reflectivity, the estimates are not
scientifically useful over land surfaces due to a lack of reliable path-integrated at-
tenuation [Matthew Lebsock, CloudSat product developer, personal communication].
However, the CPR did measure significant reflectivity from the surface through the
deepest part of the pyroCb, and Figure 2.6c shows a vertical gap between the surface
return and the cloud base. These features provide additional confidence that there
44
was little-to-no precipitation during the mature stage of this event. Note that the
narrow column of radar echoes observed below the cloud base in Figure 2.6c is likely
dense smoke/lofted debris as discussed previously.
2.2.2 PyroCb and Cb Comparisons
Previous studies of pyroconvection have rarely had the ability to leverage ob-
servations of the interior during the active convective stage. Some exceptions to this
generality include the following: aircraft have flown through developing pyrocumulus
to take radiative flux measurements (Gatebe et al., 2012), ground-based operational
radar has been used to determine pyroconvective cloud tops from the upper-most
echoes (Dowdy et al., 2017), and doppler lidar has been deployed to remote locations
to estimate entrainment rates and density currents within pyroconvection (Lareau
and Clements , 2016). To our knowledge, however, only one other pyroCb case has
had a direct penetration of the active convective column by CloudSat and CALIOP
within the ? 12-year data record: the 2006 Wollemi case in Australia presented
in Fromm et al. (2012). However, that CloudSat intersection occurred over 40 km
to the east of the fire as the pyroCb was being advected away by strong winds, and
it is possible that the CPR did not capture the full vertical depth of the cloud (see
Figure 2.12). Note, however that Figure 2.12a shows the UVAI from the Ozone
Monitoring Instrument (OMI) to be well correlated with the pyroCb.
Thus, the current GSL pyroCb case is strategic because of contemporaneous
active profiling through the updraft core of a pyroCb and nearby Cb. In this section,
45
Grose Valley Wollemi Cirrus
LCL
Figure 2.12: (Left) CloudSat reflectivity and CALIOP ??532 through the Wollemi
and Grose Valley pyroCb columns. (right) Trajectories initialized at the high OMI
UVAI values through the ground track of these two nadir-looking instruments (black
LCL
line) with the MODIS fire hot spots (orange outlines). This figure adapted from
Figures 3 and 6 of Fromm et al. (2012).
CPR reflectivity is used to qualitatively estimate the internal hydrometeor structure
of the GSL pyroCb by partitioning the profile according to the thermodynamic levels
outlined in Section 2.2.1, and comparing it to the Alberta Cb and several other Cbs
that were observed over the course of two months in the same region.
The mean CloudSat radar reflectivity profiles for the ?deep convection? por-
tions of the GSL pyroCb and concurrent Alberta Cb are shown in Figure 2.13a
(blue and red lines, respectively). The shading about each of these two profiles
is a horizontal standard deviation of CloudSat reflectivity within these portions of
each storm. Here, ?deep convection? is defined using the following two sequences
46
in an attempt to select the most intense profiles. This methodology is similar to
the algorithm used in CloudSat Level 2 Product 2B-CLDCLASS with the following
variations: the assignment of vertical cloud boundaries, the application of MODIS
thermal infrared BT selection criteria, and the additional reflectivity thresholds.
First, use the co-located CloudSat ECMWF-Aux product to a) determine the LCL
altitude to use as the cloud base and b) determine the uppermost EL, and then use
the 2B-CLDCLASS to c) find profiles with a single continuous cloud layer between
the cloud base and EL. Second, use 2B-GEOPROF to d) confirm radar reflectivity
values ? 6 dBZe at 8 km a.m.s.l. or ? 4 dBZe at 9 km a.m.s.l., and then use Aqua
MODIS to e) confirm BT ? ?40 ?11 C and f) BTD ?12?11 > ?0.5 C. After the deep
convection CPR profiles are identified, the reflectivities are horizontally-averaged to
reduce noise.
Also shown in Figure 2.13a are 15 additional deep convection core (DCC)
profiles from June-August identified using the above algorithm for the same region
in Canada (light gray lines). Table 2.1 gives the latitude, longitude and CloudSat
?granule? identification number of all 17 DCC cases, including the GSL pyroCb, Al-
berta Cb. All of these DCC profiles are plotted as a function of altitude above the
cloud base (local LCL) to normalize the differences since convective cloud growth
begins at the cloud base. The horizontal lines in Figure 2.13a represent the relevant
thermodynamic levels for the pyroCb (blue) and Cb (red), and show the freezing
level (FL; 0?; dot-dashed), homogeneous freezing level (HFL; ?38?; dashed), the
equilibrium level (EL; solid) and the CloudSat cloud top (dotted). Here, the Cloud-
Sat cloud top is defined as the lowest level above which the median reflectivity drops
47
a) b)
Figure 2.13: CloudSat horizontal-mean reflectivity profiles plotted as a function
of height above the LCL through deep convection during the period 1 June?31
August 2014 near the GSL pyroCb. Shown in (a) are the GSL pyroCb (blue),
Alberta Cb (red) and the additional 15 meteorological DCC profiles identified. Blue
and red shading represent the standard deviation about the mean for GSL pyroCb
and Alberta Cb. Horizontal lines in a) show FL (dot-dashed), HFL (dashed), EL
(solid) and CloudSat cloud top (dotted) for both the pyroCb and Cb. Green line
is the radar reflectivity from ARW Full Simulation pyroCb model run. (b) Black
line shows the linear regression slope of mean reflectivity profiles between these
thermodynamic levels and gray shading shows the standard deviation for all cases.
Individual regressions of the GSL pyroCb and Alberta Cb cases are shown in blue
and red, respectively.
below the noise threshold (?28 dBZe) for a distance of of 1 km. Defining the cloud
top this way prevents spurious reflectivity values that may exist above a DCC from
being identified as the cloud top. Situations with multiple cloud layers are flagged
as non-DCC and ignored (i.e., the reflectivity increases above the noise threshold
48
Table 2.1: June?August 2014 deep convective core cases found using the modified
CloudSat identification algorithm described in the text within a rectangular region
between 50?65?N and 103?120?W, including the GSL pyroCb and Alberta Cb.
Date approx. lat/lon (?N, ?W) CloudSat granule
06/02 (52,110) 43073
06/09 (57,144) 43175
06/11 (58,111) 43204
06/22 (52,107) 43364
06/25 (53,112) 43408
06/27 (55,110) 43437
06/29 (51,105) 43466
07/04 (59,114) 43539
07/06 (51,107) 43568
07/06 (55,108) 43568
07/06 (62,113) 43568
07/20 (54,111) 43772
07/24 (53,104) 43830
07/24 (61,109) 43830
08/05 (53,111) - Alberta Cb 44005
08/05 (61,115) - GSL pyroCb 44005
08/09 (56,106) 44063
for a continuous altitude range at heights > 1 km above a cloud top).
The pyroCb has a reflectivity of ?10 dBZe at cloud base compared with values
ranging from approximately ?5 to 5 dBZe for the Alberta Cb and other meteoro-
logical DCC observations. The peak reflectivity for all DCC (including the GSL
pyroCb) is ? 10 dBZe, and generally occurs between 5 to 8 km above the cloud
base. The Alberta Cb shows a profile consistent with mid-level precipitation with
two bright bands just beneath both the FL and HFL, whereas the GSL pyroCb pro-
file has a more monotonic increase between the FL and HFL. Included for reference
in Figure 2.13a is a simulated CloudSat radar reflectivity profile from the modeled
49
GSL pyroCb discussed in Chapter 3 (green profile).
To better quantify the differences in these profiles, a linear regression slope
of reflectivity between the aforementioned thermodynamic levels is computed. Fig-
ure 2.13b shows the values of these regression slopes for the GSL pyroCb (blue),
Alberta Cb (red) and all DCC cases (black; this includes the Alberta Cb). The
gray shading is the standard deviation about the mean of all meteorological DCC.
The pyroCb stands out with a few interesting features: 1) reflectivity between the
LCL and FL decreases by ?2 dBZe, whereas the meteorological DCC have positive
slopes, 2) there is a large positive slope (+5 dBZe) between the FL and HFL, with
DCC having a neutral or only slightly positive slope, and 3) the pyroCb has a neu-
tral slope between the HFL and EL, whereas all meteorological DCC are strongly
negative.
Radar reflectivity at 94 GHz through a cloud is proportional to the total
condensed water content, so the profile regression slopes are representative of the
change in integrated hydrometeor volume with height. Therefore a weak(strong)
positive{negative} slope indicates a slowly(rapidly) growing{diminishing} hydrom-
eteor distribution between the thermodynamic levels. Using this interpretation, the
GSL pyroCb most likely has relatively weak droplet growth just above the cloud
base that rapidly increases in the layer between the FL and HFL. This could be
caused by an overabundance of CCN as predicted by the invigoration effect dis-
cussed in Rosenfeld et al. (2008), but it may also be a result of intense updrafts
from surface heating and a subdued saturation ratio at the cloud base. The large
positive slope in reflectivity (? 5 dBZe) between the FL and HFL indicates a rapid
50
growth in total condensed water between these levels. This is likely due to the strong
updrafts from the latent heat of freezing and/or buoyancy driven by surface heat
flux. The slopes between the EL and cloud top are all strongly negative for all cases,
including the GSL pyroCb. This is because of the rapidly diminishing condensate
above the EL.
In the GSL pyroCb case, the regression slope between these levels is similar
to the meteorological DCC slopes, even though the ice particles in these levels were
shown to be smaller than the Cb, and should therefore have weaker reflectivity.
Considering the difference in CALIOP and CloudSat cloud tops, however, the mag-
nitude of this regression slope value is probably biased low because the actual cloud
top (as seen by CALIOP) is about 1 km above the CPR-observed cloud top. If the
regression was computed between the EL and CALIOP cloud top, the slope would
be approximately ?10 dBZe per the depth of the layer.
These CloudSat reflectivity profiles augment the passive imagery to help show
the GSL pyroCb contained a large number of very small ice particles near the cloud
top. The warm BT3, cold BT11 and weak |BTD12?11| confirmed the presence of
small ice re, and since the ? 10 dBZe reflectivity at EL is similar to other DCC
(with larger ice particles) it is likely a result of a larger abundance of those small
particles.
One result of this shift toward smaller, more numerous ice particles is an in-
crease in anvil lifetime. Figure 2.14 contains four snapshots of GOES-15 IR bright-
ness temperatures starting at 20:30 UTC on 5 August (a) and progressing every five
hours until 11:30 UTC on 6 August. Similar to the results of (Lindsey and Fromm,
51
2008), the GSL pyroCb anvil (blue arrow) remains detectable with cold brightness
temperatures for much longer than the Alberta Cb (red arrow) by several hours.
The pyroCb anvil actually has brightness temperature values well below the ?40
?C threshold (Fromm et al., 2010) for at least 24 hours, at which point it becomes
indistinguishable from nearby meteorological cirrus. On the other hand, the Cb
anvil dissipates almost completely within the 12 hours depicted in Figure 2.14. A
25-hour animation of this GOES-15 imagery beginning at 19:30 UTC on 5 August
is provided as Supporting Information (Movie S1) in Kablick et al. (2018). It is
possible that upper-level descending air from a synoptic ridge contributed to the Cb
anvil dissipating quicker than the pyroCb anvil, but the magnitude of this effect is
unknown. However, in the next section it is shown that the ice within the pyroCb
anvil remained detectable for greater than five days post-detrainment, which is sev-
eral days longer than mid-latitude meteorological Cb anvils (Lindsey and Fromm,
2008).
2.3 Detrained Stratospheric Plume Properties
The GSL pyroCb plume was observed multiple times by A-Train instruments
over the subsequent two weeks following 5 August. In particular, CALIOP and
MLS were able to capture the particle and gaseous constituents, respectively. One
of the objectives of this study is to quantify what effect this plume had on the
downstream water VMR concentrations. To that end, a climatology of water VMR
is computed using all available Level 2, Version 4 MLS H2O data from the years
52
a) b)
c) d)
Figure 2.14: 5-hourly GOES-15 IR brightness temperature imagery starting at 20:30
UTC on 08/05 comparing the anvil lifecycle of the GSL pyroCb (blue arrows) and
Alberta Cb (red arrows).
2005?2014 (Livesey et al., 2018) for the months of June-August.
MLS H2O data is retrieved on pressure surfaces, and is considered scientifically
valid for pressures ? 316 hPa. Because the focus is on the UTLS, the MLS Temper-
ature product is used to convert these pressures to potential temperatures (?), and
then average the vertically-resolved H2O profiles on isentropic levels. All data from
this time period is then averaged on a 10??5? (longitude?latitude) grid. Individual
plume anomalies are then computed by subtracting the three-dimensional climatol-
ogy from observed values in each MLS profile. The MLS GPH product (geopotential
53
height, Schwartz et al., 2008) is used to approximate the altitude above mean sea
level (a.m.s.l.) of the vapor anomalies for co-location with the CALIOP backscatter
profiles. This step ensures that the analyzed vapor concentrations are spatiotempo-
rally concomitant with any aerosol/ice layer identified using CALIOP.
Figure 2.15 shows the path the plume followed, and the individual observations
made by CALIOP and MLS over the two weeks post-UTLS injection. The map in
the upper panel has contains a Hybrid Single-Particle Lagrangian Integrated Trajec-
tory model (HYSPLIT) forward trajectory (Stein et al., 2015) initialized from the
first A-Train observation segment (denoted with an ?A?) on 5 August. Additional
observations are noted along the trajectory with eight orbit segment nadir-tracks
denoted with letters B?I.
The plume made its way eastward over the Atlantic during the week following
injection, and then it made a cyclonic loop around western Europe between 10-15
August as the winds in the UTLS were being driven by a baroclinic storm. Over
the following days it moved over northern Asia as the CALIOP backscatter signal
weakened, until it became undetectable after 20 August. These detections were
corroborated using the nearest HYSPLIT hour that matched the orbit segment (red
circles along the trajectory). MLS water VMR was synchronized with CALIOP
backscatter, and the VMR anomaly was computed using the climatology at that
location.
The nine panels below the map in Figure 2.15 correspond with the A-I obser-
vations locations. Each of these panels contains the ?vertical feature mask? (VFM)
product derived from CALIOP ??532, the MLS water vapor anomaly in percent (open
54
A B C
D E F
G H I
Figure 2.15: (Top) Map of HYSPLIT trajectory showing intersections with relevant
A-Train orbits. Segments are labeled A?I, and the date, segment time stamp and
corresponding HYSPLIT hour are given in the inset table. (A?I) CALIOP VFM
curtains showing the ?stratospheric features? in black with MLS water vapor anoma-
lies overlaid as colored rectangles (% relative to local climatology). Black dashed
line is 2.5 PVU dynamical tropopause and green contours are potential temperature
(K) from ERA-I reanalysis. Abscissae of A?I are in degrees of latitude, and orange
vertical lines bound the curtain profiles used in Figures 2.17, 2.18. Corresponding
??532 curtains are shown in Figure 2.16.
55
A B C
D E F
G H I
Figure 2.16: As in Figure 2.15, but showing the ??532 instead of ?stratospheric
features? identified using the VFM product. Note the altitude range is extended
compared with 2.15 to show more of the backscattering differences between strongly
attenuating clouds in the troposphere with the stratospheric pyroCb plume.
squares plotted to match the limb-sounding measurement volume geometry; color
bar below each panel), the 2.5 PVU isosurface used to approximate the dynamical
tropopause location (white dashed line), and the ? = 360, 380 and 400 K isentropes
(green lines). Note the good correspondence between the positive VMR anomalies
and the plume locations in the stratosphere at each observation (segments B-I) after
the primary injection from the active pyroCb (segment A). The nighttime observa-
tions by CALIOP (segments D, E, F, and H) contain less noise, and the backscatter
plume is more readily discernible. Many of the MLS observations contain anomalous
VMR values > 80% surrounding the CALIOP plume locations. Figure 2.16 shows
the native CALIOP ??532 used in determining the VFM areas in Figure 2.15.
Individual profiles of the VMR values corresponding with these plume inter-
56
sections are shown in Figure 2.17, plotted here using potential temperature as the
vertical coordinate. Approximate altitude from the GPH product is shown on the
right-side ordinate axis. The red lines are the absolute magnitude (not the anomaly)
of the VMR observation, and the blue lines represent the climatology values that
correspond with the plume-portion of the orbit segment shown in the Figure 2.15
map.
To help orient the viewer, the individual curtains shown in Figure 2.15A?I
span a larger latitude range than the plume width denoted on the Figure 2.17A?I
panels. Horizontal gray dashed lines mark the 360 and 400 K isentropic boundaries
to which the plume was confined during its observable lifetime. Note the increase
in VMR at ? = 390 K ranging from 1 to 5 ppmv greater than the climatology as
the plume ages and undergoes the cyclonic movement in segments D?G. In the two
final segments (H, I), individual VMR anomalies are still as large as 2 ppmv above
climatological values.
