Projecting Impacts on Southeastern U.S. Air Quality U. Shankar1, J. Prestemon2, D. McKenzie3, E. Stavros4, K. Talgo1, A. Xiu1, B. H. Baek1, M. Omary1, and D. Yang1 Regionally#Relevant#Wildfire#Modeling#for#the#U.S.# Background AAB Projections 2010-2060 2010 Annual PM2.5 Performance The increasing frequency and intensity of over the past few decades has greatly concerned land and air quality managers, due to the Lightning-caused Human-caused FSJVA_base10b_12k − State: NEI All, Network: IMPROVE, Species: PM25_TOT FSJVA_base10st_12k − State: ST All, Network: IMPROVE, Species: PM25_TOT FSJVA_base10dy_12k − State: DY All, Network: IMPROVE, Species: PM25_TOT Forest#Service#and#Bureau#of#Land#Management# 125 125 125 125 125 125 ● JAN IMPROVE ● JAN IMPROVE ● JAN IMPROVE FEB FEB FEB environmental and health impacts, as well as the potential impacts on the radiation budget due to the short-lived climate forcing pollutants produced in MAR MAR MAR APR APR APR 100 MAY 100 100MAY 100 100MAY 100 these fires, and climate change is increasingly seen as a driver. The U.S. Forest Service (USFS), in its current projections of the condition of rangeland JUN JUN JUN • Helped#USFS#develop#staAsAcal# JUL JUL JUL AUG AUG AUG SEP SEP SEP 75 75 ● 75 75 75 ● 75 75 75 ● 75 and forests required by the Resources Planning Act (RPA) of 1974, has included the impacts of climate change on wildfires in its current assessment. Regionally#Relevant#Wildfire#Modeling#for#the#U.S.#● OCT ● OCT ● OCT NOV NOV ● NOV models#to#project#annual#acres# DEC DEC ● DEC ●

USFS has developed statistical models of annual areas burned (AAB) over the Southeast at the county level, to project their trends from 2010-2060. 50 50 50 burned#from#2010#G#2060#using# 50 50 50 50 50 50 Normalized Mean Error (%) Normalized Mean Error (%) Normalized Mean Error (%) Normalized Mean Error (%) These projections account not only for changes in regional meteorology, downscaled from an ensemble of nine climate simulations, but also changes in Forest#Service#and#Bureau#of#Land#Management#35 Normalized Mean Error (%) 35 Normalized Mean Error (%) 35 land use, population and economic growth patterns over the region, consistent with the greenhouse gas (GHG)climate#model#inputs#and# emission scenarios used in those 25 25 25 25 25 25 socioeconomic#projecAons#for# • Helped#USFS#develop#staAsAcal#0 0 0 0 0 0 simulations. This study used these AAB projections in a stochastic model of fire generation that disaggregates them into daily areas burned, to estimate −75 −60 −45 −30 −15 0 15 30 45 60 75 −75 −60 −45 −30 −15 0 15 30 45 60 75 −75 −60 −45 −30 −15 0 15 30 45 60 75 fire emissions needed to drive air quality simulations in the present (2010), and in selected future years, and analyzedthe#Southeast# the regional air quality impacts for models#to#project#annual#acres# Normalized Mean Bias (%) Normalized Mean Bias (%) Normalized Mean Bias (%) the 2010 baseline and future years. This work is supported by USFS Joint Venture Agreement # 11-JV-11330143-080. burned#from#2010#G#2060#using#PM2.5 model performance against IMPROVE is nearly identical for the ST- • Used#these#to#constrain# Lightning#(L)G#and#human#(R)Gcaused#burned#areas# and DY-based fire emissions relative to simulations with the NEI wildfire PM2.5 summer FSJVA_base10dy_12k (State: Southeast) AABs in the Southeast from human-caused fires seenclimate#model#inputs#and# to be ~ 3

stochasAc#models#of#12 daily#acres# inventory, with slightly better summertime performance, and slightly worse Othertimes those from lightning-caused fires, and decreasing in the future Project Goal and Objectives Conceptual Modeling System PM2.5 summer FSJVA_base10dy_12k (State: Southeast) socioeconomic#projecAons#for# PM2.5OC summer FSJVA_base43dy_10_12k (State: Southeast)

PM2.5 summer FSJVA_base10dy_12k12 (State: Southeast) burned#over#the#Southeast#to# PM2.5 summer FSJVA_base48dy_10_12k (State: Southeast) PM2.5 summer FSJVA_base58dy_10_12k (State: Southeast) performance in the fall. Wintertime performance (when there are no fires) is PM2.5 summer FSJVA_base53dy_10_12k (State: Southeast) 12 EC Other 12 12 12 + Other 12 OC Other the#Southeast# NH4 OC Other Other Other

