Climate impacts of biomass burning in high NH latitudes

1
Climate impacts of biomass burning in high NH
latitudes
Phil Rasch (NCAR)

• With thanks to Dani Bundy-Coleman, Gabi Pfister,
  Mark Flanner, Natalie Mahowald, Fabrizio Sassi,
  Peter Hess, Jim Randerson, Karl Taylor, Mike
  Fromm (NRL), Omar Torres, Dave Diner, Ralph
  Kahn, John Martonchik (NASA)

2
Highlights from 3 studies

• Potential for pyrocumuli on stratospheric
  temperature and winds (Rasch, 2005, AGU)
• Role of biomass burning on radiative forcing and
  surface temperature at high latitude (Pfister,
  2007, submitted)
• Role of soot deposition on snow on climate
  (Flanner etal, 2007, JGR, this week?)

3
Pfister etal study
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
4
inverse modeling of CO emissions

• MOPITT CO Observations
• CTM Simulations with MOZART
• Data Assimilation
• A Priori Emissions based on daily MODIS fire
  counts

A Posteriori US Anthr. Emissions
Change in Tropospheric O3 Column due to Fire
Emissions (July 15-25, 2004)
Strong Perturbation in Tropospheric
Composition and Chemistry over large
parts of the Northern Hemisphere
Pfister et al., 2005 2006
Pfister et al., 2006
5
Rasch etalCan lack of soot injection into UTLS
region explain cold bias at Summer pole?
6
The Chisholm Fire 29 May 2001
Stratospheric smoke 12.5 km (wind corrected)
7
TOMS Aerosol Index analysis

• Aerosol index of plumes associated with most
  intense forest fires 12-30

plume height assumed at 6km
Most palatable aerosol properties
plume height assumed at 10 km
8

• Analysis uses single particle soot photometer
• Uncertainty in mass 50
• Accuracy very significant improvement over wire
  impactors (estimated uncertainty of 1000)
• Previous estimates (eg IPCC-TAR and references
  therein) suggested aerosol concentrations of lt 1
  ng/m3

9
Model estimate using Liousse (1996) BB emissions
10
SP
NP
Ensemble Mean of 23 models
20 of 23 IPCC-FAR models have cold bias at NP
exceeding 4.5K during JJA
11
Randerson et al, BGC, 2005.
Pfister et al, GRL, 2005

• Inverse model study
• Emissions constrained by MOPPIT CO data
• Rescaled using Andreae and Merlet Emission Factors

12
Temperature Change 2 days
13
Experiment 2 Sustained background emissions by
Forest fires

• Goal to examine the impact of continuous seasonal
  emissions of forest fires on model climate.
• Emissions will have impact at time of injection
  (summer/fall), but lifetime is long enough that
  absorbing aerosols will remain in spring, can act
  to warm polar vortex, change ozone destruction
  also.
• emission designed to
• Raise Absorbing Aerosol in lower stratosphere to
  10-50 ng/m3
• Injection altitude around 12km (100-300 hPa)
• Compare to injection at surface
• Used Randerson emission inventory 1998.
• Assume 30 of emission injected into UTLS
• Note that this would be more like 6 of an
  inventory produced by an inverse method!

14
JJA Average Temperature Change
15
Pfister etal study

• Compare model behavior
• Summer 2000 (low fire year)
• Summer 2004 (very high fire year)
• Evaluate
• Broadband flux retrievals from CERES (long and
  shortwave)
• Aerosol optical depth (AERONET, MODIS and MISR)
• Explore consequences of changes in
• Meteorology
• Emission factors

16
wildfires in alaska
MOPITT CO at 850 mbar forJuly 2002, 2003 and 2004
2004 - record fire season in AK burned area
6.6e6 acres in AK 4.5e6 acres in Yukon
17
comparing AOD from MODIS, MISR and CAM

• Model underestimates observations over the fire
  region, and also outside the region and time
  period of the fires.
• MODIS and MISR also differ with MODIS gt MISR

18
clear-sky TOA fluxes 2004 vs 2000
CAM
CERES
Longwave
-4.4?2.5 Wm-2
-6.6?2.6 Wm-2
Shortwave
-2.5?5.6 Wm-2
-3.5?2.5 Wm-2
19
RF for individual components
Model Estimate for RF due todifference in Tsrf
for 2004 and 2000

• Longwave effect mostly explained by higher
  Tsrf in 2004

Model Estimate for RF due to carbonaceous
aerosols from fire
-6.7?2.6 Wm-2

• Shortwave effect mostly due to BC and OC
  from fires

-4.4?2.4 Wm-2
20
Exploring aerosol emission uncertanties

• Size of symbol indicates AOD
• Color indicates net RF (also shown by diagonal
  lines
• Obs suggest forcing in this region order 2W/m2
• To stay near obs requires either more BC or more
  absorbing OC

21
Conclusions

• Monthly emission smooth the model response to
  emissions, but it is always low compared to
  AERONET, MODIS, MISR
• TOA cooling over the 2004 Alaska fire region
  mostly due to higher surface temperatures and
  carbonaceous aerosols emitted from the fires.
• Simultaneous observations and model simulations
  of TOA fluxes and aerosol loading can be used to
  constrain aerosol emissions.
• Model uncertainties in assumptions of aerosol
  optical properties, transport and removal
  processes and in observations place large error
  bars on results.
• Need for additional observations of aerosol
  speciation and optical properties of biomass
  burning aerosols.

