Transcript PPT - Atmospheric Chemistry Modeling Group

Wildfire in the western United States in the mid-21st
century and consequences for air quality:
Results from an ensemble of climate model projections.
Loretta J. Mickley
Xu Yue
Jennifer A. Logan
Jed Kaplan, Univ of Geneva
NASA Air Quality Applied
Sciences Team serves as a
bridge between NASA
science and environmental
management.
Millions of people in US live in areas with unhealthy
levels of ozone or particulate matter (PM2.5).
Number of people living in areas that exceed the national
ambient air quality standards (NAAQS) in 2010.
PM2.5
24-hour average
or annual average
Ozone
daily maximum
8-hour average
Bars on barplot will change with changing emissions.
Climate change could also change the size of these bars, by changing the dayto-day weather.
SO2 -- sulfur dioxide
NOx -- nitrogen oxides
Life cycle of particulate
matter (PM, aerosols)
precursor gases
ultra-fine
(<0.01 mm)
nucleation
Soup of
chemical
reactions
fine
(0.01-1 mm)
. . coagulation
.
. . .
condensation
Organic
carbon
SO2
NOx
SO2
NOx
VOCs
combustion
volcanoes
NOx
VOCs
cloud
(1-100 mm)
cycling
coarse
scavenging
(1-10 mm)
Black
carbon
VOCs
VOCs
NH3
VOCs -- volatile
organic compounds
NH3 -- ammonia
NOx
NOx
agriculture
biosphere
wildfires
combustion
soil dust
sea salt 3
Effects of wildfires on air quality in cities in Western US can
be very dramatic.
 Hayman fire, June 8-22, 2002
 56,000 ha burned
 30 miles from Denver and Colorado Springs
Unhealthy air quality in Denver
June 8, 2002
PM10 = 40 μg/m3
PM2.5 = 10 μg/m3
June 9, 2002 PM10 = 372 μg/m3
PM2.5 = 200 μg/m3
Standard = 35 µg/m3
Colorado Dept. of Public Health and Environment
Vedal et al., 2006
Fire activity had a big impact on
California air quality in 2013.
Timeseries of 3-hour average PM2.5
concentrations in Foothills Area
PM2.5 (mg m-3)
Rim Fire
Hazardous levels > 250 mg m-3
Very unhealthy
Aug 28
August 20
August 31
Will fire change in the future climate?
Unhealthy air
Aug 30
Very unhealthy air
Observations suggest that fires are increasing in North America.
obs temperature
area burned
5 yr means
1960
Area burned in Canada has
increased since the 1960s,
correlated with temperature
increase.
Gillett et al., 2004
2000
Increased fire frequency over the
western U.S. since 1970,
related to warmer temperatures
and earlier snow melt.
1970
2000
Westerling et al., 2007
IPCC AR4 models show increasing temperatures across North
America by 2100 in A1B scenario.
Change in surface temperatures in 2100, relative to present-day.
Results for precipitation changes are not so clear.
IPCC, 2007
Future regional predictions for precipitation show large variation
across North America.
Percent change in 2100 precipitation relative to present-day
Annual
DJF
JJA
Number of models showing increasing precipitation
most
models
few
models
IPCC 2007
How do we predict fires in a future climate?
We don’t have a good mechanistic approach for modeling wildfires.
Start with the past.
Relationship between observed
meteorology + area burned
+
Future
meteorology
2 approaches
Future area
burned
Use ensemble of
climate models to
gain confidence in
prediction.
JJA Temperature
increase by 2100
Approach 1. Regression method.
Regress meteorological variables and fire indexes onto annual mean area burned
in each of six ecoregions with a stepwise approach.
RMF
ERM
PNW
NMS
CCS
DSW
Ecoregions are aggregates of those in
Bailey et al. (1994)
Identify the meteorological
variables and fire indexes that
best predict area burned.
Include lagged met variables.
Best predictors: Temp, RH, precip, Build-up Index, Drought code, Duff
moisture code.
E.g., Area burned in Nevada/ semi-desert = f ( + T summer max that year
+ RH and rainfall previous years)
Predicted fires match observed area burned
reasonably well. Least best fit is in Southern
California.
Obs
Fit
Area burned in many ecoregions depends on
previous year’s relative humidity, rainfall, or temp.
RMF
ERM
PNW
NMS
CCS
DSW
Yue et al., 2013
Approach 2. Parameterization method.
Parameterize daily area burned in each grid square as a function of that day’s
temperature, relative humidity, and rainfall.
Area burned = f1(Temp) x f2(RH) x f3(rainfall)
ì a × T× (1.0 - RH) 2
ï
if T > Tt and R < Rt
ln( AB) = í Tt × ( R + 0.2)
ï no fire
else
î
Fire potential coefficients are (α) chosen to match long-term area
burned for each ecoregion.
Scheme can predict seasonality of fires on the 1°x1° grid.
Parameterization works for forested regions,
but not for regions where meteorology from
previous seasons matter.