10 OC EC − OC OC OC identical. GHG NO EC NH + EC 3 4 + EC esAmate##consumpAon#and# 10 + EC − NH EC 2− NH NO 4 + + 10 4 3 − + 10 Wildfire − PM EmissionNH4 NH TrendsNH SO 4 10 NO 4 Global climate 2− 10 − Emission 4 NO3 10 3 − − SO4 2.5 2− NO3 NO NO 2− • 3 3 Used#these#to#constrain# Goal: To assess the impact of climate change on wildfire SO SO4 4 2− 2− 2− SO4 SO SO fire#emissions#to#project#daily# 4 4 Lightning#(L)G#and#human#(R)Gcaused#burned#areas# 26.9% −− Scenarios 26.9% −− 8 PM2.5 summer FSJVA_base10dy_12k (State: Southeast)

8 26.9% −− 8 # of sites: 38 8 #

8 Annual PM Composition Trends (DY) # of sites: 38 33.3% −− 8 # of sites: 38 # of sites: 38 8 3

activity, emissions and air quality in the Southeast. 12 33.3% −− 33.3% −− # of obs: 863 # of sites: 38 # of sites: 38 # ofstochasAc#models#of# sites: 38 daily#acres# 2.5 33.3% −− 33.1%# −−of obs: 863 33.3%# of −− obs: 863 33.3% −− Downscaling # of obs: 863 # of obs: 928 # of obs: 863 # of obs: 863 Other air#quality#in#selected#years#Chemistry 6.5% −− 49.8% −− Radiative Feedback of Aerosols & clouds 6.5% −− PM2.5 summer FSJVA_base10dy_12k (State: Southeast) 3.5%56.1% −− −− 51.3% −− 52.8% −− OC 6 3.5% −− PM2.5 summer FSJVA_base43dy_10_12k (State: Southeast) 6 PM2.5 summer FSJVA_base10dy_12k12 (State: Southeast) 6 g/m 6.5% −− 6 PM2.5 summer FSJVA_base48dy_10_12k (State: Southeast) PM2.5 summer FSJVA_base58dy_10_12k (State: Southeast) 6 & transport 6 burned#over#the#Southeast#to# PM2.5 summer FSJVA_base53dy_10_12k (State: Southeast) 12 EC Other 17.1% −− 13.1% −− 12

Objectives 12 12 µ 17.1% −− 17.1%13.1% −− −− + 12 Other 3.5% −− 17.2% −− 17.1% −− Other OC 2040#G#2060#6 17.1% −− NH Other Other Other FSJVA_base10dy_12k 4 OC

10 OC 3.9%FSJVA_base10dy_12k −− FSJVA_base43dy_10_12k − EC OC 3.9%3.1% −− −− RMSE: 4.65 FSJVA_base48dy_10_12k FSJVA_base53dy_10_12k FSJVA_base58dy_10_12k + EC OC OC Regional climate Fire GHGs 3.1% −− 3.1% −− 3.2% −−RMSE: 4.65 5.7% −−RMSE: 6.23 NO3 EC NH 4 RMSE: 6.12 4 + Concentration (ug/m3) Concentration 3.1% −− 3.1% −− EC RMSEs: 2.06 RMSE: 6.01 esAmate#fuel#consumpAon#and# 10 + EC RMSE: 6.03 4 EC 4 − Concentration (ug/m3) Concentration 13.1% −− (ug/m3) Concentration RMSEs: 2.06 RMSEs: 5.51 NH 4 2.7% −− 5.9% −− 2− NH 4 + + Concentration (ug/m3) Concentration NO 4 10 17.1% −− 12.5% −− 4 RMSEs: 5.01 5.9% −− 4 + Concentration (ug/m3) Concentration Concentration (ug/m3) Concentration 3 − RMSEu: 4.17 RMSEs: 5.27 RMSEs: 5.16 10 NH 12.5% −− 12.5% −− 6.4% −−RMSEu: 4.17 RMSEu: 2.92 SO − 4 NH NH4 Emissions 10 NO 4 12.5% −− 2% −− 2− 10 − 12.5%9.5%r: 0.57 −− −− 12.5%RMSEu: −− 3.52 2% RMSEu:−− 2.89 RMSEu: 3.12 4 NO3 10 SO 3 − − • Examine the impacts of changes in climate and 2.7% −− r: 0.57 r: 0.22 4 2− NO3 NO NO – 2.5% −− r: 0.19 2− 3 3 9.3% −− SO Wildfires#seen#to#decrease#slightly#2.5% −− 2.5% −− r: 0.25 r: 0.31 2− 9% −− SO 4 3.9% −− FSJVA_base10dy_12k 2.5%8% −− −− 1.4%2.5% −− −− 2.5% −− 4 SO 2− SO 2− 1.3% −− 4 SO4 4 Mixing 3.1% −− 1% −− 1.4% −− fire#emissions#to#project#daily# RMSE: 4.65 46.2% −− 26.9% −− 46.2% −− 26.9% −− 8 2