The pyroCb injected a large abundance of small ice particles combined with
smoke that underwent diffusion over the subsequent days. To test the hypothesis of
ice sublimation increasing down-stream absolute humidity within the plume these
VMR anomalies are combined with MLS retrievals of ice water content (IWC).
Figure 2.18 shows a time series of the MLS VMR anomalies (black line) along with
corresponding MLS IWC observations (blue line) and CALIOP depolarization ratio
(?532, red line). Vertical gray dashed lines denote the times of each orbit segment.
The MLS values plotted in Figure 2.18 are averaged between 360?400 K for
the plume locations within each orbit segment, and the CALIOP depolarization
57
A B C
D E F
G H I
Figure 2.17: MLS water vapor profiles through the pyroCb plume (red lines) ob-
servations shown in Figure 2.15. Blue lines are the profiles of background WVMR
from the 2005?2014 JJA gridded climatology that fall within the region bounded
by orange lines in panels A?I in Figure 2.13, blue shading shows the standard de-
viation of these background profiles. The gray dashed lines delineate the potential
temperature boundaries of the plume over the course of its observable lifetime.
ratio values represent an averaged value corresponding to all backscatter pixels that
fall within to the MLS measurement volume. The initial A-Train observation (A;
hour 0) has an IWC value of 8.3 mg m?3, and then over the next 50 hours reduces
to 1 mg m?3, at which point the VMR anomaly has become positive. Over the
subsequent 100 hours, IWC diminishes to 0 mg m?3 as the VMR anomaly increases
to a plume-averaged value of ? 2 ppmv (45%), and remains large for the next three
58
Figure 2.18: Mean values of MLS WVMR anomalies and IWC (black and blue lines,
respectively) and median CALIOP ?532 (red lines) over the two-week observable
lifetime of the GSL pyroCb plume. Each MLS point in this time series represents
a per-profile vertical mean between 360?400 K for measurements encompassing the
plume (dashed lines in Figure 2.14), and each CALIOP point is the median value
for the entire ?stratospheric feature? between 360?400 K.
observations (segments E?G). After the plume has aged 350 hours, there are no more
detections made by the A-Train that can confidently be attributed to the pyroCb
source.
CALIOP depolarization ratio is examined as an additional test for ice presence.
On a pixel-by-pixel basis, values of ?532 should approach ? 50% when significant
amounts of ice are present (Hu et al., 2009). However as shown in Figure 2.18,
the average values remain less than 2% at all observation times. These values are
artificially small because of the large number of insignificantly depolarizing pixels
59
that fall within the MLS measurement volume included in the average. Therefore,
the ?532 values shown in Figure 2.18 should be viewed as a qualitative representation
of the combined depolarization from all particles (and molecules) that fall within
the MLS retrieval volume.
As was shown in Section 2.2, the pyroCb anvil is comprised of ice, but the
average ?532 of segment A is small because at that initial observation, the anvil
had a very small footprint within the MLS volume. Additionally, it is likely that
in noisy daytime observations when bright clouds are present in the troposphere
below the plume (e.g., segment G) the increase in ?532 would be affected by multiple
scattering within those clouds, so the relatively large ?532 seen in G is probably
unrealistic. When accounting for the issues with segments A and G, the behavior
of ?532 loosely resembles the MLS IWC behavior. Segments B?F and H?I show a
diminishing ?532 as the VMR initially increases and then decreases with age.
60
Chapter 3: Cloud-Aerosol Modeling of Active PyroCb Convection
The non-hydrostatic and compressible Advanced Research Weather Research
and Forecasting (ARW) model (Skamarock et al., 2005) is used to estimate the
sensitivity of the GSL pyroCb to specific variables. The model is run for four
conditions. First, a control pyroCb simulation is established using the geography
and weather forcing of the GSL pyroCb event (Full Simulation). The remaining three
model runs are repeats of the Full Simulation with individual alterations to initial
conditions. The second simulation has aerosol concentrations over the fire reduced
to background values (Low Aerosol), the third has reduced moisture advection in
the PBL and free troposphere (Low Moisture), and the fourth has surface sensible
and latent heat fluxes reduced to background values (Low Heat Flux). The purpose
of these simulations is to test the effect of surface CCN concentrations, moisture
entrainment, and surface heating on pyroCb cloud properties.
3.1 Model Experiments
The simulations are run for a 24-hour period starting at 12:00 UTC on 5
August and ending at 12:00 UTC on 6 August. A fifth-order monotonic advection
scheme is used for the advection of cloud variables (Wang et al., 2009), and radiation
61
is handled with the one-dimensional Rapid Radiative Transfer Model (Fouquart and
Bonnel , 1980; Mlawer et al., 1997). To capture the mesoscale structure of the py-
roCb, and to resolve cloud processes, the model horizontal domain is set to 200?200
km2 using a 500 m resolution and a domain depth of 20 km at 200 m resolution. A
bulk double-moment microphysics parameterization is used that emulates a bin mi-
crophysics scheme for the calculation of collection and sedimentation processes. This
parameterization is generally referred to as the bin-bulk scheme and was first im-
plemented into Regional Atmospheric Modeling System (RAMS) at Colorado State
University (Walko et al., 1995; Saleeby and Cotton, 2008).
For the Full Simulation, the surface sensible and latent heat fluxes are set to
150 and 310 W m?2, respectively in the non-fire regions of the domain. A hot-spot
representing the fire is modeled as a circle centered in the domain with diame-
ter=20 km, and sensible and latent heat fluxes are set to 15000 and 1800 W m?2,
respectively. These values are based on the previous modeling studies by Trentmann
et al. (2006) and Luderer et al. (2006). Beringer et al. (2003) noted that sensible
heat fluxes are much greater than latent heat fluxes over fire (large Bowen Ratio),
and Trentmann et al. (2006) noted that there exists a positive feedback between
these fluxes due to increased entrainment of low level moisture as a fire emits more
sensible heat.
In the present model setup, this interaction is ignored, and all heat values
are prescribed without any meteorological coupling. Therefore, the results here are
not designed to mimic reality with respect to fire-atmosphere interaction, but are
idealized so that sensitivity of convective properties between model runs can be
62
more readily compared. Directly over the fire, the aerosol concentration is set to
15000 cm?3 within the PBL, and decreases exponentially with height in the free
troposphere. Within the PBL at non-fire locations, the concentration is set at 150
cm?3, also decreasing exponentially with height above this level.
A big question regarding pyroCb development is the individual effects of sur-
face heat fluxes, fire-produced aerosol particles, and moisture entrainment illustrated
by Peterson et al. (2017b). Toward the goals of assessing convective intensity, prop-
erties of detrained cirrus, and impacts on UTLS moisture, it is useful to re-run
the control simulation with high heat flux/low aerosol and with low heat flux/high
aerosol. To estimate the effect of the former, the Low Aerosol simulation is done by
repeating the Full Simulation with PBL aerosol concentrations reduced to the back-
ground level (150 cm?3). Then, to estimate the role played by surface heat fluxes,
the Low Heat Flux run maintains the large aerosol concentrations of the Full Sim-
ulation, but has reduced surface latent and sensible heat fluxes to the background
values (310 and 150 W m?2, respectively) in the within fire spot.
However for these conditions, it is found that surface latent and sensible heat
fluxes in the fire spot are too low to form a cloud, so a potential temperature
perturbation is prescribed over the fire spot to trigger a cloud, following Weisman
and Klemp (1982). The horizontal extent of this perturbation is identical to that of
the fire spot and the vertical extent is 2.8 km, and the maximum perturbation is 1.8
K. This perturbation has been used by numerous previous studies and considered to
have negligible influences on cloud development, although it triggers the formation
of a cloud.
63
As discussed in Chapter 2, the GSL pyroCb occurred during a substantial ad-
vection of moisture in the mid-troposphere. NARR data also showed an immediate
increase in surface relative humidity at the time of convection (20?30%? 50?60%).
It is possible this advection contributed a significant amount of moisture to the
atmosphere over the fire?and therefore instability?just before pyroCb formed, or
that it was entrained during convection.
To test whether these moisture sources had any impact on convection, the Full
Simulation run is repeated again by reducing the level of moisture advection by a
factor of 5 in the PBL and free troposphere throughout the simulation period (Low
Moisture run). Since the simulations are started two hours prior to the formation
of the pyroCb, this reduction is applied to the period before and after convection
starts to limit entrainment from both the surface through the cloud base and the
free atmosphere through the sides of congestus.
3.2 Results
In Figure 2.1, the pyroCb anvil is observed to advect to the northeast of
the fire spot due to the southwesterly winds at the outflow altitude. In the Full
Simulation, a convective column forms over the fire spot, and is accurately advected
northeastward as the GSL pyroCb does in reality. Figure 3.1 shows the field of
cloud-ice mixing ratio (representing the outflow anvil) at the top of the simulated
pyroCb and at a time that corresponds to the satellite image in Figure 2.1f. The
modeled anvil cirrus cloud is ? 80 km in diameter and is in fairly good agreement
64
(g m-3)
616.61. 6N N 0..330
0..2255
0..220
606.30. 3N N 0..1155
00..110
0..0055
595.09. 0N N
111188..55 W W 111515.5.5  WW 111122..55 W W
Figure 3.1: ARW Full Simulation ice particle mixing ratio within the cirrus out-
flow at the equilibrium level (? 13 km) during the mature stage of pyroconvection
(approximately 3 hours after convection began). Red circle denotes fire hot-spot lo-
cation used to initialize the simulation. Compare with VIIRS image in Figure 2.1f.
with the VIIRS observation, giving confidence in the ability of the ARW setup to
reproduce the morphology.
Among the observed CloudSat 94 GHz radar profiles shown in Figure 2.13 is a
simulated 94 GHz radar reflectivity profile from the Full Simulation modeled pyroCb
(shown as green line). This profile is shown plotted above the cloud base, and the
simulation time corresponds with the CloudSat overpass time of 20:20 UTC, which
is ? 2 hours after the simulated convection began. The shape of the Full Simulation
radar profile reasonably matches the shape of the GSL pyroCb CloudSat-observed
profile, indicating the model represents the high cloud base reasonably well.
As mentioned previously, heat fluxes over fires can produce extremely large
65
a) b)
Figure 3.2: Profiles from four pyroCb model simulations showing the averaged ver-
tical distributions of the a) spatiotemporally-averaged updraft mass fluxes, and b)
ice mass densities. See text for descriptions of the four simulations.
updrafts from enhanced buoyancy, but increased CCN concentrations are also known
to enhance updrafts from latent heat release. The results of Luderer et al. (2006)
showed heat fluxes dominate the updraft velocities over aerosol influences, and that
result is confirmed here. Figure 3.2a shows the vertical distributions of the time-
and domain-averaged updraft mass flux?a standard representation of the cloud
dynamic intensity?for all four model runs. The black line is the Full Simulation,
which has a peak value of 3.2 kg m?2 s?1 at the top of the PBL.
The bump in the updrafts between 8?11 km is due to a buoyancy push from
the latent heat of freezing. The Low Aerosol run (blue line) closely follows the Full
Simulation profile throughout the depth of the troposphere with minimal deviation.
The impact of limiting moisture entrainment (green line) is slightly stronger than
limiting aerosol concentrations. Although this profile is similar to the Full Simula-
66
tion mass fluxes, the model shows a greater reduction in updrafts, especially between
7?9 km where mass fluxes are 10?15% less than the Low Aerosol run. This implies
that any invigoration from an aerosol increase is likely dependent on the availability
of moisture. However, the most significant negative impact on the strong updraft
mass fluxes demonstrated in the Full Simulation comes from a reduction in surface
heat fluxes (red line), which is ? 4 times smaller than the other runs. Limiting sen-
sible and latent heating from the fire down to background values produces a peak
updraft mass flux of 1.1 kg m?2 s?1 and a upper tropospheric peak of ? 0.3 kg m?2
s?1. Among the variables tested here, this result confirms fire-induced surface heat
fluxes play the most important role in pyroCb intensity.
The different model runs produce comparable differences in cloud-ice mass
densities (Figure 3.2b). Similar to the updraft mass fluxes, the fire-induced surface
heat fluxes play the most important role in the amount of cloud ice particularly
just below the tropopause. The ice mass density peaks ? 12.8 g m?3 at the EL for
the Full Simulation with minimal reductions in the Low Aerosol and Low Moisture
runs (? 11.5 and ? 10.8 g m?3, respectively). Of the various ways that surface heat
fluxes, moisture entrainment and aerosol concentrations may influence ice cloud
properties, the common mechanism is updraft buoyancy. These ice mass densities
are not surprising given the updrafts shown in Figure 3.2a.
The additional mechanism to consider is an increased number density of ice
particles caused by an increased droplet number density from CCN nucleation. Fig-
ure 3.2b shows that this mechanism does not seem to have much influence on the
ice mass density because the Low Aerosol result does not deviate significantly from
67
Figure 3.3: Profiles of WVMR spatiotemporally averaged for each simulation for all
cloud regions within the domain (solid lines), and surviving water post-detrainment
(dashed lines), estimated here to be 30%. The latter includes vapor contributions
from ice sublimation and existing absolute humidity. The MLS observation during
the active pyroCb?when no WVMR enhancement is yet present?is shown as a
gray line as a reference to the background profile.
the Full Simulation result. It is possible the microphysics scheme is unable to sub-
stantially account for changes in surface aerosol concentrations, but it is more likely
that the updraft velocity is overwhelmingly more important on determining the ice
mass density. Therefore, the results indicate that increased CCN concentrations
have much less importance on the ice distribution than the surface heat flux.
Vertical distributions of the averaged water vapor mixing ratio from all four
simulations are shown in Figure 3.3, plotted above the model tropopause (? 13
km). These values (solid lines) are horizontally averaged in the non-cloud areas,
which are defined as columns with a zero liquid and ice water path. Again, there
68
are differences between vapor detrainment values from the Full Simulation and the
other three runs, but the Low Heat Flux run has the most significant deviation.
For example, at 14 km the Full Simulation has a clear-sky average VMR=13 ppmv,
the Low Aerosol run has 12 ppmv, the Low Moisture run has 9 ppmv and the Low
Heat Flux has < 2 ppmv. These values, however, are instantaneous values after
the pyroCb ceases, and do not account for sublimation from the aging ice cloud
detrained at these altitudes.
Because MLS observations showed the GSL pyroCb plume absolute vapor
concentrations peak between 8?9 ppmv after one week, it appears there should
be approximately a 30% survival of total water from both detrained vapor and
sublimated ice in order to reach these values. This 30% survival amount is applied
to each model run and shown as dashed lines in Figure 3.3. The cloud ice mass
density from each simulation is converted to parts per million and is combined with
the molecular water vapor and multiplied by 0.3. At 14 km, the value from the Full
Simulation is between 7?8 ppmv, the Low Aerosol is ? 7 ppmv, the Low Moisture
is ? 6 ppmv, and the Low Heat Flux is ? 3 ppmv.
69
Chapter 4: Diabatic-Lofting of Injected PyroCb Plumes
The physics behind plume self-lofting are straight forward (Crutzen and Birks ,
1982; Radke et al., 1990). The basic idea is that a sufficiently optically-dark aerosol
pall absorbs shortwave electromagnetic radiation and undergoes diabatic heating
within a volume of air that otherwise would not have such a heating anomaly. This
diabatic heating increases the temperature within the air volume and generates
buoyancy. Upward motion of the air volume continues until it reaches thermal
equilibrium, either from reaching an altitude at which ambient temperature causes
radiative equilibrium or from turbulent diffusion of the aerosol reducing the volume
absorption coefficient.
This radiatively-induced self-lofting of dark aerosol plumes was studied exten-
sively throughout the 1980s when the so-called Nuclear Winter phenomenon gained
traction during the Cold War (Crutzen and Birks , 1982; Turco et al., 1983; Sagan
et al., 1983; Thompson et al., 1984; Turco et al., 1984; Thompson and Schneider ,
1986; Turco et al., 1990). During that time, pyroCb had not yet been discovered,
and the smoke that made it into the stratosphere in the Nuclear Winter model-
ing scenarios was based on a best-estimate of the density and optical properties of
tropospheric plumes that would emanate from urban conflagration (typically not
70
from the direct stratospheric injection of plumes from detonation columns of bombs
themselves).
The studies relied on extremely dense smoke plumes and highly absorptive
optical properties to get the smoke to gradually make its way into the stratosphere
through diabatic self-lofting, overcoming the stability of the tropopause. More re-
cent Nuclear Winter research has relied on the same hypothetical smoke condi-
tions (Robock et al., 2007a,b; Toon et al., 2007). It should be noted that several of
the Nuclear Winter studies referenced the World War II bombing campaign in both
Europe and Japan and the resulting firestorms as an event that could be scaled-
up to replicate a Nuclear Winter precursor. No convincing evidence has ever been
published showing that the Dresden Firestorm in Germany, the intensive sustained
bombing campaigns in Japan, nor the Hiroshima and Nagasaki nuclear detonations
had any impact on stratospheric radiative budgets or surface climate (Robock and
Zambri , 2018).