22
Flanner et al study

• First study that treats coupled snow aerosol
  heating, snow aging, meltwater scavenging in a
  climate model

23
Soot in Snow
24
Temperature response to Forest fires
25
Conclusions

• 1998 showed higher forcing than 2002 by 17.
  Temperature response factor of 3 higher
• Efficacy (temperature response to forcing
  compared to CO2) is about 3.
• Disproportionate response due to
• Ability to warm snow, and prime it for earlier
  melt (most important effect)
• Enhanced snow grain metamorphasis, which also
  darkens snow, plus photons penetrated further
  into snow
• Impurities move to surface during melt.

26
Aerosol Forecasting

• Can be used to
• influence field project
• help interpret results
• Successful in
• INDOEX
• ACE-Asia
• PACDEX (last week)

27
Extras
28
summer 2004 fires in alaska
How much pollution is coming from the fires?
What is the local, regional and global impact on
atmospheric composition ?
What is the radiative impact of the fires?
29
Aerosol Optical Depth (500nm)
Day 2
Day 5
Day 10
Day 1
30
Changes in Solar Flux day 2
Surface
Top of Atm
31
Radiative Forcing of the 2004 Alaska Fires

• Gabriele Pfister
• Atmospheric Chemistry Division

32
The Chisholm Fire, Alberta, May 28-29, 2001
May 28 1230 UT
May 29, 02 UTC AVHRR RGB
33
In the beginningthe earliest evidence
POAM SAGE
Sweden lidar
July 9-18 1998
34
Whole Atmosphere CAM (WACCM) Ozone January vs
Observed Ozone
Sassi et al 2005
Observed
height
latitude
Because of Summer Cold bias Ozone Depletion
persists through summer
35
Ozone -gt Temperature -gt Zonal WindFeedbacks are
important
Sassi et al 2005
36
Model Configurations

• Radiation sees standard aerosols plus two new
  aerosol types
• Standard Aerosols are prescribed for this run
• Sulfate, Dust, Sea Salt, Black and Organic Carbon
• New aerosols
• black carbon from forest fires injected between
  100-300 hPa as a hydrophobic aerosol
• ages to a hydrophilic aerosol with 1.5 day
  e-folding time.
• Hydrophilic form aggressively scavenged as both
  CCN and IFN

37
Experiment 1 Short Intense Forest fire

• Goal to reproduce the areal extent and optical
  depth of an intense pyro-cumuli event. Examine
  impact of plume on model evolution over a short
  time.
• Typical intense plume has
• Optical depth 1, single scattering albedo of
  0.6-0.8
• Areal extent 5 degrees latitude by 10 degrees
  longitude
• Plume located at order 12km (100-300 hPa)
• All emissions used Randerson climatology for
  Norman Wells fire event during August 1998.
• All aerosol emissions for a month were injected
  in a 5x10 degree region at 65N in 1 time step at
  start of run
• Randerson emissions were multiplied by a factor
  of 5 to produce a reasonable starting aerosol
  optical depth!

38
Experiments used the Community Atmosphere Model
(CAM)

• Component of NCAR CCSM (Community Climate Systems
  Model), a coupled ocean, atmosphere, land, sea
  ice model that now includes various options for
  more elaborate physical representations (eg
  isotopic fractionation H, O, C), chemistry,
  biogeochemistry (N and C cycles).
• http//www.ccsm.ucar.edu/models/atm-cam
• Can run as standalone model or as a component
  model
• General Circulation Model (GCM)
• Chemical Transport Model (CTM)
• Aerosol formulation follows Barth et al (2000),
  Rasch et al (2001), Collins et al (2001), Rasch
  et al (2001), Mahowald et al (2005), Collins et
  al (2003)

39
Conclusions

• Injection of black carbon into UTLS a viable
  mechanism for reducing longstanding bias in
  virtually all GCMs
• Temperature
• Cirrus formation
• Chemistry
• Dynamical feedbacks
• Amplitude of emissions is probably underestimated
  in most forest fire inventories
• Mechanism for lofting of aerosol should be
  handled explicitly in models

• Variety of instruments that can help us nail this
  down. Instruments of this class include
• SAGE
• POAM
• MISR
• TOMS
• CALIPSO
• MODIS
• More in-situ measurements using instruments with
  low detection limits (ltlt 100ng/m3)
• Frequency of occurrence of fires
• Altitude of injection
• Areal extent
• Thickness of plumes
• PDF of optical thickness

40
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
41
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
42
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
43
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
44
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
45
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
46
work flow
Inverse Modeling of CO Fire emissions using
MOPITT CO and MOZART
Derive emissions for other trace species and
aerosols by scaling using ER from literature.

• CAM-Chem Simulations (offline)
• Simulations for 2000 (reference year, low
  fire activity) and 2004 (fire year)
• Simulations with individual emissions off
• Range of emission scenarios for OC BC

MODIS MISR AOD
CERES TOA LW SW Fluxes
Separate radiative impacts of Tsrf,fire ozone
and aerosols
Constrain emissions and ER for OC and BC
47
Ensemble mean of 23 models
20 of 23 IPCC-FAR models have cold bias gt 7K at
SP during DJF
48
Zonal average Temperature difference between
experiment and control (DJF)
Injection in UTLS
Injection at Surface
49
comparing AOD from MODIS, MISR and CAM