8 26.9% −− 2 2 8 # of sites: 38 8 4 2 LSFs #

Growth (ug/m3) Concentration 2 2 RMSEs: 2.06 8 # of sites: 38 33.3% −− 8 socioeconomic variables relevant for fire activity on annual # of sites: 38 # of sites: 38 8 due#to#socioeconomic#influences;#31.5% −− 31.5%31.5% −− −− 33.3% −− 3 33.3% −− # of obs: 863 # of sites: 38 # of sites: 38 # of sites: 38 12.5% −− RMSEu: 4.17 25.8%31.5% −− −− 31.5%31% −− −− 31.5%30.1% −−−− 33.3% −− 33.1%# −−of obs: 863 33.3%# of −− obs: 863 33.3% −− 28.9% −− # of obs: 863 # of obs: 928 # of obs: 863 # of obs: 863 air#quality#in#selected#years# 6.5% −− 49.8% −− r: 0.57 6.5% −− 2.5% −− 3.5%56.1% −− −− 51.3% −− 52.8% −− 6 3.5% −− 6

areas burned in the next five decades over the Southeast. 6 0 0 0 g/m 6 0 6.5% −− 6 0 PM#concentraAons#track#this#trend# 6 0 17.1% −− 13.1% −−

IMPROVE IMPROVEIMPROVECMAQ CMAQCMAQ 17.1% −− µ 17.1%13.1% −− −− 46.2% −− IMPROVE IMPROVECMAQ IMPROVECMAQ CMAQ 3.5% −− 17.2% −− 17.1% −− LSF – Land-surface feedback 2040#G#2060#6 17.1% −− FSJVA_base10dy_12k 2 3.9%FSJVA_base10dy_12k −− FSJVA_base43dy_10_12k Vegetation Wildfires Smoke 3.9%3.1% −− −− RMSE: 4.65 FSJVA_base48dy_10_12k FSJVA_base53dy_10_12k FSJVA_base58dy_10_12k 3.1% −− 3.1% −− 3.2% −−RMSE: 4.65 5.7% −−RMSE: 6.23 4 RMSE: 6.12 Concentration (ug/m3) Concentration RMSEs:3.1% 2.06 −− 3.1% −− RMSE: 6.01 RMSE: 6.03 4 4 Concentration (ug/m3) Concentration 13.1% −− (ug/m3) Concentration RMSEs: 2.06 RMSEs: 5.51 4 2.7% −− 5.9% −− 31.5% −− (ug/m3) Concentration 4 Not included in actual model 17.1% −− 12.5% −− 4 RMSEs: 5.01 5.9% −− Concentration (ug/m3) Concentration – 2010# 2043# (ug/m3) Concentration •Use the downscaled meteorology to project daily fire activity, RMSEs: 5.27 RMSEu: 4.17 RMSEs: 5.16 Now#refining#these#methods#to# 2048# 12.5% −− 12.5% −− 6.4% −− 2053# 2058# RMSEu: 4.17 RMSEu: 2.92 12.5% −− 2% −− 2% −− 2.7% −− 12.5%9.5%r: 0.57 −− −− 12.5%RMSEu: −− 3.52 RMSEu: 2.89 RMSEu: 3.12 – 2.5% −− r: 0.57 r: 0.22 r: 0.19 9.3% −− Wildfires#seen#to#decrease#slightly#2.5% −− 2.5% −− r: 0.25 r: 0.31 9% −− 3.9% −− FSJVA_base10dy_12k 2.5%8% −− −− 1.4%2.5% −− −− 2.5% −− realizaon 3.1% −− 1.3% −− 1.4% −− RMSE:46.2% −− 4.65 46.2%1% −− −−

and estimate fire emissions in selected years. include#decadal#fuel#load#changes# 2 2 2 4 2 Concentration (ug/m3) Concentration 2 RMSEs: 2.06 2 3 31.5% −− Mortality 0 due#to#socioeconomic#influences;#31.5% −− 31.5% −− 12.5% −− 31.5% −− 31% −− PM #concentraAons#(µg/m )#and## RMSEu: 4.17 25.8% −− 31.5% −− 31.5%30.1% −−−− 28.9% −− 2.5 r: 0.57 2000#G#2050,#and#a#more#IMPROVE CMAQ 2.5% −− 0 0 0 0