PyroCb events, on the other hand, are known to directly inject smoke into the
stratosphere because of the impact that the surface heat fluxes have on convective
updraft velocities (Luderer et al., 2006; Trentmann et al., 2006, and Chapters 2, 3
herein). It remains unknown whether the Nuclear Winter scenarios and their pro-
posed plume evolutions are valid because such a high-yield event thankfully has
never occurred, but based on peer-reviewed pyroCb research both modeling and
observational plumes from massive wildfires are stratospherically-injected through
deep convective processes and not through slow ascent across the tropopause by
radiatively-induced self-lofting.
71
However, once the pyroCb plume is in the LS, self-lofting can be observed.
Previous research has attempted to associate pyroCb plumes with a self-lofting
mechanism, specifically with the 2009 Black Saturday plume (Laat et al., 2012). Un-
fortunately that study misinterpreted the maximum height of the active convection
columns from several pyroCbs on Black Saturday, and assumed a high rate of radia-
tive heating in order to put the plume into the stratosphere. Ground-based radar
observations show that the conclusions of Laat et al. (2012) were flawed because a
majority of the plume was directly injected into the LS prior to self-lofting (Dowdy
et al., 2017).
This chapter explores pyroCb plume diabatic self-lofting. The focus is on
the extreme 2017 PNE case, but comparisons are made with the Black Saturday
plume and the comparatively much weaker 2014 GSL pyroCb. LS observations are
examined in the days and months that followed these events, and a diabatic model
framework is developed using solar radiative anomalies from extinction observations
and an empirical heat accumulation model. Prior to that analysis, it is insightful to
discuss the amount of aerosol stratospherically-injected. Examining the very large
amount mass of aerosol injected during the 2017 PNE will put the diabatic-lofting
discussion in better context.
4.1 Stratospheric Injection Mass
As was demonstrated in Peterson et al. (2018), the 2017 PNE plume was un-
precedented in the amount of pyroCb aerosol directly injected into the stratosphere.
72
The aerosol mass loading was estimated to be 0.2?0.1 Tg based on a combined pas-
sive and active remote sensing methodology, using CALIOP lidar and OMPS Nadir
Mapper (NM) instrument. The estimation leveraged the fact that OMPS UVAI
(and products of similar instruments like OMI and TOMS) is larger when either the
plume optical thickness and/or altitude is increased, and that for sufficiently thick
UTLS plumes there is an empirical threshold of UVAI that delineates stratospheric
vs. tropospheric presence (Torres et al., 2002). In other words, when a plume is
optically opaque, physically deep and has layers above and below the tropopause,
there exists a UVAI value that would determine which pixels are in the stratosphere
and which pixels are in the troposphere.
This methodology is made possible by a fortuitous combination of earth-
observing satellite instruments making nearly simultaneous measurements in sun-
synchronous orbits. For example, the OMPS NM on-board Suomi NPP collects a
wide swath of imagery in the afternoon ascending orbital node that was typically
within a 1?2 hr of vertical profiles from the CALIOP lidar (at the time of this
writing, CALIPSO has now been moved out of the A-Train to a lower orbit with
CloudSat, and the orbital characteristics cause a periodic overlap with NPP that
will continuously change with correction maneuvers). The rest of this section briefly
details the main results of Peterson et al. (2018)?that a pyroCb injected an aerosol
plume on the scale of a small volcano?and provides the methodology-validation
that the present author undertook as a part of that study.
The total mass of aerosol injected into the stratosphere is defined as the prod-
uct of the mass density (M?) and the stratospheric plume volume (V ). Mass density
73
of the aerosol plume is defined as
??532SM? = , (4.1)
?
where the S is the extinction-to-backscatter ratio (colloquially referred to as the
?lidar ratio?), and ? is the mass extinction coefficient. Vaughan et al. (2005) state
in the CALIOP Level 1 algorithm theoretical basis document (ATBD) that S is
an empirical parameter ranging between 40?60 sr?1 for smoke plumes, depending
on age, and ? 70 sr?1 for volcanic sulfate. The lidar observation (Figure 4.1)
used for determining aerosol mass is within 48 hours of the plume injection, and
was shown in Peterson et al. (2018) to be mostly free of clouds and ice. With this
assumption of a smoke-only plume, all computations presented herein utilize S = 60
sr?1 and a single scatter albedo ?0 = 0.9, which are appropriate for aging boreal
fire plumes (Reid et al., 2005). For this case, a range of ? between 3.0?6.0 m2
g?1 (Nikonovas et al., 2017) is used for the final M? estimate.
To estimate V (= plume area ? thickness), each component are determined
individually from different instruments. Stratospheric area is determined by com-
puting the area subtended by all OMPS pixels containing UVAI values above the
threshold, and thickness is determined from the depth of stratospheric component
of the plume in concurrent CALIOP lidar profiles. A UVAI threshold = 15 has
been shown to be an accurate cutoff for stratospheric pyroCb plumes (Fromm et al.,
2008b), and is used here. Using this cutoff produces a plume area approximately
8?105 km2, and the lidar profiles resolved a plume between 1.0?2.0 km thick, giving
a stratospheric V ' 1.475?1015 km3. CALIOP ??532 are input into Equation 4.1,
74
and profiles of M? are averaged, resulting in total mass estimate between 0.1?0.3
Tg.
Figure 1 from Peterson et al. (2018) shows how this value compares to the
2008 Kasatochi eruption and the large Chisholm pyroCb event. Kasatochi?s 2008
eruption is considered a moderate volcanic event with a Volcanic Explosivity Index
(VEI) of 4, and had an estimated initial aerosol mass between 0.2?0.5 Tg ignoring
secondary particle formation. The Chisholm pyroCb was estimated to have a mass
between 0.018?0.109 Tg (Fromm et al., 2008b), depending on the choice of optical
properties and using an UVAI threshold = 15.
The first reliable CALIOP overpass of the PNE plume occurred on 14 August
2017 at approximately 11:30 UTC, and is shown in Figure 4.1. Previous overpasses
on 12 and 13 August had ambiguous ??532 and ?532 properties that reduced confidence
in the pyroCb source attribution. The majority of the plume is situated between 12
and 13 km in altitude with strong backscatter and full attenuation of the lidar. The
overlaid, open boxes show the simultaneous MLS CO retrievals (similar to the water
vapor anomalies shown in Figures 2.15 and 2.16). The maximum CO value for this
scene (? 678 ppbv; white ???) occurs near the highest plume altitude near 70.5?N.
Columns labeled ?1? and ?2? are used in the radiative transfer simulations detailed
below. Because Figure 4.1 is a nighttime observation, the closest NPP overpass
does not provide an OMPS UVAI; the leftmost panel in Figure 2.3d (14 August
19:12 UTC) is the best representation of a UVAI measurement corresponding to
this CALIOP curtain.
Figure 4.2 shows columns 1 and 2 from Figure 4.1 in terms of their average
75
1 2 75
60
45
30
15
a) b) 120 90
Figure 4.1: The first reliable CALIOP ??532 intersection with PNE stratospheric
plume on 14 August (?1.5 days after injection) at 11:30 UTC. Overlaid are MLS
observations of CO. Maximum CO mixing ratio observation is > 670 ppmv (white
??? near 70.5 ?N). Columns of light vertical shading labeled 1 and 2 are loca-
tions used to assess total aerosol extinction for comparison with OMPS UVAI (see
Figures 4.2 and 4.3).
scattering ratio, and the determined particle extinction profiles. The vertically-
dependent threshold recommended by the official CALIOP ?feature finding algo-
rithm? (Winker et al., 2009) is shown as the discontinuous blue line in (a, c). The
scattering ratios (green line) greater than this threshold are considered ?features?
(i.e., non-molecular), and are used to determine particle extinction profiles (b, d)
and stratospheric AOD using the following steps: i) aerosol feature layers are iden-
tified using the scattering ratio threshold, ii) ??532 values within the identified layers
are multiplied by S to estimate particle-only optical extinction coefficients, and iii)
vertical integration of these coefficients in the stratosphere determines stratospheric
AOD. For this case the tropopause as determined by the nearby 12:00 UTC Norman
76
a) column 1: 70.1? - 71.0? N b) column 1: 70.1? - 71.0? N
scattering ratio
non-molecular threshold
zero scattering 
c) column 2: 64.5? - 65.0? N d) column 2: 64.5? - 65.0? N
scattering ratio
non-molecular threshold
zero scattering 
Figure 4.2: Profiles of scattering ratio (left) and derived non-molecular extinction
(aerosol particulates, right) from the PNE plume intersection in Figure 4.1. (a, b)
correspond with column 1 and (c, d) with column 2. Vertical integration of plume
extinction for tropopause height >11.5 km yields AOD ' 3.5 and' 1.5, respectively.
77
Wells radiosonde is 11.5 km, which is approximately ? ' 370 K. This agrees with
the initial injection potential temperature determined from surface weather radar
at Prince George, British Columbia (discussed in the next section).
Radiative transfer simulations using the Santa Barbara DISORT Radiative
Transfer (SBDART; Ricchiazzi et al., 1998) model are used to give confidence in
the mass estimation algorithm. The validity of UVAI threshold (= 15) is tested
with two channel ultraviolet simulations (? = 330 and 390 nm) using the CALIOP
stratospheric extinction profiles shown in Figure 4.2. Figure 4.3 shows the strato-
spheric AOD (a) and simulated UVAI from SBDART (b) along the CALIOP orbit
track (red orbit segment on Figure 4.1b). According to the CALIOP ATBD, the
lidar typically is fully attenuated by optical thicknesses > 3, denoted by the red line
across Figure 4.3a. There is a large section of the plume that is consistently at or
above this value between 66?73?N and is at an altitude high enough to create an
UVAI greater than the 15 threshold (red line on panel (b)).
Comparing this independent UVAI estimation with OMPS shows that the
known stratospheric components of the plume (as determined with CALIOP) are
well thresholded using UVAI = 15. In other words, computing the area of the
plume using this threshold does indeed subset the UVAI field into a stratospheric-
only component. Therefore, stratospheric mass estimates for this pyroCb plume
using this methodology are more likely to be an underestimate than an overestimate
because the lidar is fully attenuated along thicker sections of the plume and result
in an underestimate of plume depth.
78
a)
b)
Figure 4.3: Empirical results of mass estimation of the PNE plume shown in Fig-
ures 4.1 and 4.2 showing the a) along-track AOD derivation and b) simulated UVAI
using the 330 nm and 390 nm channels. Red lines show a) the value of ?total optical
attenuation? of the downward-looking vertical CALIOP instrument (note a majority
of the plume between 66-73 ?N has fully attenuated the lidar) and b) stratospheric
UVAI threshold. There is good agreement with the plume?s simulated UVAI ? 15
and the stratospheric component of the CALIOP observation in 4.1.
79
4.2 Plume Rise Observations and Simulations
4.2.1 Satellite Observations
This section details the dramatic diabatic rise of the 2017 PNE plume by
combining satellite observations with radiative transfer modeling. The plume is
tracked for several months after the 12 August pyroCbs occurred using dual MLS
thresholds of concurrent H2O and CO, and confirmed with CALIOP extinction
observations. Figure 4.4 shows the plume on 3 September about three weeks after
the pyroCbs injected it. The white dashed line is the maximum injection height
observed on 12 August at ? 11.5 km. The white ??? once again shows the location
of the maximum MLS CO for this particular scene (? 170 ppbv), and at this time is
around z = 22 km where CALIOP ??532 confirms the plume?s presence. Computing a
velocity from this observation and the injection height on 12 August puts the average
rate of rise over the first three weeks at roughly 400 m day?1, an extraordinary rate
never-before observed with a pyroCb plume.
Anomalously large stratospheric mixing ratios of both H2O and CO are a
defining characteristic of this event, and the reliable, global observations of Aura
MLS make tracking this plume straightforward. This instrument is used to per-
form a plume-tracking analysis (as opposed to analyzing spatiotemporal averages of
MLS species) for the following two reasons. First, the best way to track stratospheric
aerosol would be with the high-resolution CALIOP lidar. Other instruments such as
limb-profilers would work, but CALIOP has reasonably good spatiotemporal cover-
80
injection Zmax
35 38 41 44 47
a) b)
Figure 4.4: CALIOP/MLS observation of the PNE plume on 3 September 2017 at
21:00 UTC (? 3 weeks post-injection). Location of CO maximum (? 170 ppbv) and
large ??532 values show plume has risen at least 10 km above the injection height
(white dashed line).
age, and has high vertical and horizontal resolution unmatched by solar occultation
or limb-scatter instruments for stratospheric aerosol (Fromm et al., 2014). However,
CALIOP suffered a > 1 week data gap during the month of September 2017, and
connecting observations with HYSPLIT trajectories (as was done in Chapter 2) is
problematic over such a long time without observations to validate them. Also,
HYSPLIT does not account for diabatic heating of plumes, so forecasting a track
position can lead to significant errors when self-lofting may be involved. The con-
tinuous monitoring by MLS alleviates this problem. Second, a goal of this research
is to estimate the radiative forcing impact of stratospheric pyroCb plumes, and
water vapor anomalies need to be quantified in order to accomplish this goal (see
Chapter 1). Thus, in addition being the primary tracking method of the plume,
MLS observations are used to suss out individual species anomalies and estimate a
?real-time? radiative forcing of the plume (Chapter 5).
81
To track the plume, a dual threshold of MLS H2O ? 7 ppmv with a concurrent
CO ? 70 ppbv is applied to all MLS observations at isentropic levels 600 ? ? ? 380
K. After 25 September the CO threshold is lowered to 40 ppbv, and after 10 October
it is lowered to 30 ppbv. On 19 November both thresholds are lowered: H2O ? 6
ppmv and CO ? 20 ppbv. Because MLS retrieves species mixing ratios on pressure
surfaces, the Level 2 Temperature product is used to compute ?. Figure 4.5 shows
four time steps in this tracking process. Panel (a) starts one month post-event
on 12 September, at which time a large section of the pyroCb plume (shown in
Figure 4.4) separated into three distinct and trackable sections over central Asia.
From this point onward in the study?including Chapter 5?these three sections
are referred to as the ?Triangle,? ?Circle,? and ?Square? plumes for convenience
and ease of identification. The trajectory shown in Figure 4.5a is an approximation
of the actual path the plume took over the course of one month post-stratospheric
injection. It was tracked during this time using the MLS threshold mentioned herein,
but the individual dates between 12 August?12 September are omitted for space.
Note that typical zonal mean wind velocity in the LS at these latitudes is ? 15 m
s?1 (Wallace, 2003), which agrees with the observed wind velocity ? 16 m s?1.
In Figure 4.5b?d, these three sections are tracked by connecting the MLS
observations of points that exceed the dual threshold criteria (green trajectory;
Triangle plume, blue; Circle plume, and red; Square plume). The colored dots on
the maps are locations where MLS observations exceed the dual thresholds. These
dots are colored according to their altitude as determined with the geopotential
height (GPH) product. The last observable day of each plume is shown in the
82
a) 9/12/2017 - plume separation 
1 month
b) 10/11/2017 - last observation of ?Triangle? plume (green)
c) 10/14/2017 - last observation of ?Circle? plume (blue)
d) 11/25/2017 - last observation of ?Square? plume (red)
Figure 4.5: Hand-drawn trajectories of three sections of the 2017 PNE plume tracked
using the dual MLS H2O/CO threshold. (Continued on next page)
83
Figure 4.5: (continued) Panel (a) shows the start date (12 September) when the
plume could be sufficiently identified as three independently moving sections. The
?1 Month? trajectory between the 12 August pyroCb date and the plume loca-
tions on 12 September is an approximation based on the plume tracking method
described in the text. Panels (b?d) show termination dates of each plume section?s
observability. Colored dots are the GPH of the MLS plume observation (bottom
colorbar). Inset in panel (d) focuses on the area over the mid-Atlantic where the
Square plume completed a circumnavigation; locations are marked when observa-
tions occurred between 29 September and 2 October (bearing east toward west) and
11?25 November (bearing from southeast toward northwest, then abruptly shifting
toward northeast).
bottom panels; (b) Triangle, (c) Circle, and (d) Square. Note that panel (a) shows
the start of the trajectories in (b?d). MLS and CALIOP were also used prior to 12
September, but in the interest of space, they are left off this figure. However data
from those dates are included in the self-lofting analysis below. Each of these plume
sections circumnavigated the globe at least once, and the Square plume was actually
lofted high enough in the tropical lower stratosphere to encounter the Easterlies, and
shifted course back toward the west prior to circling the globe.