Obs
Fit
Yue et al., 2013
Start with the past.
Relationship between observed
meteorology + area burned
+
Future
meteorology
Future area
burned
Use of an ensemble of 15 climate models
improves confidence in the results.
Changes in 2050s climate in the West.
• Temperature increases 2-2.5 K.
• Changes in precip and relative humidity
are small and not always robust.
Next step: apply meteorology from climate
models to the two fire prediction schemes.
Yue et al., 2013
Wildfire area burned
increases across the
western United States by
the 2050s timeframe.
Results from regressions approach.
Shown are median results.
Yue et al., 2013
Relationship between observed
meteorology + area burned
+
Future
meteorology
Future area
burned
1986-2000
2051-2065
+40%
1986
2065
Obs
+20%
Models predict increases in
area burned of 20-100%
across the West, depending
on ecoregion.
Area burned (ha)
Median of models
spread of models
100%
+60%
Results are from regressions.
+70%
The models match the mean area
burned fairly well in present-day.
+60%
Yue et al., 2013
Median model results show an increase in
area burned in all regions, but some scatter.
Median changes:
Forest
Forest
Ratio of area burned
Ratio of future to present-day area burned
median
40-70% increase in
forested regions
60% increase in grasslands
Doubling in Southwest
RMF
ERM
PNW
NMS
CCS
DSW
Yue et al., 2013
Parameterization method shows similar increases in area
burned, and also provides seasonality.
Units = 104 hectares
future
present-day
X4 increase
X2 increase
Predicted area burned
shows large increases in
2050s during peak months.
Yue et al., 2013
Parameterization method also reveals a lengthening of
wildfire season.
observations
Predicted number of days of fire
season across the Western US.
Mean increase is ~3 weeks.
present-day
future
Fire season is defined as those days when daily area burned
is larger than 100 ha in at least one grid box.
Yue et al., 2013
What about trends in vegetation in the future climate
and the issue of reburning?
Results from LPJ dynamic vegetation model show relatively little change in
vegetation in 2050s atmosphere.
We placed fires randomly within each sub-ecoregion to minimize reburning
of gridsquares.
How will changing area burned affect air quality?
Ensemble of
climate models
Median area
burned
Emissions = area burned x fuel
consumption x emission factors
Future
meteorology
GEOS-CHEM
Global chemistry
model
Future air
quality
Organic particles increase in future atmosphere over the
western U.S. in summer, especially during extreme events.
D Organic Carbon, OC
D Black Carbon, BC
Change in summertime mean PM2.5 in ~2050s,
relative to present-day.
Yue et al., 2013
Organic particles increase in future atmosphere over the
western U.S. in summer, especially during extreme events.
Cumulative probability of daily mean
concentrations of organic particles
2050s
doubling
Presentday
Distribution of daily OC
concentrations, May-Oct in
Rocky Mountains.
Visibility decreases during
extreme events from 130
km to 100 km.
These changes are
underestimates due to
model resolution.
Yue et al., 2013
What do these increases in wildfire aerosol mean for
human health?
% area
burned
% OC
particles
Ongoing project with Yale will look at health
impacts of these increases.
Yue et al., 2013
How do we improve fire predictions in S. California?
The largest fires in California are associated with Santa Ana events.
Fire plumes (Oct. 2007)
Composite Santa Ana winds
Need finely resolved wind fields to
capture Santa Ana in meteorological
record.
Hughes and Hall (2010)
Area burned
Improving predictions of area burned
in Southern California.
• Divide up California into smaller
ecosystems
• Use finer grid
• Parameterize effects of elevation,
population, and Santa Anas.
1980-2009 fire occurence
Santa Ana winds are associated with
surface pressure anomalies that can
be captured by climate models.
Surface pressure anomalies
Ratio of future area
burned to present-day.
Area burned increases
by 50-100% in
Southwest California.
R
P
2 methods
Yue et al., in review
Main findings.
• Area burned by wildfires increases 20-170% across the West by the
2050s, depending on the ecosystem.
• During peak months, area burned nearly doubles in Eastern
Rockies/Great Plains and quadruples in the forested Rockies.
• Wildfire season may increase by ~3 weeks in the West in future.
• Increased smoke from wildfires may thwart efforts to regulate air
quality in coming decades. This is a climate penalty.
Haze over Boston on May 31, 2010
Wildfires in Quebec the same day.
Seasonality of fires in Southern California
Fire regions
South-West Cal.
area
num. fires
Central Western Cal.
Largest area burned in SW California.
October peak associated with the Santa
Ana winds, which are underestimated by
large scale models as they lack the
detailed topography: need large-scale
approach
Sierra Nevada
New parameterization predicts yearly variability and
seasonality in south west California
Southwest CA
R2=0.64
Seasonality
Area burned in ~2050 / Present-day
Area burned in Southern
California increases 20-100% by
2050s relative to present-day.
R
P
South west
California
R
Central
California
R
P
Two approaches used in each
ecoregion.
Sierra Nevada
Yue et al., 2013