PM#concentraAons#track#this#trend# 0 • Examine the trends in fire emissions and air quality over the 0 consAtuent#mass#fracAons# 46.2% −− IMPROVE IMPROVEIMPROVECMAQ CMAQIMPROVECMAQ IMPROVECMAQ IMPROVECMAQ CMAQ

representaAve#set#of#years#for#the# 2 Southeastern U.S. in modeled years. – Now#refining#these#methods#to#31.5% −− 2010# 2043# 2048# Other natural sources Anthropogenic NEI emissions are much higher in spring, and much lower in summer for 2053# 2058# BLM#project#sources almost all years than for HIS, ST and DY cases. DY meteorology-basedinclude#decadal#fuel#load#changes# Total PM shows a slight decrease in the 2040-20603 period relative to

0 2.5 PM2.5#concentraAons#(µg/m )#and## AAB estimates yield the lowest emission values in all years2000#G#2050,#and#a#more# simulated.IMPROVE 2010.CMAQ The decrease in SO is mainly from reductions in the energy sector. Ensemble Model Setup representaAve#set#of#years#for#the# consAtuent#mass#fracAons#4 Summer DY Wildfire PM EmissionsBLM#project# (T) Conclusions Statistical models of AAB projections are built on original • Regional climate: Dynamically downscaled from CGCM3 A2 scenario, the only 2.5 2010 2043 2048 models of Mercer and Prestemon (2005) and related work. ensemble member available in the North American Regional Climate Change • Socioeconomic factors play a major role in decreasing future wildfires Statistically downscaled meteorological inputs to these models Assessment Program (NARCCAP) archive, using the Weather Research and • Projected PM2.5 emissions track AAB trends well, showing a slight come from a nine-member ensemble of 3 climate models, Forecasting (WRF) model at 12-km resolution over the Southeastern U.S. for decrease in future for all three methods of calculating daily fires each run with 3 GHG emission scenarios from the PRISM selected years: 2010, 2043, 2048, 2053 and 2058 • Projected emissions significantly differ from 2010 NEI in all seasons (Daly et al., 2002) database at 8’x8’ resolution. • Fuel Loads: Static, from Fuel Characteristic Classification System (FCCS) • DY estimates of daily AB produce lower fire emissions in all seasons General Circulation Models: • Fire activity: Fire Scenario Builder (McKenzie et al., 2006), adapted for the and years compared to the other methods Canadian General Circulation Model Version 3.1 (CGCM31) Southeast; disaggregated three ways to examine emission sensitivities, from: 2053 2058 • 2010 PM2.5 performance is very similar for all methods, implying that the Commonwealth Scientific and Industrial Research o AABs estimated using statistically downscaled meteorological inputs in all years fire emissions projection methodology is reasonably robust; however, Organization Mark 3.5 (CSIRO MK35) model 2010 - 2060 (labeled ST) more in-depth evaluation is needed for all pollutants Model for Interdisciplinary Research on Climate o AABs estimated using WRF model inputs to replace statistically downscaled Version 3.2 (MIROC32) inputs in the selected years (labeled DY) References o actual historical mean (1992-2010) AABs (labeled HIS) Daly, C., W. P. Gibson, G.H. Taylor, G. L. Johnson, and P. Pasteris, Emission Scenarios (from IPCC AR4): • Smoke emissions: BlueSky (USFS); dynamic plume rise in CMAQ 2002, Climate Research 22, 99-113. McKenzie, D., S. M. O’Neill, N. A. Larkin, and R. A. Norheim, 2006, A1B: Moderate population growth, high economic growth • Chemistry-transport: CMAQ v5.0.2; Carbon Bond 2005 (gas phase) and Future DY-based summertime emissions are seen to decrease slightly Ecol. Modell., 199, 278–288. A2: Moderate economic growth, high population growth AERO6 (aerosol) mechanisms; lateral boundary inputs from an existing 2010 relative to 2010 in Southeastern coastal areas and in West Virginia, Mercer, D. E., and J. P. Prestemon, 2005, Forest Policy and B2: Moderate economic growth, low population growth Continental U.S. simulation at 36-km resolution for the same mechanisms. although the trend is not uniform, consistent with AAB trends in these years. Economics 7, 782–795.

1 University of North Carolina at Chapel Hill — Institute for the Environment; 2 Southern Research Station, USFS; 3 Pacific Wildland Fire Sciences Laboratory, USFS; 4 NASA - Jet Propulsion Laboratory [email protected] http://www.ie.unc.edu/cempd