Figure 4.6 combines all MLS dual threshold observations from the northern
hemisphere (NH) between 12 August and 18 November, and displays them at their
? level. The vertical dashed line is the 12 September plume separation date (Fig-
84
N.H. Potential temperature of MLS H2O ?  7 ppmv
CO ? 70 ppbv CO ? 40 ppbv CO ? 30 ppbv
Figure 4.6: Potential temperatures of all northern hemisphere MLS thresholded
observations during the first three months of the PNE plume evolution. A constant
H2O threshold of 7 ppmv is applied with diminishing CO thresholds as the plume
ages. The vertical dashed line marks the point at which the plume separates over
central Asia, and the three dashed lines emanating from this time show the different
ascent rates of three tracked plume sections.
ure 4.5a), and the three dashed lines emanating from this date represent from top-
to-bottom the approximated diabatic rise of the Square, Circle, and Triangle plume
sections, respectively. The apparent diabatic increase is greater than 200 K for this
H2O/CO thresholds, and from these observations it is apparent that the rate of
plume rise is greater in the early stages of the plume lifetime (before 12 September).
This is most likely due to aerosol diffusion as the plume ages, which would reduce
radiative heating throughout the solar daytime.
85
?? ??
?t ?0 ?t >0
Active pyroCb: Injection-? NP: strong absorption EP: diminishing absorption
Figure 4.7: Illustration of pyroCb smoke plume evolution from onset through decay.
Left: available observations are used to define the injection-? at a specific location
and time. Middle: the NP-phase is when the earliest-possible 3D observations of
the plume can be characterized, and coincides with strong diabatic-lofting potential
due to high concentrations of absorbing aerosol. Right: the EP-phase is the period
during which advection and dispersion are reducing the concentration of aerosol to
the point of non-observability, and diabatic-lofting potential is rapidly diminishing.
A schematic of this process is shown in Figure 4.7, where the early stages of
plume life are denoted as the ?nascent plume? (NP) and the latter stages as the
?evolving plume? (EP). Black arrows represent incoming solar irradiance, which
is more greatly absorbed during the NP stage than in the EP stage. The next
section will mathematically demonstrate this concept and use SBDART to model
the radiatively-forced plume rise.
86
4.2.2 Modeling Plume Rise
To simulate the lofting observed in the MLS and CALIOP observations, SB-
DART is used to calculate longwave and shortwave radiative heating rates using
the stratospheric extinction profiles. This analysis focuses on the Square section of
the 2017 PNE plume, but two additional cases are analyzed for comparison: 1) the
much smaller August 2014 GSL pyroCb, and 2) the February 2009 Australian Black
Saturday pyroCb event. Extinction coefficients at 532 nm are derived from CALIOP
??532 and S = 60 sr
?1, and plume optical depth is calculated using vertical integra-
tion for heights where ? > 380 K. SBDART is set up to calculate full-spectrum
irradiance using these optical depths and plume density profiles, using ?0 = 0.9,
and mie scattering asymmetry parameter = 0.7. The model atmosphere provided
with CALIOP data (derived from NASA?s GEOS-5) is used for temperature, hu-
midity and upper stratospheric O3 profiles (Winker et al., 2009). To calculate the
diabatic heating anomaly, model runs are performed for the same atmospheres at
each observation except with clear skies, and the resulting profiles are subtracted
from the aerosol results.
In simple terms, each CALIOP observation is used as an independent data
point in a time series of plume heating anomalies. These heating rate anomalies
are assumed to proportionally increase the temperature of the plume, are thus are
accrued to the initial UTLS-injection potential temperature (?0) to estimate the
rate of diabatic-rise. At each observation, the modeled difference between the clear
sky and the plume heating rate profiles yields the aerosol-only heating rate anomaly,
87
so the diabatic rise at the ith model time step is then estimated by accumulating all
these anomalous values (units of K day?1) from the 0th through the (i ? 1)th time
step.
The total change in plume potential temperature can be represented by the
thermodynamic energy equation:
D? g0
= ?(Q? w). (4.2)
Dt cp
Here, the coefficient ? is defined as the efficiency that the sum of internal heating
and work has on affecting ? changes within the plume, Q is the radiative heating
rate, which is simply
Q = ? 1 d (Fnet), (4.3)
?cp dz
g0 is the acceleration due to gravity, cp is the specific heat at constant pressure, ? is
the air density, Fnet is the net flux as defined in Chapter 1, and w is the isothermal
vertical velocity.
An appropriate scale assumption is that w is time-independent because of the
highly stratified environment of the LS. It is possible that for long duration events
the Brewer-Dobson uplift in the lower stratosphere could play a small roll in the
diabatic rise of plumes. However, the time-scale for plume duration would need to
be measured in years because observations have shown Brewer-Dobson velocities at
these altitudes are approximately 0.02?0.03 cm s?1 (17?25 m day?1), which means
a parcel would gain 1 km in altitude about every two months (Minschwaner et al.,
2016). For all documented pyroCb aerosol lifetimes this time scale is too long, and w
can be ignored as a term that would do work on the system. Therefore, Equation 4.2
88
can be reduced to
?? = ??Q?t, (4.4)
and the simulated potential temperature at the ith observation step of the plume
can be written as ?N
??,i = ? (n?Qi?1 + ?i?1), (4.5)
i=1
where n is the number of lapsed days between the ith and (i? 1)th CALIOP extinc-
tion observations and N is the total number of observations. Because ? is assumed
to be time-independent due to the nature of stratospheric stability, the approxi-
mated physical processes contained within ? include small impact quantities such
as diffusion, turbulence, latent heating, etc; processes that are on long times scales
in the LS. It also provides a simple way to bundle the errors in the optical property
assumptions, SBDART physics, CALIOP extinction estimations, etc.
The plume-rise model defined by Equation 4.5 is a simplification of the nature
of LS dynamics, but with the understanding that ? is an empirical parameter to
be used as a test against observed rates of plume-rise. This model is applied to
observations made during the 2017 PNE Square plume (Figure 4.8), the 2014 GSL
pyroCb plume (Figure 4.9) and the 2009 Black Saturday plume (Figure 4.10), and
the data from these cases are tabulated in Tables 4.1?4.3, respectively.
The radiative heating anomaly (?Q) is computed from SBDART by taking
the difference between the heating of the aerosol layer and a clear sky with the same
temperature and humidity profiles. Aerosol layer boundaries used in the SBDART
profile inputs are determined from the CALIOP extinction profiles used to retrieve
89
stratospheric aerosol optical depth.
Potential temperature levels of each CALIOP plume extinction observation
(at which ?Qi is computed) are not considered when calculating ??,i (the plume?s
heat accumulation summation). These observations are only used as input into the
SBDART model that produces heating rate profiles. In other words, this model time
series is ?independent? in that there is no attempt to correct a prognosticated ??,i
with observations of plume ?. The empirical ? is the only parameter by which the
model results are ?corrected? to match observations. A range of ? values have been
applied, and ? = 100%, 50%, 30% are shown in Figures 4.8?4.10, with the additional
? = 10% shown in Tables 4.1?4.3.
Figure 4.8 shows observed and modeled diabatic plume rise of the Square
plume section of the PNE case (see red trajectory in Figure 4.5). This is the highest,
and most optically thick section of the PNE plume that is trackable. It also moves
south into the tropics (? 30 ?N), and slowly moves westward in this region for about
one month between early October and early November.
In Figure 4.8a, CALIOP ??532 is used to determine the ? centroid at each
observation, and is shown as a blue ???. Blue lines extending from each CALIOP
??? denote the range of plume thickness. Light blue bars during the month of
September are periods of CALIOP data gaps. Also, replotted here (as in Figure 4.6)
are all NH MLS measurements greater than the dual H2O/CO threshold (green
???) to provide information during these CALIOP data gaps. Corresponding plume
AOD is shown in (b). The pyroCb injection ? = 370 K is shown on the plot, and
is determined from the highest surface weather radar echo top altitude from the
90
2017 PNE
a)
model observations
? = 1.0 MLS plume theta
? = 0.5
? CALIOP plume theta = 0.3
b)
Figure 4.8: (a) Observed and modeled diabatic rise of the PNE Square plume (see red
trajectory in Figure 4.5). Blue ??? show the potential temperature of the CALIOP-
observed plume centroid with error bars denoting the upper and lower backscatter
boundaries, augmented with MLS plume observations of colocated HCN/CO max-
ima (green dots). Simulation of radiatively-forced diabatic lofting for various ? are
shown over observations. Prince George radar echotop-based injection ? = 370 K is
noted. (b) The stratospheric AOD derived from CALIOP extinction.
91
Table 4.1: 2017 PNE pyroCb plume potential temperature observations (?) and
model-predicted (??) using various efficiency parameters (?). Also shown are
CALIOP stratospheric AOD and SBDART radiative heating rate anomalies (?Q)
in K day?1. n?Q is the maximum temperature change assuming a steady state over
n days between observations (K; only shown where n > 1, otherwise ?Q = n?Q).
Last row gives total change in ? and each ??. ?? values in bold are closest match to
plume ?.
date plume ? AOD ?Q (n?Q) ??=100% ??=50% ??=30% ??=10%
08/13 352 8.51 18.30 ? ? ? ?
08/14 347 6.46 22.95 371 361 358 354
08/15 361 6.62 13.02 394 372 365 356
08/16 357 3.34 33.79 407 379 369 358
08/17 388 10.65 37.62 440 396 379 361
08/18 392 10.47 36.84 478 415 390 365
08/19 415 8.76 25.55 515 434 401 369
08/20 418 3.76 13.76 540 446 409 371
08/21 438 1.75 22.64 554 453 413 372
08/22 420 3.28 30.31 577 465 419 375
08/23 432 3.91 16.35 607 480 429 378
08/24 407 2.01 30.85 623 488 434 379
08/25 437 4.25 23.88 654 503 443 383
08/26 437 2.74 23.38 678 515 450 385
08/27 448 2.65 33.07 702 527 457 387
08/28 448 3.56 19.79 735 543 467 391
08/29 458 2.03 18.34 754 553 473 393
08/30 482 1.64 22.67 773 563 478 394
08/31 493 1.82 27.23 795 574 485 397
09/01 480 1.93 23.28 823 587 493 399
09/02 496 1.96 9.47 846 599 500 402
09/03 508 0.78 16.55 855 604 503 403
09/04 501 1.18 9.02 (99.23) 872 612 508 404
09/15 521 0.49 5.51 971 662 538 414
09/16 498 0.40 4.65 977 665 540 415
09/17 530 0.28 5.39 981 667 541 415
09/18 538 0.35 3.80 (117.92) 987 670 543 416
10/19 586 0.18 3.39 (30.52) 1105 728 578 428
10/28 593 0.17 10.76 (86.06) 1135 744 587 431
11/05 593 0.64 5.44 (70.77) 1221 787 613 439
11/18 596 0.33 ? 1292 822 634 446
?? 244 ? ? 940 470 282 94
92
Environment Canada WSR-98R site at Prince George.
Also shown in Figure 4.8a are the model-simulated plume rise using three
different ? values (red; 30%, yellow; 50%, gray; 100%). Solid lines represent the
time prior to the CALIOP data gap, after which interpolation is used to estimate
extinction for input into SBDART. Before this gap, there is a fairly linear and
consistent rise seen in the observations, and computed by each model estimate.
? = 0.3 gives the best agreement, matching both the ? and ?? almost exactly. This
time period (mid-August?early-September) is a relatively early part of the plume?s
life when extinctions were large, as shown where stratospheric AOD > 1.0 (b). After
the gaps, CALIOP shows a slower ascent, and the ? = 0.3 modeled uplift is still in
agreement. The earliest four days show very little diabatic rise (12?16 August; gray
bar), and are denoted as the NP time period (See Figure 4.7), and the days that
follow are labeled as the EP as the plume begins to undergo strong ascent.
Table 4.1 lists the observed plume ? and AOD, and the model heating rates
and predicted plume ?? at each time step. There are several days early in the plume?s
lifetime where ? = 50% or 30% yield a closer match to observations, but on the whole
? = 30% has better agreement. The last line in Table 4.1 shows the net change in
the plume potential temperature and the net change in each ?-value prediction.
??=30% is the closest match of the choices here, but still has a net difference = 38 K.
However, the observation on 5 November had a jump in AOD from 0.17 to 0.64 that
may be causing an overestimation of heat accumulation on that day. Ignoring the
last two observations would reduce this disagreement (244? (587? 352)) K = 9 K
for ? = 30%. It is unknown why there is a jump in AOD during this last couple of
93
2014 GSL pyroCb
a)
model observations
? = 1.0
? = 0.5 CALIOP plume theta
? = 0.3
b)
Figure 4.9: As in Figure 4.8, but for the GSL pyroCb plume. NP phase is shorter
than the 2017 PNE case before stratospheric ascent is observed. Also, this plume
was only trackable for approximately two weeks, and had a much smaller total ??.
Injected plume ? = 357 K, averaged over depth of CALIOP observation is noted.
CALIOP observations. As shown in Figure 4.5d (inset) these AODs were calculated
during a ?recurve? in the plume?s trajectory as it move back toward the northeast
and out of the tropics. It is possible that the model atmosphere (from GEOS-5) has
changed in a way that increases the extinction estimation from CALIOP profiles.
94
Table 4.2: As in Table 4.1, but for the 2014 GSL pyroCb plume.
date plume ? AOD ?Q (n?Q) ??=100% ??=50% ??=30% ??=10%
08/05 357 10.61 19.45 (38.91) ? ? ? ?
08/07 359 3.94 9.36 384 371 365 360
08/08 367 1.61 6.50 (19.51) 403 380 371 362
08/11 373 1.10 2.19 412 384 373 362
08/12 376 0.37 2.07 418 387 375 363
08/13 373 0.35 2.90 (5.81) 420 388 376 363
08/15 377 0.49 2.67 (8.00) 422 389 376 363
08/18 383 0.44 0.59 (1.18) 425 391 377 364
08/20 375 0.10 ? 427 392 378 364
?? 18 ? ? 70 35 21 7
The 2014 GSL and the 2009 Black Saturday Event pyroCb plumes are also an-
alyzed with this observation/modeling framework (Figures 4.9, 4.10, respectively).
Using the same values, ? = 30% appears to be the best approximation of the
instrument-observed diabatic rise for these events as well, despite each event having
a different scale and duration. Figure 4.9 shows that the GSL plume had fewer ob-
servations, and was only trackable for a little over two weeks. The NP stage lasted
less than three days before significant plume rise was observed. Again, the AOD (b)
was strong initially with values > 3 during the first two days, and then gradually
reduced to < 1 after the third day. Table 4.2 shows the observed ?? = 18 K with
a small overestimation by ???=30% of only 3 K.
Similar results are seen in the 2009 Black Saturday case (Figure 4.10). This
case was similar to 2017 PNE in that there were several pyroCb plumes contributing
to the total aerosol loading. An early AOD ? 8 was observed by CALIOP during
95
2009 Black Saturday
a)
model observations
? = 1.0 MLS plume theta
? = 0.5
? CALIOP plume theta = 0.3
b)
Figure 4.10: As in Figure 4.8, but for the 2009 Black Saturday pyroCb plume. Green
??? in this case are MLS observations colocated HCN/CO maxima to fill CALIOP
gap (light blue vertical bar between mid-February to mid-March). Blue ??? are
CALIOP observations, and black ??? is surface lidar observation from Sa?o Jose? dos
Campos, Brazil. Melbourne surface radar echotop-based ? = 370 K is noted.
the NP (b). AOD dropped to ? 2 within about 3 days, and was steady for about
one week. During this week, the plume experienced a significant rise (= 42 K),
and ???=30% applied over this time period yields 44 K, a minimal overestimation.
96
Table 4.3: As in Table 4.1, but for the 2009 Black Saturday pyroCb plume.
date plume ? AOD ?Q (n?Q) ??=100% ??=50% ??=30% ??=10%
02/08 376 8.46 18.75 ? ? ? ?
02/09 381 2.00 24.46 394 385 381 377
02/10 395 2.72 17.70 419 397 388 380
02/11 400 1.86 20.86 436 406 394 382
02/12 411 1.99 21.89 457 416 400 384
02/13 407 2.19 19.54 479 427 407 386
02/14 410 2.00 23.93 499 437 412 388
02/15 418 2.47 3.80 522 449 420 390
02/16 424 0.33 7.16 (200.46) 526 451 421 391
03/16 496 0.48 10.81 (21.62) 727 551 481 411
03/18 510 0.72 1.77 (46.45) 749 562 487 413
04/13 495 0.12 ? 795 585 501 417
?? 119 ? ? 419 209 125 41
Unfortunately, after 15 February there was a prolonged data outage from CALIOP,
and when data acquisition resumed in mid-March there was significant ambiguity in
connecting the plume to the observations before the gap. This case did not have a
significant MLS H2O anomaly (discussed further in Chapter 5), so another biomass
burning tracer HCN was thresholded with CO to document ? increases (green ???
in Figure 4.10a). When the model is reapplied to CALIOP observations after the
data gap, ???=30% underestimates ? by about 15?20 K. At the end of CALIOP?s
trackable lifetime, however, the net difference is a small overestimation by 6 K (last
line in Table 4.3).
The take home message from these cases is that pyroCb plumes, regardless of
the scale and duration of the initial convective event, appear to have an empirical
heat accumulation efficiency of 30%. Once again, this model is an oversimplification.
97
Potentially useful because of its simplicity, but it does not explicitly account for LS
dynamics, and the effects that instrument retrievals and radiative transfer assump-
tions such as optical properties have on the heating. For example, the delineation
between black and brown (organic) carbon contained within the aerosol plume is
known to have a strong effect on the overall absorption and scattering properties
of the plume. For the sake of simplicity and portability, these properties herein
are assumed to be constant over the life of the plume in this empirical algorithm.
Thus by design, this is an unrealistic accounting of the ?efficiency? of smoke plumes
to convert solar insolation to diabatic lofting, but has the benefit of computational
efficiency and avoidance of non-confirmable assumptions about aerosol optical prop-
erties and local turbulence/diffusion. It must be stressed that this approach does not
attempt to quantify the dynamic nature of evolving Mie-theory single scatter albedo
or asymmetry parameter, only to account for the diminishing extinction coefficient
as measured by CALIOP.
Additional pyroCb cases with observed diabatic rise are shown in Table 4.4
with citations. Up until now there has been very little focus on diabatic self-lofting
after stratospheric injection. With the large scale 2017 PNE case, and the under-
standing that pyroCb are a seasonal occurrence, this work will be useful in the future
as more evidence is gathered. The stratospheric aerosol mass estimation combined
with a plume-rise modeling framework is a good first step in understanding the
net impact that pyroCb have on climate through stratospheric processes. Another
remaining question is the role of radiatively active gases within these plumes. As
was shown in Chapters 2 and 3, the local WVMR contribution can be significant,
98
Table 4.4: PyroCb events with significant observational evidence of diabatic plume
rise that have been published in peer-reviewed literature. Included are three addi-
tional cases not analyzed herein (all values of z in km and ? in K).
pyroCb event z0 (?) z1 (?) zmax. (?) duration (months)
Chisholm1,2,3, 2001 14 (379) 12 (337) 18 (450) 3
Canberra4, 2003 15 (367) 16 (385) 19 (450) 2
Great Divide5, 2006 12 (340) 13 (343) 18 (425) 1-2
Black Saturday6,7,8, 2009 15 (364) 16 (376) 22 (525) 4
GSL9, 2014 14 (357) 14 (359) 15 (383) 0.5
PNE10,11, 2017 14 (378) 14 (373) 24 (> 600) > 6
1. Rosenfeld et al. (2007)
2. Fromm et al. (2008b)
3. Fromm et al. (2006)
4. Dirksen et al. (2009)
5. Dirksen et al. (2009)
6. Siddaway and Petelina (2011)
7. Pumphrey et al. (2011)
8. Dowdy et al. (2017)
9. Kablick et al. (2018)
10. Khaykin et al. (2018)
11. Peterson et al. (2018)
through both direct specific humidity enhancements and down stream sublimation
of injected ice (Figure 2.18). The next chapter discusses the radiative impact of this
process.
99
Chapter 5: Radiative Effects of Residual PyroCb-Injected UTLS Wa-
ter Vapor
The global energy budget contains an average clear sky radiative forcing from
water vapor of RF = +71 W m?2 (Kiehl and Trenberth, 1997). Compounding the
climate issue of direct radiative forcing is the possibility of feedbacks from increased
WVMR, thereby making the problem nonlinear. As discussed in Chapter 1, the
sources of LS H2O are thought to be mostly constrained, but recent studies have
disagreements about the trends since the early 1980s. Quantifying the ability of deep
convection, if any, to increase moisture in the UTLS under certain climate change
scenarios is a problem that cannot be answered without understanding how deep
convection itself would respond to climate forcing. The observationally-based water
vapor feedback of +0.3 W m?2 as concluded by Dessler et al. (2013) was inferred
from a correlation between the increase in observed stratospheric WVMR and an
increase in tropospheric temperature, and was reproduced in a global chemistry-
climate model. Their conclusion is a start to the conversation, but it cannot be
considered axiomatic simply because of the disagreement on what the UTLS H2O
trend actually is; this conflicting data is assimilated by their model.
It is yet to be shown?or studied for that matter?what effect climate would
100
have on pyroCb frequency, seasonality, or intensity. To undertake such a study would
require several decades (centuries, even) of pyroCb and fire data, and a complete
understanding of the mechanics behind pyroconvection so that a detailed climate
model could be parameterized to handle these events. Using the climate questions
surrounding cloud-radiation and cloud-feedback as analogs to questions about py-
roCbs, it seems likely that such a study would require several decades of research,
dedicated observation platforms/campaigns and closure experiments.
On the other hand, the first-order question of ?what effect do pyroCbs have
on climate?? is one that can probably be answered with current instruments and
models. The first things to investigate are the impact of stratospheric aerosol (dis-
cussed in Chapter 4), and other relevant climate forcing agents such as water vapor.
The work contained in this chapter seeks to assess what the ?real-time? water vapor
forcing is of the 2017 PNE pyroCb plumes. Chapter 4 discussed the enhancements of
H2O in the context of tracking the plume, and this chapter will put these enhance-
ments in context with a long-term global data record, compute the stratospheric
WVMR RF using a highly-detailed radiative transfer model, and compare the sen-
sitivity of this RF using two different techniques to determine the H2O anomalies
associated with this case.
5.1 Observed Water Vapor Enhancements
The WVMR impact of the 2017 PNE plume is put into better context when
first analyzing statistics of all H2O observations in the northern hemisphere. Fig-
101
a) b)
c) d)
Figure 5.1: August 2017 histograms of all Aura MLS H2O observations in the North-
ern Hemisphere lower stratosphere using coincident CO thresholds (a) ? 20 ppbv,
(b) ? 30 ppbv, (c) ? 40 ppbv, (d) ? 70 ppbv. Count densities shown for each 1 km
layer between GPH = 15?25 km for observations constrained by 380 ? ? ? 600 K.
(Continued on following page)
102
Figure 5.1: (continued) Note logarithmic scale. Observations to the left of 7 ppmv
threshold (dashed line) were ignored for plume identification used in Chapter 4.
ures 5.1-5.4 show these statistics. MLS H2O count densities are presented in 1 km
altitude bins between 15?25 km for August through November 2017. In each Figure,
panel a) shows WVMR distributions where coincident CO? 20 ppbv, b) CO? 30
ppbv, c) CO? 40 ppbv, and d) CO? 70 ppbv. The vertical dashed line is the 7
ppmv H2O threshold used in tracking the plume throughout the observation period
(see Figures 4.5 and 4.6).
In August, prior to the plume reaching above 20 km, there are relatively
few H2O observations at altitudes greater than this height for all coincident CO
values. In fact, all H2O observations at z ? 22 km are ? 6 ppmv, even for low CO
threshold, and the water vapor distributions in this part of the lower stratosphere
consistently have a peak at ? 5 ppmv with an approximate range of ?1 ppmv,
irrespective of the coincident CO (see orange/red distributions in Figure 5.1). It
appears that there is little correlation between H2O/CO during the month of August
in the northern hemisphere at these altitudes because the H2O count densities are
consistently centered at 5 ppmv in (a?d) even with absolute counts decreasing with
greater CO mixing ratios.
This lack of correlation in August makes sense in general because the major
sources of LS H2O and CO are not correlated in the northern hemisphere at this
time (Filipiak et al., 2005). CO has a photochemical lifetime between 1?3 months
103
a) b)
c) d)
Figure 5.2: As in Figure 5.1 but for September 2017.
in the stratosphere, and approximately two months in the troposphere, so it can
be a useful tracer of tropospheric air entering the stratosphere if the time scale of
transport is shorter than these lifetimes; e.g., deep convection. Strong correlations
between stratospheric CO and H2O are typically only present within the polar vor-
104
a) b)
c) d)
Figure 5.3: As in Figure 5.1 but for October 2017.
tex during the winter months when air is descending from the upper stratosphere
and lower mesosphere. CH4 oxidation is primarily responsible for H2O produc-
tion and photolytic chemistry of CO2 in the thermosphere is the primary source of
stratospheric CO. Other correlations, especially in the lower stratosphere tend to be
105
a) b)
c) d)
Figure 5.4: As in Figure 5.1 but for November 2017.
indicative of a tropospheric source (Randel et al., 2010).
At z ? 19 km, the concurrence of CO and H2O is apparent in the distributions
for August, when deep convection is strongest in the NH (gray, blue, light green
distributions in Figure 5.1). The majority of the CO ? 70 ppbv H2O counts occur
106
in the lower-most stratosphere (15?16 km), which is to be expected since there
is a relatively large amount of continental and monsoon convection during that
month. However, since most of these large CO mixing ratios are confined to the
lower altitudes it does not appear that much, if any, tropospheric air intrudes into
z >? 20 km, at least not on the time scale of 1?2 months when the convection
is most active during July?August. During September?November, the concurrence
is not as apparent at the altitudes between 15?20 km, but at higher altitudes the
distributions indicate a stronger correlation. For example, in September when CO
concentrations within the plume are still large, there are many more concurrent
H2O counts at z > 22 km (compare Figure 5.2d where water vapor ? 10 ppmv
and CO ? 70 ppbv coexist at high altitude with Figure 5.1d where there are no
concurrences).
In October (Figure 5.3), the count density distributions at lower altitudes
become narrower as tropospheric sources from deep convection become less frequent,
but there are still several observations of large H2O/CO concentrations at z > 22 km.
In November (Figure 5.4), all high-altitude observations of H2O now have coincident
CO concentrations < 70 ppbv. It is interesting to note that it appears the median
value of H2O has increased at by ? 1 ppmv by this point. This is demonstrated
when comparing the upper altitude distributions in Figure 5.1 to those in Figure 5.4.
Median values of count density are centered around 5 ppmv in August, and have
increased to 6 ppmv in November for the entire NH.
As shown in Chapter 4, even intense radiatively-induced self-lofting can be
? 1?2 km week?1 in the LS. So, diminishing CO concentration is to be expected
107
over this time period as the plume ages. On the other hand water vapor is longer
lived, but undergoes the same fate as all concentrated plumes over time: dilution
through diffusion. Using the plume tracking described in the previous chapter, the
next section uses these H2O enhancements to estimate changes to boundary fluxes,
and stratospheric heating rates.
5.2 Line-by-Line Simulations
The Line-By-Line Radiative Transfer Model (LBLRTM, Clough et al., 1992)
is the framework used to estimate the longwave water vapor forcing associated with
the 2017 PNE plume. This model was developed by Atmospheric and Environmental
Research corporation, and uses the extremely detailed HITRAN 2016 database of
line spectra for H2O and about 50 additional species (Gordon et al., 2017). The
model and spectral database are freely available, and have both been validated
several times over using both laboratory experiments, and long term atmospheric
data records from the Department of Energy?s Atmospheric Research Program site
at the Southern Great Plains (Clough et al., 2005).
The methodology of the forcing simulation experiments used herein is based
on plume-centric H2O anomalies. These anomalies are created by first constructing
a spatiotemporal stratospheric climatology of all 2006?2017 Aura MLS H2O obser-
vations, and subtracting it from plume observations. It is similar to the climatology
used in Chapter 2 with a couple of key differences: i) it is zonally-averaged for both
the tropical (30 ?S to 30 ?N) and mid-latitude (30 ?N to 60 ?N) bands, and ii) the
108
troposphere temperature, pressure and humidity profiles are held constant using the
AFGL Tropical (TRP), MidLatitude Summer (MidLatS) and MidLatitude Winter
(MidLatW) standard soundings. The TRP sounding is applied during the entire
year in the 30 ?S to 30 ?N band, but for 30 ?N to 60 ?N the MidLatS is applied
between August?October, and the MidLatW is applied between November?March.
The stratospheric component to this climatology is individual-date-specific, accord-
ing to the value determined by averaging the MLS record on that date for that
latitude band for every year during 2006?2017.
This climatology is used to calculate WVMR anomalies for the Square, Trian-
gle and Circle plume lifetimes (see Figure 4.5) using two different methods. The first
method is called the ?Fixed Anomaly,? and computes an anomaly that is confined
to the plume altitudes only within the stratosphere. The plume altitude is deter-
mined using the H2O/CO threshold, and then the WVMR climatology is subtracted
from the MLS plume observations at those altitudes only. Other altitudes in the
stratospheric portion of the profile are assigned climatological WVMR values. This
technique removes the small scale variation in day-to-day UTLS H2O at altitudes
that are not associated with the plume. Additionally, the Square plume uses the
TRP tropospheric profile for its entire lifetime for the Fixed Anomaly. The rea-
son for this choice will become apparent when making comparisons, but the short
explanation is that this removes the seasonal variation in RF .
The second methodology is termed the ?Lagrangian Anomaly,? and does not
place a plume-altitude constraint on the stratospheric anomaly profile. In other
words, this anomaly does consider the small H2O variations at all stratospheric
109
altitudes, regardless of whether they are associated with the plume or a CO en-
hancement. Also, the Square plume?s tropospheric profile is allowed to vary as it
is tracked from the tropics to the midlatitudes during September?November 2017.
It therefore contains tropospheric profiles from any of the three AFGL standard
atmospheres mentioned above, depending on the date and latitude of the MLS ob-
servation. This is in contrast to the TRP-only troposphere for the Square plume in
the Fixed Anomaly methodology.
The idea behind this Lagrangian Anomaly is that RF should be closer to real-
ity since the seasonality, longitude, and full-depth stratospheric H2O are taken into
account when considering the ?background? H2O conditions. The purpose of these
calculations is to put a finite real-time RF estimate on stratospheric water vapor
intrusions for the three separate long-lived plume sections of the 2017 PNE event.
If one were to have a broadband radiometer at the tropopause looking upward and
one at the top of the atmosphere (TOA) looking downward at the same column, the
actual RF could be measured and compared with this estimate. Unfortunately, such
a system is not in place, but these estimates nonetheless are a starting benchmark.
The next two sections detail the results for the three plume sections using the above
two methodologies.
5.2.1 Fixed Anomaly Results
The Fixed Anomalies following the Square, Triangle, and Circle plume sections
are shown in the three left-hand panels Figure 5.5. The red shaded contours are
110
a) b)
c) d)
e) f)
Figure 5.5: (left column) Aura MLS WVMR anomalies (red contours) and fixed
local climatologies (black contours) within the three 2017 PNE pyroCb plumes after
the 12 September separation. All contours in ppmv. (right column) Longwave
LBLRTM heating rate anomalies throughout the troposphere and lower stratosphere
computed from these observations. Square plume (top row), Triangle plume (middle
row) and Circle plume (bottom row). Vertical gray bars indicate where MLS missed
observations.
111
the WVMR anomalies: values greater than the climatological background amounts
(dark gray contours). Light gray vertical bars show dates when MLS did not observe
the plumes. The most apparent aspect of all three plumes is the upward motion of
each as it ages (See Chapter 4). Note the temporally-invariant WVMR climatology
shows typical values ? 4?5 ppmv within the altitude of the Square plume at 20 <
z < 25 km (a). This plume has anomalies > 7 ppmv in the early days of this time
series. Similar results are seen for the midlatitude Triangle and Circle plumes with
slightly smaller anomalies, and shorter durations. As each plume ages, the anomalies
reduce to values between 2?3 ppmv. Because the climatological background values
are between 4?5 ppmv, this puts the total concentrations less than the 7 ppmv
threshold, and they are no longer tracked. Note that tropospheric climatological
values are not displayed here because the WVMR profile gains several orders of
magnitude at lower altitudes (see Figure 1.1a for reference).
The right-hand panels in Figure 5.5 are the corresponding longwave heating
anomalies (?QLW ) from LBLRTM, and profiles of these QLW and ?QLW are shown
in Figure 5.6. The model is run for both the MLS observation profile (with Fixed
Anomaly vertical constraints) and the climatology profile, and then differenced to
determine ?QLW . There is a cooling enhancement within the Square plume down
to ?0.5 K day?1 during the early days (Figure 5.5b), and a very small warming
enhancement throughout the depth of the troposphere of < +0.05 K day?1. Dur-
ing mid-to-late October, there is an increase in warming at the tropical tropopause
(? 16 km) and slight cooling below in the troposphere. There is cooling within and
above the upper section of WVMR plume anomaly because it of its enhanced absorp-
112
tion and re-emission at colder temperatures of terrestrial longwave radiation from
below. Results from this plume recreate the classic positive RF scenario discussed
in introductory radiative transfer texts used to illustrate first-order global warming.
The Triangle and Circle ?QLW time series panels (d,f) show similar results.
Individual heating rate and the anomaly profiles from this methodology using
these plume-centric MLS observations are shown in Figure 5.6. Profiles are colored
according to their date (see inset legend for each plume?s dates). The ?QLW in the
right-hand column show the cooling within and above the plumes and the warming
at and below the tropopause. The two midlatitude plumes have a slightly smoother
warming in the UTLS below the plume (d,f) than the tropical plume (c). The
MidLatS profile used to compute the Fixed Anomaly has a climatologically lower
tropopause height (? 13 km) than the TRP profile (? 16 km). An interesting
result seen here is that the level of maximum cooling from the plumes (minimum
in ?QLW ) rises to the same distance from the tropopause (9 km) in each case just
prior to dissipation (with these thresholds). Stratospheric cooling is strongest in the
early observations of the plumes when WVMRs are largest.
Figure 5.7 shows longwave RF at TOA and the tropopause for the Fixed
Anomaly of these three plumes. The square plume is above the level of maximum
cooling in the standard TRP atmosphere, and its presence slightly decreases the
outgoing longwave radiation (OLR) in the tropics throughout the duration of its
lifetime. Square plume TOA RF (a) is between 0.00 and ?0.15 W m?2 for the
entire time series. The midlatitude Triangle and Circle plumes on the other hand
have weakly positive RF at TOA (< +0.05 W m?2) for both of these plumes (c,e).
113
a) Square b) Square
c) Triangle d) Triangle
e) Circle f) Circle
Figure 5.6: (Continued on following page)
114
Figure 5.6: (continued) Fixed Anomaly H2O longwave radiative heating (left col-
umn) and radiative heating anomalies (right column) for the individual dates of the
three plume sections. Note the significant cooling within and above the plume and
mild warming below the plume to the tropopause.
This positive RF is from an increase in OLR because the peak in the WVMR
anomalies remain within a fairly weak layer of radiative cooling between 17?24 km
for the MidLatS atmosphere (see Figure 5.6c, d). The converse is true for the
Square plume. There is a decrease in OLR because the H2O anomaly exists at the
top of the minimum (maximum) in longwave cooling (heating) between 20?25 km,
and temperature decreases with height much quicker in the TRP lower stratosphere
above 25 km, which enhances the ?cooling to space? aspect of stratospheric energy
budgets (Figure 5.6a).
However, with the exception of the latter half of the Square plume the associ-
ated tropopause RF is strongly positive for all cases (b, d, f), and is approximately
an order of magnitude greater than TOA forcing. All three plumes show a stronger
initial RF with a maximum value reaching between +0.5 and +1.0 W m?2 in the
first two weeks of this time series, indicating a substantial increase in downward
longwave irradiance. Recall from Equations 1.1 and 1.2 that RF is a measure of the
change in net flux between the anomaly and control atmospheres, so RF > 0 at the
tropopause is an increase in F into the troposphere.
115
a) b)
c) d)
e) f)
Figure 5.7: TOA (left) and tropopause-level (right) longwave RF using the H2O
Fixed Anomalies from Figure 5.5. Note the Square plume is confined to the tropics
for most of its observable duration, and has a negative RF at TOA (a decrease in
OLR), whereas the Triangle and Circle plumes are in the mid-latitudes, and have a
positive RF at TOA. In all cases, there is a positive RF at the tropopause.
5.2.2 Lagrangian Anomaly Results
Figure 5.8 shows the WVMR anomalies and climatology using this method-
ology. The dark gray contours on panels (a,c,e) vary with time and location as
the plumes are tracked, and the anomalies (red shaded contours) are computed
against these background values. The results shown here are similar to the Fixed
Anomaly results shown in Figure 5.5 with some small differences. First, the maxi-
mum anomaly in the Square plume (= +7.0 ppmv) is +0.5 ppmv greater than the
Fixed Anomaly. Absolute WVMR in this case is 11 ppmv with background con-
centrations between 4.0?4.5 ppmv. Another difference is that there is no positive
116
a) b)
c) d)
e) f)
Figure 5.8: As in Figure 5.5, but vapor concentration anomalies and radiative heat
flux differences are calculated using the Aura MLS Lagrangian H2O climatology.
anomaly between z = 25?30 km around in the first two weeks of the Square plume
as there is in the fixed anomaly (see Figure 5.5a). The most apparent difference is
the mitigation of ?QLW .
The effect of using a different temperature and vapor climatology is apparent
when comparing Figure 5.9 with Figure 5.6. QLW profiles in Figure 5.9a,c,e show
different clusters of heating rates that correspond to the latitude/time regime the
117
a) Square b) Square
c) Triangle d) Triangle
e) Circle f) Circle
Figure 5.9: (Continued on following page)
118
Figure 5.9: (continued) As in Figure 5.6, but for the Lagrangian climatology. (a)
Note the Square plume has three general QLW profiles, and (c) the Triangle has
two as these plumes move to different latitude bands through summer and winter
months.
plume is in. Easily visible at z < 15 km, for example, the Square plume moves from
the MidLatS to TRP to MidLatW background profiles over its lifetime, and therefore
has different QLW profiles that represent these regimes. Profiles in Figure 5.9b,d,f
show that ?QLW is approximately half as intense as the Fixed Anomaly at the plume
centroid. Once again, there is cooling within an above each of the plumes, but the
extrema in ?QLW are approximately ?0.25 K day?1 for all cases. The warming
between the tropopause and the plume is significantly mitigated with values < 0.1
K day?1 for all cases.
Figure 5.10 illustrates that using a more realistic i) tropospheric climatology,
and ii) full-depth stratospheric WVMR has strong effects on the radiative outcomes
of the simulation. In the case of these three plumes, the sign of the forcing is the
same for all TOA and tropopause RF (Figure 5.7), but the magnitude is again
reduced by ? 50% as the heating rate anomalies were. These values are smaller
than the Fixed Climatology, but are hypothetically more likely to represent reality.
It would be very useful to have validation in these results with observations of
longwave boundary fluxes.
These results serve as a starting point for the estimation of pyroCb-generated
119
a) b)
c) d)
e) f)
Figure 5.10: As in Figure 5.7, but for the Lagrangian climatology.
stratospheric water vapor forcing. An obvious conclusion from these H2O RF results
is that this component is not as important as the aerosol absorption for plume-
lofting. In fact, within the plumes, the presence of anomalous H2O at these altitudes
has a cooling effect on the air at the top, and just above the plume; the opposite
effect of shortwave absorption by aerosol. This would have a stabilizing effect by
reducing the buoyancy of the parcel. Empirically, therefore it must be that the
aerosol presence is more important for plume dynamics than H2O. Although in the
long run, H2O enhancements outlive the relatively large aerosol particles, and would
likely contribute for a longer time period to tropopause RF as demonstrated here.
5.3 The Big Picture
Having established estimates of H2O-induced boundary RF and stratospheric
?QLW for the different 2017 PNE plumes, the next question is how important is this
120
phenomenon for climate? The answer to this question remains unknown at this time
largely because of inadequate observational capability of tropopause flux anomalies
on the spatiotemporal scale of the plume. However, given our understanding that
lower stratospheric water vapor is a climate relevant variable, it is useful to analyze
a longer dataset to determine the frequency with which events of this magnitude
occur. Figure 5.11 shows this analysis. The dual H2O/CO threshold of 7 ppmv/70
ppbv was reapplied to the Aura MLS data record between 2005?2017 for both the
northern and southern hemispheres. The data was gridded onto ?-coordinates using
the co-located Level 2 MLS Temperature product and the pressure surface of the
MLS retrieval, and all data in the range 380 ? ? ? 600 are analyzed for concurrent
measurements meeting this dual threshold criteria.
Immediately apparent in Figure 5.11 is the diabatic rise from the 2017 PNE
plumes towards the end of the time series in the northern hemisphere (a). Also
noted in Figure 5.11b is the date of the 2009 Black Saturday plume. This latter
case does not appear to have any measurements that meet the threshold. In this
figure there are annual signals at 550 < ? < 600 in both hemispheres from the
polar vortex, and annual northern hemisphere signals 380 < ? < 400 are from
the Asian Monsoon. The polar vortex contains H2O/CO-rich air from descent at
higher stratospheric altitudes. The Monsoon is a time when convection and tropical
features can cause pressure increases and temperature decreases at the levels of
neutral buoyancy, which can lead to a lower ? at the tropopause level, even thought
the geometric height may not significantly change.
Although the PNE plumes are significant, this analysis shows that pyroCb
121
Northern Hemisphere Aura MLS: H2O ? 7 ppmv, CO ? 70 ppbv
2017 PNE
Southern Hemisphere Aura MLS: H2O ? 7 ppmv, CO ? 70 ppbv
2009 Black Saturday
Figure 5.11: The global Aura MLS H2O/CO data record with constant thresholds.
All potential temperature observations are shown for the a) northern hemisphere and
b) southern hemisphere where MLS observes H2O ? 7 ppmv and CO ? 70 ppbv.
Highlighted are the 2017 PNE plume in the northern hemisphere and the lack of
2009 Black Saturday signal in the southern hemisphere. Note the annual signals
between 380 < ? < 400 K are from the Monsoon and those between 550 < ? < 600
K are from each hemispheres polar vortex.
122
events of this scale are certainly not common. One caveat is the dual threshold used
to construct this time series is somewhat conservative for UTLS mixing ratios. An
additional analysis was performed with the H2O threshold lowered to 6 ppmv (not
shown), but it did not show an increase in frequency of occurrence. Admittedly, the
2017 PNE case from which the three H2O-rich plumes emanated was an extremely
large and unprecedented event. A seasonal integration of the more typical smaller
enhancements from individual events would be a way to estimate the net impact
of RF . But given that the largest anomalies from this event produced a highly-
localized RF < +1.0 W m?2, it seems unlikely that individual events would have a
measurable global impact on climate, at least in terms of instantaneous forcing.
However, local effects are still relevant, and the slow-moving, long-lived nature
of the 2017 PNE plumes may have significant effects on LS dynamics and heat
budgets, which would constitute a second-order effect. The question does remain
whether there is a net pyroCb contribution to background stratospheric H2O in
any meaningful way, especially considering the conflicting estimates in water vapor
trends mentioned in Chapter 1. This is another question that could be answered
by seasonal integration of the smaller events, which is reserved for future research.
But the RF < +1.0 W m?2 result of these plumes is significant when comparing
the observationally-based results of Solomon et al. (2010) that showed a 1 ppmv
stratospheric H2O reduction ?paused? increases in global surface temperature over
2000?2010.
123
Chapter 6: Summary of Work, Conclusions, and a Future Direction
for PyroCb Impact Studies
6.1 PyroCb Convection and Cloud Modeling
A case study of the Great Slave Lake (GSL) pyrocumulonimbus (pyroCb) from
5 August 2014 in Northwest Territories of Canada was analyzed using satellite- and
ground-based observations, and multiple ARW cloud-resolving simulations. This py-
roCb was an intense storm that penetrated the tropopause, reaching up to ? 14 km
(?'380 K). Passive imagery from the polar-orbiting MODerate resolution Imag-
ing Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite
(VIIRS) and the Geostationary Operational Environmental Satellite (GOES)-West
detailed the convective lifecycle of this pyroCb, and allowed for comparisons with a
concurrent meteorological cumulonimbus (Cb) that formed to the south in Alberta.
Brightness Temperatures (BT) at thermal- and near-infrared channels indicated the
cloud had a large number density of ice particles with a very small effective radius.
The CloudSat 94 GHz Cloud Profiling Radar (CPR) and Cloud-Aerosol Lidar
with Orthogonal Polarization (CALIOP) instruments made fortuitous intersections
within a few minutes of the active core of both the GSL pyroCb and the Alberta
124
Cb, and allowed for direct comparison of the internal structure of the storms based
on radar reflectivity profiles. Additional deep convective core (DCC) cases from
June?August 2014 in the same region were identified using the CPR and also ana-
lyzed. The pyroCb CPR profiles indicated a suppression of precipitation, and rapid
hydrometeor growth between the freezing level (FL) and homogeneous freezing level
(HFL), whereas the Alberta Cb and 15 other meteorological DCC CPR profiles
showed the presence of precipitation, and smaller changes in radar reflectivity be-
tween the FL and HFL.
An analysis of the meteorology indicated that the GSL pyroCb formed in fa-
vorable convective conditions, but fire radiative power (FRP) retrievals from GOES-
West and a lack of meteorological trigger showed that the fire itself likely initiated
the convection. Surface observations from the Buffalo Junction ground station, over
which the pyroCb advected during the active convection stage, showed no precipita-
tion reached the surface. GOES-West showed the GSL pyroCb anvil was detectable
in thermal infrared imagery for at least 24 hours until it became indistinguishable
from nearby (and lower-altitude) cirrus clouds. This lifetime was approximately
50% longer than the Alberta Cb anvil, which is in agreement with anvil lifetime
results published by Lindsey and Fromm (2008).
The detrained GSL pyroCb anvil was tracked in the lower stratosphere over
two weeks using a combination of HYbrid Single Particle Lagrangian Integrated
Trajectories (HYSPLIT) and CALIOP observations of the aerosol/ice plume. Con-
current Microwave Limb Sounder (MLS) observations of this plume indicated that
ice was present initially, but sublimated within a week. MLS also observed a sub-
125
stantial increase in water vapor mixing ratio (WVMR) values over this time that
produced plume-averaged anomalies as large as +2 ppmv (> 50%) relative to a
2005?2014 climatology. In general, these WVMR anomalies appeared to be anti-
correlated with ice water content (IWC) observations, which implies aging pyroCb
plumes that undergo ice sublimation could enhance WVMR on short timescales.
Simulations of the GSL pyroCb using Advanced Research WRF (ARW) showed
that updraft mass fluxes, cloud ice concentrations, and detrained WVMR were heav-
ily influenced by the surface sensible and latent heat fluxes prescribed over the fire.
Sensitivity tests on other initial conditions, such as lowering the surface aerosol
concentrations and limiting the moisture entrainment indicated these had a smaller
effect. These tests were performed using a full simulation of the GSL pyroCb as
a control. Although the pyroCb was less sensitive to both aerosol and moisture
entrainment, the aerosol concentration had the least influence.
This approach is unique from most convective modeling studies in that the
control is based on the most intense situation (pyroCb). Other convective modeling
studies have found that aerosol increases substantially impact updraft intensity and
microphysics (Khain et al., 2005; Fan et al., 2013), but when starting with a pyroCb,
the model showed that aerosols are not nearly as important as the surface heat fluxes.
This case agrees with previous idealized pyroCb modeling (Reutter et al., 2014). The
ARW simulations shown here also indicated there is a substantial amount of vapor
detrainment in the case of the GSL pyroCb, but by itself the absolute humidity is
not enough to match MLS observations of the plume one week later. We estimate
that a 30% survival of all water produced in the model (including both sublimation
126
from ice clouds and absolute humidity detrained at cloud top levels) would account
for the 7-9 ppmv values observed in the aging plume.
6.2 Stratospheric PyroCb Aerosol
The methodology of estimating the stratospheric injection mass of the August
2017 Pacific Northwest pyroCb Event (PNE) published in Peterson et al. (2018)
was demonstrated to be valid by simulating the plume?s Ultra-Violet Aerosol Index
(UVAI) signal from the Ozone Mapping and Profiler Suite (OMPS) instrument.
Total mass estimates were 0.2?0.1 Tg (?50% error) based on a mass density (M?)-
lidar backscatter relationship using a range of optical properties and vertical plume
depth estimates. The 532 nm total attenuated backscatter (??532; CALIOP?s main
measurement) suggested the stratospheric depth of the plume was 1.5?2.0 km. This
information was colocated with OMPS UVAI imagery to determine that UVAI = 15
is a sufficient threshold for determining stratospheric aerosol presence in locations
where the nadir-viewing CALIOP lidar is not able to measure depth. The area of
the plume equals the total area of pixels subtended by this UVAI threshold and was
multiplied by the stratospheric depth to determine an overall initial plume volume
V ' 1.475?1015 km3.
The Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model
was used to simulate the two-wavelength OMPS UVAI signal using a CALIOP
stratospheric aerosol optical depth (AOD) estimate along the track of the observa-
tion. This simulation validated the total mass estimate, and provided confidence in
127
both the empirical UVAI threshold of 15, and the total volume estimate of the 2017
PNE plume.
The radiatively-induced self-lofting of this plume was analyzed both from an
observational and modeling standpoint. The plume was tracked using a dual MLS
threshold for CO/H2O, and was shown to separate into three distinct plume sections
(referred to herein as the ?Square,? ?Circle,? and ?Triangle? plumes) over central
Asia one month after the 12 August pyroCb date. Each of these plumes was shown
to circumnavigate the globe in either the midlatitudes (Circle and Triangle) or the
tropics (Square). The Square plume was lofted to approximately 25 km (? ' 600
K) over the course of three months, and encountered the Easterly winds in the
tropical lower stratosphere (LS). WVMR for these plumes was ? 7 ppmv for the
entire three months of observation, which is approximately 2?3 ppmv (+40?75%)
above the background values for this part of the LS.
This diabatic rise was simulated using a simple 1D modeling framework that
combined heating rate anomalies (?Q) computed from SBDART with an empirical
heat accumulation efficiency parameter (?). Using CALIOP stratospheric extinction
profiles as input into SBDART, ?Q was estimated at each observation time, and
? = 30% was shown to closely match the rate of observed diabatic rise. Two
additional cases were examined for this phenomenon: the much smaller GSL pyroCb
plume, and the relatively-large 2009 Black Saturday event from Australia. Both
cases had similar results showing ? = 30% best matched the observed rates of rise,
in spite of their differences in overall aerosol loading and the large difference in
lifetimes.
128
6.3 Water Vapor Forcing
The 2017 PNE plumes were shown to have significant enhancements in strato-
spheric H2O. Monthly global statistics showed the plume?s high correlation with
CO (an excellent biomass burning tracer), and the upward movement of the plume
as it aged. During August 2017, there are very little enhancements at altitudes
20 ? z ? 25 km, and the H2O that is present at these altitudes is between 4?6
ppmv with coincident CO mixing ratios mostly ? 40 ppbv. In October and Novem-
ber at these higher altitudes, median WVMR has increased to > 5 ppmv with a
larger width in count density histograms. From these statistics alone, it appeared
that this event increased the LS median WVMR concentration by ? 20% (for all
locations where CO mixing ratios ? 20 ppbv.
The longwave (25 ? ?? ? 3000 cm?1) radiative effects of this WVMR increase
were tested for each of the Square, Circle and Triangle plumes using two different
anomaly computations: ?Fixed? and ? Lagrangian? frameworks. Both anomaly
methodologies had similar results, but the more realistic Lagrangian Anomaly re-
duced the stratospheric ?QLW from ?0.5 to ?0.25 K day?1, and also reduced
the tropopause radiative forcing RF from +0.5?1.0 W m?2 to +0.25?0.5 W m?2.
These boundary RF were shown to slightly weaken as the plume aged and the
WVMR enhancements were reduced. The tropical Square plume showed RF '
?0.025 W m?2 at the top of the atmosphere (TOA) corresponding to a decrease in
outgoing longwave radiation (OLR). This is in contrast with the midlatitude Circle
and Triangle plumes that showed RF ' +0.025 W m?2 for the observable lifetimes.
129
The reasons behind this difference were explained to be the local (tropical or mid-
latitude) levels of maximum radiative cooling (for the longwave). The Square plume
existed toward the top of this level, and had a steeper lapse rate above it than did
the Circle and Triangle plumes.
The water vapor effects of the PNE event were put into context by showing
global statistics for both hemispheres. This event is unprecedented in the Aura
MLS data record, even when comparing with the previously-known largest pyroCb
event: Black Saturday. The record showed that this type of event is very uncommon.
However, the localized effects of many smaller pyroCbs over the course of a season
may be important.
6.4 Future Work
The observations made by the A-Train and the ARW model results indicate
that lower stratospheric water vapor may be influenced by pyroCb activity, at least
on short time scales. As recent studies have shown, these pyroCb events are a
relatively common occurrence every season (Peterson et al., 2017a). Given the ra-
diative significance of this greenhouse gas in the UTLS, it is worth considering a
larger research effort into understanding the net impact that these events may have.
Integrating the smaller events into a larger estimate of the seasonal impact of py-
roCb on stratospheric water vapor forcing, and further advancements in tracking
and modeling the self-lofting mechanism are needed. Tracking individual plumes
is a valuable technique, but it is now possible to undertake a seasonal, global ap-
130
proach to the impact of these events. There are rich and large observational datasets
from various earth-observing satellites in place that are underutilized for this type
of research. Satellite-based sounding instruments such as MLS, Atmospheric In-
frared Sounder (AIRS, Chahine et al., 2006), Infrared Atmospheric Sounding In-
terferometer (IASI, Blumstein et al., 2004) and Measurements of Pollution in the
Troposphere (MOPITT, Emmons et al., 2004) have different capabilities that com-
plement each other on long time scales when quantifying tropospheric and UTLS
H2O and CO. It would be beneficial to both the meteorological-, climate-, and fire
behavior-communities to explore the local and global impact of pyroCb.
131
Bibliography
Austin, R., A. Heymsfield, and G. Stephens (2009), Retrieval of ice cloud micro-
physical parameters using the CloudSat millimeter-wave radar and temperature,
J. Geophys. Res., 114 (D8).
Bates, D. R., and M. Nicolet (1950), The photochemistry of atmospheric water
vapor, J. Geophys. Res., 55 (3), 301?327.
Bedka, K. (2011), Overshooting cloud top detections using MSG SEVIRI Infrared
brightness temperatures and their relationship to severe weather over Europe,
Atmos. Res., 99 (2), 175?189.
Beringer, J., L. Hutley, N. Tapper, A. Coutts, A. Kerley, and A. O?grady (2003),
Fire impacts on surface heat, moisture and carbon fluxes from a tropical savanna
in northern Australia, Int. J. Wildland Fire, 12 (4), 333?340.
Blumstein, D., G. Chalon, T. Carlier, C. Buil, P. Hebert, T. Maciaszek, G. Ponce,
T. Phulpin, B. Tournier, D. Simeoni, et al. (2004), IASI instrument: Technical
overview and measured performances, in Infrared Spaceborne Remote Sensing XII,
vol. 5543, pp. 196?208, International Society for Optics and Photonics.
Brasseur, G., and S. Solomon (1986), Aeronomy of the Middle Atmosphere, 452 pp.,
Reidel, Dordrecht.
Brewer, A. (1949), Evidence for a world circulation provided by the measurements
of helium and water vapour distribution in the stratosphere, Q. J. Roy. Meteor.
Soc., 75 (326), 351?363.
Chahine, M., T. Pagano, H. Aumann, R. Atlas, C. Barnet, J. Blaisdell, L. Chen,
M. Divakarla, E. Fetzer, M. Goldberg, et al. (2006), Airs: Improving weather
forecasting and providing new data on greenhouse gases, B. Am. Meteorol. Soc.,
87 (7), 911?926.
Clough, S., M. Iacono, and J. Moncet (1992), Line-by-Line Calculation of Atmo-
spheric Fluxes and Cooling Rates: Application to Water Vapor, J. Geophys. Res.,
97 (D14), 15,761?15,785.
132
Clough, S., M. Shephard, E. Mlawer, J. Delamere, M. Iacono, K. Cady-Pereira,
S. Boukabara, and P. Brown (2005), Atmospheric radiative transfer modeling: a
summary of the AER codes, J. Quant. Spectrosc. Ra., 91 (2), 233?244.
Crutzen, P. J., and J. W. Birks (1982), The atmosphere after a nuclear war: Twilight
at noon, Ambio, 11 (2/3), 114?125.
Damoah, R., N. Spichtinger, R. Servranckx, M. Fromm, E. Eloranta, I. Razenkov,
P. James, M. Shulski, C. Forster, and A. Stohl (2006), A case study of pyro-
convection using transport model and remote sensing data, Atmos. Chem. Phys.,
6 (1), 173?185, doi:10.5194/acp-6-173-2006.
Delanoe?, J., and R. Hogan (2010), Combined CloudSat-CALIPSO-MODIS retrievals
of the properties of ice clouds, J. Geophys. Res., 115 (D4).
Dessler, A., M. Schoeberl, T. Wang, S. Davis, and K. Rosenlof (2013), Stratospheric
water vapor feedback, P. Natl. Acad. Sci. USA, 110 (45), 18,087?18,091, doi:10.
1073/pnas.1310344110.
Dessler, A. E. (2002), The effect of deep, tropical convection on the tropical
tropopause layer, J. Geophys. Res, 107 (4033), 9169?9205.
Dirksen, R. J., K. F. Boersma, J. De Laat, P. Stammes, G. R. Van Der Werf, M. V.
Martin, and H. M. Kelder (2009), An aerosol boomerang: Rapid around-the-world
transport of smoke from the December 2006 Australian forest fires observed from
space, J. Geophys. Res., 114 (D21), D21,201, doi:10.1029/2009JD012360.
Dowdy, A., M. Fromm, and N. McCarthy (2017), Pyrocumulonimbus lightning and
fire ignition on Black Saturday in southeast Australia, J. Geophys. Res., 122 (14),
7342?7354.
Emmons, L., M. Deeter, J. Gille, D. Edwards, J. Attie?, J. Warner, D. Ziskin, G. Fran-
cis, B. Khattatov, V. Yudin, et al. (2004), Validation of Measurements of Pollu-
tion in the Troposphere (MOPITT) CO retrievals with aircraft in situ profiles, J.
Geophys. Res, 109 (D3).
Fan, J., L. Leung, D. Rosenfeld, Q. Chen, Z. Li, J. Zhang, and H. Yan (2013),
Microphysical effects determine macrophysical response for aerosol impacts on
deep convective clouds, P. Natl. Acad. Sci. USA, 110 (48), E4581?E4590.
Field, R., M. Luo, M. Fromm, A. Voulgarakis, S. Mangeon, and J. Worden (2016),
Simulating the Black Saturday 2009 smoke plume with an interactive composition-
climate model: Sensitivity to emissions amount, timing, and injection height, J.
Geophys. Res., 121 (8), 4296?4316.
Filipiak, M., R. Harwood, J. Jiang, Q. Li, N. Livesey, G. Manney, W. Read,
M. Schwartz, J. Waters, and D. Wu (2005), Carbon monoxide measured by the
eos microwave limb sounder on aura: First results, Geophys. Res. Lett., 32 (14).
133
Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D. W. Fahey, J. Hay-
wood, J. Lean, D. C. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz,
and R. Van Dorland (2007), Changes in Atmospheric Constituents and in Ra-
diative Forcing, Climate Change 2007: The Physical Science Basis. Contribution
of Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change.
Fouquart, Y., and B. Bonnel (1980), Computations of solar heating of the earth?s
atmosphere: A new parameterization, Contrib. Atmos. Phys., 53, 35?62.
Fromm, M., A. Tupper, D. Rosenfeld, R. Servranckx, and R. McRae (2006), Violent
pyro-convective storm devastates Australia?s capital and pollutes the stratosphere,
Geophys. Res. Lett., 33 (5), doi:10.1029/2005GL025161.
Fromm, M., E. Shettle, K. Fricke, C. Ritter, T. Trickl, H. Giehl, M. Gerding,
J. Barnes, M. O?Neill, S. Massie, et al. (2008a), Stratospheric impact of the
Chisholm pyrocumulonimbus eruption: 2. Vertical profile perspective, J. Geo-
phys. Res., 113 (D8), D08,203, doi:10.1029/2007JD009147.
Fromm, M., O. Torres, D. Diner, D. Lindsey, B. Hull, R. Servranckx, E. Shettle, and
Z. Li (2008b), Stratospheric impact of the Chisholm pyrocumulonimbus eruption:
1. Earth-viewing satellite perspective, J. Geophys. Res., 113 (D8), D08,202, doi:
10.1029/2007JD009153.
Fromm, M., D. T. Lindsey, R. Servranckx, G. Yue, T. Trickl, R. Sica, P. Doucet,
S. Godin-Beekmann, et al. (2010), The untold story of pyrocumulonimbus, B.
Am. Meteorol. Soc., 91 (9), 1193, doi:10.1175/2010BAMS3004.1.
Fromm, M., R. McRae, J. Sharples, and G. Kablick (2012), Pyrocumulonimbus
Pair in Wollemi and Blue Mountains National Parks, 22 November 2006, Aust.
Meteorol. Ocean. J., (62), 117?126.
Fromm, M., G. Kablick, G. Nedoluha, E. Carboni, R. Grainger, J. Campbell, and
J. Lewis (2014), Correcting the record of volcanic stratospheric aerosol impact:
Nabro and Sarychev Peak, J. Geophys. Res., 119 (17), doi:10.1002/2014JD021507.
Fromm, M. D., and R. Servranckx (2003), Transport of forest fire smoke above
the tropopause by supercell convection, Geophys. Res. Lett., 30 (10), 1542, doi:
10.1029/2002GL016820.
Fromm, M. D., R. M. Bevilacqua, R. Servranckx, J. Rosen, J. P. Thayer, J. Herman,
and D. Larko (2005), Pyro-cumulonimbus injection of smoke to the stratosphere:
Observations and impact of a super blowup in northwestern Canada on 3?4 August
1998, J. Geophys. Res., 110 (D8), doi:10.1029/2004JD005350.
Fueglistaler, S., A. E. Dessler, T. J. Dunkerton, I. Folins, Q. Fu, and P. W.
Mote (2009), Tropical tropopause layer, Rev. Geophys., 47, RG1004, doi:10.1029/
2008RG000267.
134
Gatebe, C., T. Varnai, R. Poudyal, C. Ichoku, and M. King (2012), Taking the pulse
of pyrocumulus clouds, Atmos. Environ., 52, 121?130.
Gordon, I., L. Rothman, C. Hill, R. Kochanov, Y. Tan, P. Bernath, M. Birk,
V. Boudon, A. Campargue, K. Chance, et al. (2017), The HITRAN2016 molecular
spectroscopic database, J. Quant. Spectrosc. Ra., 203, 3?69.
Haines, D. (1988), A lower atmosphere severity index for wild-land fires, Natl. Wea.
Dig., 13 (3), 23?27.
Hegglin, M., D. Plummer, T. Shepherd, J. Scinocca, J. Anderson, L. Froidevaux,
B. Funke, D. Hurst, A. Rozanov, J. Urban, et al. (2014), Vertical structure
of stratospheric water vapour trends derived from merged satellite data, Nat.
Geosci., 7 (10), 768, doi:10.1038/NGEO2236.
Holton, J., P. H. Haynes, M. E. McIntyre, A. Douglass, R. Rood, and Pf. (1995),
Stratosphere-troposphere exchange, Rev. Geophys., 33 (4), 403?439, doi:10.1029/
95RG02097.
Hu, Y., D. Winker, M. Vaughan, B. Lin, A. Omar, C. Trepte, D. Flittner, P. Yang,
S. Nasiri, B. Baum, et al. (2009), CALIPSO/CALIOP cloud phase discrimination
algorithm, J. Atmos. Ocean. Tech., 26 (11), 2293?2309.
Hurst, D., S. Oltmans, H. Vo?mel, K. Rosenlof, S. Davis, E. Ray, E. Hall, and
A. Jordan (2011), Stratospheric water vapor trends over Boulder, Colorado: Anal-
ysis of the 30 year Boulder record, J. Geophys. Res., 116 (D2), D02,306, doi:
10.1029/2010JD015065.
Jensen, E., P. Lawson, B. Baker, B. Pilson, Q. Mo, A. Heymsfield, A. Bansemer,
T. Bui, M. McGill, D. Hlavka, et al. (2009), On the importance of small ice
crystals in tropical anvil cirrus, Atmos. Chem. Phys., 9 (15), 5519?5537.
Jensen, E. J., L. Pfister, A. S. Ackerman, A. Tabazadeh, and O. B. Toon (2001),
A conceptual model of the dehydration of air due to freeze-drying by optically
thin, laminar cirrus rising slowly across the tropical tropopause, J. Geophys. Res.,
106 (D15), 17?237.
Kablick, G., M. Fromm, S. Miller, P. Partain, D. Peterson, S. Lee, Y. Zhang, A. Lam-
bert, and Z. Li (2018), The Great Slave Lake pyroCb of 5 August 2014: obser-
vations, simulations, comparisons with regular convection, and impact on UTLS
water vapor, J. Geophys. Res., 123 (21), 12?332, doi:10.1029/2018JD028965.
Khain, A., D. Rosenfeld, and A. Pokrovsky (2005), Aerosol impact on the dynamics
and microphysics of deep convective clouds, Q. J. Roy. Meteor. Soc., 131 (611),
2639?2663.
Khaykin, S., S. Godin-Beekmann, A. Hauchecorne, J. Pelon, F. Ravetta, and
P. Keckhut (2018), Stratospheric smoke with unprecedentedly high backscatter
135
observed by lidars above southern France, Geophys. Res. Lett., 45 (3), 1639?1646,
doi:10.1002/2017GL076763.
Kiehl, J., and K. Trenberth (1997), Earth?s annual global mean energy budget, B.
Am. Meteorol. Soc., 78 (2), 197?208.
Kley, D., J. Russell III, and C. Phillips (2000), SPARC Assessment of Water Vapour
in the Stratosphere and Upper Troposphere, WMO/TD No. 1043 WCRP 113,
World Climate Research Programme, Geneva.
Laat, A., D. Stein Zweers, and R. Boers (2012), A solar escalator: Observational
evidence of the self-lifting of smoke and aerosols by absorption of solar radiation in
the February 2009 Australian Black Saturday plume, J. Geophys. Res., 117 (D4).
Lareau, N., and C. Clements (2016), Environmental controls on pyrocumulus and
pyrocumulonimbus initiation and development, Atmos. Chem. Phys., 16 (6), 4005.
Li, Z., F. Niu, J. Fan, Y. Liu, D. Rosenfeld, and Y. Ding (2011), Long-term impacts
of aerosols on the vertical development of clouds and precipitation, Nat. Geosci.,
4 (12), 888?894, doi:10.1038/NGEO1313.
Lindsey, D., and M. Fromm (2008), Evidence of the cloud lifetime effect from
wildfire-induced thunderstorms, Geophys. Res. Lett., 35 (22), L22,809, doi:10.
1029/2008GL035680.
Lindsey, D., D. Hillger, L. Grasso, and J. Knaff, J.and Dostalek (2006), GOES
Climatology and Analysis of Thunderstorms with Enhanced 3.9-? m Reflectivity,
Monthly weather review, 134 (9), 2342?2353.
Livesey, N., W. Read, P. Wagner, L. Froidevaux, A. Lambert, G. Manney, L. Milla?n,
H. Pumphrey, M. Santee, M. Schwartz, S. Wang, R. Fuller, et al. (2018), Version
4.2x Level 2 data quality and description document, JPL D-33509 Rev. D.
Lossow, S., D. F. Hurst, K. H. Rosenlof, G. P. Stiller, T. von Clarmann, S. Brinkop,
M. Dameris, P. Jo?ckel, D. E. Kinnison, J. Plieninger, D. A. Plummer, F. Ploeger,
W. G. Read, E. E. Remsberg, J. M. Russell, and M. Tao (2018), Trend differ-
ences in lower stratospheric water vapour between Boulder and the zonal mean
and their role in understanding fundamental observational discrepancies, Atmos.
Chem. Phys., 18 (11), 8331?8351, doi:10.5194/acp-18-8331-2018.
Luderer, G., J. Trentmann, T. Winterrath, C. Textor, M. Herzog, H. Graf, and
M. Andreae (2006), Modeling of biomass smoke injection into the lower strato-
sphere by a large forest fire (part ii): sensitivity studies, Atmos. Chem. Phys.,
6 (12), 5261?5277.
Mesinger, F., G. DiMego, E. Kalnay, K. Mitchell, P. Shafran, W. Ebisuzaki, D. Jovic?,
J. Woollen, E. Rogers, E. Berbery, et al. (2006), North American regional reanal-
ysis, B. Am. Meteorol. Soc., 87 (3), 343?360.
136
Miller, S. D., and G. L. Stephens (2001), CloudSat instrument requirements as de-
termined from ECMWF forecasts of global cloudiness, J. Geophys. Res, 106 (D16),
17,713?17, doi:10.1029/2000JD900645.
Minschwaner, K., H. Su, and J. Jiang (2016), The upward branch of the Brewer-
Dobson circulation quantified by tropical stratospheric water vapor and carbon
monoxide measurements from the Aura Microwave Limb Sounder, J. Geophys.
Res, 121 (6), 2790?2804, doi:10.1002/2015JD023961.
Mlawer, E., S. Taubman, P. Brown, M. Iacono, and S. Clough (1997), Radiative
transfer for inhomogeneous atmospheres: Rrtm, a validated correlated-k model
for the longwave, J. Geophys. Res., 102 (D14), 16,663?16,682.
Mote, P., K. Rosenlof, M. McIntyre, E. Carr, J. Gille, J. Holton, J. Kinnersley,
H. Pumphrey, J. Russell, and J. Waters (1996), An atmospheric tape recorder:
The imprint of tropical tropopause temperatures on stratospheric water vapor, J.
Geophys. Res., 101 (D2), 3989?4006.
Nikonovas, T., P. North, and S. Doerr (2017), Particulate emissions from large North
American wildfires estimated using a new top-down method, Atmos. Chem. Phys.,
17 (10), 6423?6438, doi:10.5194/acp-17-6423-2017.
Oinas, V., A. A. Lacis, D. Rind, D. T. Shindell, and J. E. Hansen (2001), Radiative
cooling by stratospheric water vapor: Big differences in GCM results, Geophys.
Res. Lett., 28 (14), 2791?2794.
Peterson, D., M. Fromm, J. Solbrig, E. Hyer, M. Surratt, and J. Campbell (2017a),
Detection and inventory of intense pyroconvection in western North America using
GOES-15 daytime infrared data, J. Appl. Meteorol. Clim., 56 (2), 471?493.
Peterson, D., E. Hyer, J. Campbell, J. Solbrig, and M. Fromm (2017b), A conceptual
model for development of intense pyrocumulonimbus in western North America,
Mon. Wea. Rev., 145 (6), 2235?2255.
Peterson, D., J. Campbell, E. Hyer, M. Fromm, G. Kablick, J. Cossuth, and M. De-
Land (2018), Wildfire-driven thunderstorms cause a volcano-like stratospheric
injection of smoke, Clim. Atmos. Sci., 1, 30.
Potter, B. (2005), The role of released moisture in the atmospheric dynamics asso-
ciated with wildland fires, Int. J. Wildland Fire, 14 (1), 77?84.
Pruppacher, H. R., and J. D. Klett (1997), Microphysics of Clouds and Precipitation,
Kluwer Academic Publishers.
Pumphrey, H., M. Santee, N. Livesey, M. Schwartz, and W. Read (2011), Microwave
Limb Sounder observations of biomass-burning products from the Australian bush
fires of February 2009, Atmos. Chem. Phys., 11 (13), 6285?6296, doi:10.5194/
acp-11-6285-2011.
137
Radke, L., J. Lyons, P. Hobbs, and R. Weiss (1990), Smokes from the burning of
aviation fuel and their self-lofting by solar heating, J. Geophys. Res., 95 (D9),
14,071?14,076.
Ramaswamy, V., M. L. Chanin, J. Angell, J. Barnett, D. Gaffen, M. Gelman,
P. Keckhut, Y. Koshelkov, K. Labitzke, J. J. R. Lin, et al. (2001), Stratospheric
temperature trends: Observations and model simulations, Rev. Geophys., 39 (1),
71?122.
Randel, W., and E. Jensen (2013), Physical processes in the tropical tropopause
layer and their roles in a changing climate, Nat. Geosci., 6 (3), 169.
Randel, W., M. Park, L. Emmons, D. Kinnison, P. Bernath, K. A. Walker, C. Boone,
and H. Pumphrey (2010), Asian monsoon transport of pollution to the strato-
sphere, Science, 328 (5978), 611?613, doi:10.1126/science.1182274.
Reid, J., T. Eck, S. Christopher, R. Koppmann, O. Dubovik, D. Eleuterio, B. Hol-
ben, E. Reid, and J. Zhang (2005), A review of biomass burning emissions part
III: intensive optical properties of biomass burning particles, Atmos. Chem. Phys.,
5 (3), 827?849.
Reutter, P., J. Trentmann, A. Seifert, P. Neis, H. Su, D. Chang, M. Herzog,
H. Wernli, M. Andreae, and U. Po?schl (2014), 3-D model simulations of dy-
namical and microphysical interactions in pyroconvective clouds under idealized
conditions, Atmos. Chem. Phys., 14 (14), 7573?7583.
Ricchiazzi, P., S. Yang, C. Gautier, and D. Sowle (1998), SBDART: A research and
teaching software tool for plane-parallel radiative transfer in the Earth?s atmo-
sphere, B. Am. Meteorol. Soc., 79 (10), 2101?2114.
Robock, A., and B. Zambri (2018), Did Smoke From City Fires in World War II
Cause Global Cooling?, J. Geophys. Res., 123 (18), 10?314.
Robock, A., L. Oman, G. Stenchikov, O. Toon, C. Bardeen, and R. Turco (2007a),
Climatic consequences of regional nuclear conflicts, Atmos. Chem. Phys., 7 (8),
2003?2012, doi:10.5194/acp-7-2003-2007.
Robock, A., L. Oman, and G. Stenchikov (2007b), Nuclear winter revisited with
a modern climate model and current nuclear arsenals: Still catastrophic conse-
quences, J. Geophys. Res., 112, D13,107, doi:10.1029/2006JD008235.
Rosenfeld, D., and W. Woodley (2000), Deep convective clouds with sustained su-
percooled liquid water down to ?37.5?c, Nature, 405 (6785), 440.
Rosenfeld, D., M. Fromm, J. Trentmann, G. Luderer, M. Andreae, and
R. Servranckx (2007), The Chisholm firestorm: observed microstructure, pre-
cipitation and lightning activity of a pyro-cumulonimbus, Atmos. Chem. Phys.,
7 (3), 645?659.
138
Rosenfeld, D., U. Lohmann, G. Raga, C. O?Dowd, M. Kulmala, S. Fuzzi, A. Reissell,
and M. Andreae (2008), Flood or drought: How do aerosols affect precipitation?,
Science, 321 (5894), 1309?1313.
Sagan, C., R. Turco, P. Ehrlich, H. Geiger, and P. R. Archive (1983), The nuclear
winter, Scott Meredith Literary Agency.
Saleeby, S., and W. Cotton (2008), A binned approach to cloud-droplet riming
implemented in a bulk microphysics model, J. Appl. Meteorol. Clim., 47 (2), 694?
703.
Schoeberl, M., E. Jensen, L. Pfister, R. Ueyama, M. Avery, and A. Dessler (2018),
Convective hydration of the upper troposphere and lower stratosphere, J. Geo-
phys. Res, 123 (9), 4583?4593.
Schwartz, M., A. Lambert, G. Manney, W. Read, N. Livesey, L. Froidevaux, C. Ao,
P. Bernath, C. Boone, R. Cofield, et al. (2008), Validation of the Aura Microwave
Limb Sounder temperature and geopotential height measurements, J. Geophys.
Res., 113, D15S11, doi:10.1029/2007JD008783.
Schwartz, M., W. Read, M. Santee, N. Livesey, L. Froidevaux, A. Lambert, and
G. Manney (2013), Convectively injected water vapor in the North American
summer lowermost stratosphere, Geophys. Res. Lett., 40 (10), 2316?2321.
Setva?k, M., D. Lindsey, P. Nova?k, P. Wang, M. Radova?, J. Kerkmann, L. Grasso,
S.-H. Su, R. Rabin, J. S?t?a?stka, et al. (2010), Satellite-observed cold-ring-shaped
features atop deep convective clouds, Atmos. Res., 97 (1-2), 80?96.
Sherwood, S. (2002a), A microphysical connection among biomass burning, cumulus
clouds, and stratospheric moisture, Science, 295, 1272?1275.
Sherwood, S. (2002b), Aerosols and ice particle size in tropical cumulonimbus, J.
Climate, 15 (9), 1051?1063.
Sherwood, S., and A. Dessler (2001), A model for transport across the tropical
tropopause, J. Atmos. Sci., 58 (7), 765?779.
Siddaway, J., and S. Petelina (2011), Transport and evolution of the 2009 Australian
Black Saturday bushfire smoke in the lower stratosphere observed by OSIRIS on
Odin, J. Geophys. Res, 116 (D6), D06,203, doi:10.1029/2010JD015162.
Simpson, J. (1980), Downdrafts as linkages in dynamic cumulus seeding effects, J.
Appl. Meteorol., 19 (4), 477?487.
Skamarock, W., J. Klemp, J. Dudhia, D. Gill, D. Barker, W. Wang, and J. Powers
(2005), A description of the advanced research WRF Version 2, NCAR Tech. Note
NCAR/TN-475+STR, p. 88 pp.
139
Solomon, S., K. Rosenlof, R. Portmann, J. Daniel, S. Davis, T. Sanford, and G. Plat-
tner (2010), Contributions of stratospheric water vapor to decadal changes in the
rate of global warming, Science, 327, 1219?1223.
Stein, A., R. Draxler, G. Rolph, B. Stunder, M. Cohen, and F. Ngan (2015), NOAA?s
HYSPLIT atmospheric transport and dispersion modeling system, B. Am. Mete-
orol. Soc., 96 (12), 2059?2077.
Stephens, G., D. Vane, R. Boain, G. Mace, K. Sassen, Z. Wang, A. Illing-
worth, E. O?Connor, W. Rossow, S. Durden, et al. (2002), The CloudSat mis-
sion and the A-Train, B. Am. Meteorol. Soc., 83 (12), 1771?1790, doi:10.1175/
BAMS-83-12-1771.
Thompson, S., and S. Schneider (1986), Nuclear winter reappraised, Foreign Aff.,
64 (5), 981?1005.
Thompson, S., V. Aleksandrov, G. Stenchikov, S. Schneider, C. Covey, and
R. Chervin (1984), Global climatic consequences of nuclear war: simulations with
three dimensional models, Ambio, pp. 236?243.
Toon, O., R. Turco, A. Robock, C. Bardeen, L. Oman, and G. Stenchikov (2007),
Atmospheric effects and societal consequences of regional scale nuclear conflicts
and acts of individual nuclear terrorism, Atmos. Chem. Phys., 7 (8), 1973?2002,
doi:10.5194/acp-7-1973-2007.
Torres, O., R. Decae, J. Veefkind, and G. De Leeuw (2002), OMI aerosol re-
trieval algorithm, OMI algorithm theoretical basis document: clouds, aerosols,
and surface UV irradiance, Vol. 3, version 2, edited by P. Stamnes, NASA
Goddard Space Flight Cent., Greenbelt, MD. (OMI-ATBD-03. Available at
http://eospso.gsfc.nasa.gov/eoshomepage/forscientists/atbd/docs/OMI/ATBD-
OMI-03.pdf).
Trentmann, J., G. Luderer, T. Winterrath, M. Fromm, R. Servranckx, C. Textor,
M. Herzog, H.-F. Graf, and M. Andreae (2006), Modeling of biomass smoke injec-
tion into the lower stratosphere by a large forest fire (part i): reference simulation,
Atmos. Chem. Phys., 6 (12), 5247?5260.
Turco, R. P., O. B. Toon, T. P. Ackerman, J. B. Pollack, and C. Sagan (1983),
Nuclear winter: global consequences of multiple nuclear explosions, Science,
222 (4630), 1283.
Turco, R. P., O. B. Toon, T. P. Ackerman, J. B. Pollack, and C. Sagan (1984), The
climatic effects of nuclear war, Sci. Am., 251 (2), 33?43.
Turco, R. P., O. B. Toon, T. P. Ackerman, J. B. Pollack, and C. Sagan (1990),
Climate and smoke: An appraisal of nuclear winter, Science, 247 (4939), 166.
140
Uccellini, L., and D. Johnson (1979), The coupling of upper and lower tropospheric
jet streaks and implications for the development of severe convective storms, Mon.
Wea. Rev., 107 (6), 682?703.
Vaughan, M. A., D. M. Winker, and K. A. Powell (2005), Part 2: Feature Detec-
tion and Layer Properties Algorithms, in CALIOP Algorithm Theoretical Basis
Document, PC-SCI-202, p. 87 pp., Langley Research Center, Hampton, VA.
Walko, R., W. Cotton, M. Meyers, and J. Harrington (1995), New RAMS cloud
microphysics parameterization part I: the single-moment scheme, Atmos. Res.,
38 (1-4), 29?62.
Wallace, J. (2003), General Circulation: Overview, in Encyclopedia of Atmospheric
Sciences, edited by J. R. Holton, J. A. Curry, and J. A. Pyle, pp. 821?829, Aca-
demic.
Wang, H., W. Skamarock, and G. Feingold (2009), Evaluation of scalar advection
schemes in the Advanced Research WRF model using large-eddy simulations of
aerosol?cloud interactions, Mon. Wea. Rev., 137 (8), 2547?2558.
Wang, P. (2003), Moisture plumes above thunderstorm anvils and their contributions
to cross-tropopause transport of water vapor in midlatitudes, J. Geophys. Res.,
108 (D6).
Wang, T., and A. Dessler (2012), Analysis of cirrus in the tropical tropopause layer
from CALIPSO and MLS data: A water perspective, J. Geophys. Res., 117,
D04,211, doi:10.1029/2011JD016442.
Weisman, M., and J. Klemp (1982), The dependence of numerically simulated con-
vective storms on vertical wind shear and buoyancy, Mon. Wea. Rev., 110 (6),
504?520.
Winker, D., M. Vaughan, A. Omar, Y. Hu, K. Powell, Z. Liu, W. Hunt, and
S. Young (2009), Overview of the CALIPSO mission and CALIOP data pro-
cessing algorithms, J. Atmos. Ocean. Tech., 26 (11), 2310?2323, doi:10.1175/
2009JTECHA1281.1.
World Meteorological Organization (1992), International Meteorological Vocabulary
(2nd ed.), 636 pp., doi:(ISBN)978-92-63-02182-3.
Yan, H., Z. Li, J. Huang, M. Cribb, and J. Liu (2014), Long-term aerosol-mediated
changes in cloud radiative forcing of deep clouds at the top and bottom of the
atmosphere over the Southern Great Plains, Atmos. Chem. Phys., 14 (14), 7113?
7124, doi:10.5194/acp-14-7113-2014.
141