Adverse Effects of Drought on
Air Quality in the US
Yuxuan Wang, Yuanyu Xie, Wenhao Dong,
Yi Ming, Jun Wang, Lu Shen, Zijian Zhao
3 May 2017
IGC8
Drought: complex climate extreme
Drought
Low
precipitation
High
temperature
Land
surface
Chemistry
&
deposition
Fires
Atmospheric Composition
2
Drought is a quite ‘frequent’ extreme in the US
• SPEI drought index
(Standardized Precipitation
Evapotranspiration Index)
• Drought longer than one
month (1-month SPEI < -1.3)
Mar-Oct
1990-2014
• Focus on regions with at
least 10% drought (20
months)
%
Data and method
Surface observations:
• Ozone (MDA8): EPA, CASNET;
• PM2.5: EPA, IMPROVE
• 1990-2014; March to October (growth season)
• Daily mean averaged to monthly
GEOS-Chem simulation:
• MERRA 2o x 2.5o
• 1990-2014, Monthly mean
• Year-to-year changes of US anthropogenic emissions (Xing et al.,
2013)
• GFED4 monthly fire emissions
Drought - Pollution Relationship:
• Air pollution data were detrended and deseasonalized by
removing 7-year moving averages from each month
Observed correlation between SPEI and Pollution
Slope of linear correlation at sites with more than 10% drought
Ozone
negative
More pollution
PM2.5
positive
negative
Less pollution
More pollution
positive
Less pollution
Negative correlations throughout the US: higher O3 and PM2.5 at the
surface with increasing drought severity
Wang et al., ACPD, 2017
GEOS-Chem simulated drought effects
Slope of linear correlation at sites with more than 10% drought
Ozone
Ozone
Obs
Model
PM2.5
PM2.5
Obs
Model
• Model captures the general relationship between ozone and SPEI
• Model is not able to capture the negative correlation between PM2.5 and
SPEI
Performance of climate-chemistry models
Observations
NCAR-CAM3.5
GISS-E2-R
Ozone
PM2.5
Climate-chemistry model outputs from ACCMIP archive
Wang et al., ACPD, 2017
Drought effects on ozone



Higher temperatures and lower cloud covers lead to higher rates of
ozone production (most models get this)
Drought reduces dry deposition of ozone (Kavassalis and Murphy,
2017; Huang et al., 2016) (models are lacking this response)
Response of biogenic emissions to drought is highly uncertain
Obs isoprene (regrid to model)
Collocated Model (N=2388)
EPA’s PAMS
network
Sparse
observations
•
•
drought
SPEI
drought
SPEI
Isoprene
increases with
drought (except
for Great Plains)
Model isoprene is
not sensitivity to
drought (except
for SE US)
Drought effects on aerosol species
Observed drought anomalies (drought minus normal)
 Organic aerosol (OA)
shows the largest
response to drought
in all regions
 Sulfate is the second
largest
 Dust change is
evident in the West
and Great Plains.
OA - drought relationship
Slope of linear regression between SPEI drought index and OA
Obs


Model
Observed OA-SPEI slope is positive over all regions, with larger
sensitivity over northwestern and southeastern US.
Model captures the sign and spatial pattern of the slope, but largely
underestimates the slope
Sulfate – drought relationship
Slope of linear regression between SPEI drought index and sulfate
Obs



Observations show sulfate increase
during drought, due to lower wet
deposition and higher gas-phase
oxidation
Model shows a large decrease in the
South
Model has excessive reduction in
aqueous phase SO4 in the south
Model.
Model aqueous production of SO4 vs SPEI
Conclusion

Observations show significant adverse effects of drought on ozone
and PM2.5 air quality:
3.5 ppbv (8%) for ozone and 1.6 μg m-3 (17%) for PM2.5 during drought
compared to normal conditions
 Models (CTM and ACCM) reproduce the direction of changes for
surface ozone, but not for PM2.5
 Model deficiencies include but not limited to:
 Too weak a response of surface emissions (e.g. BVOCs) to drought
 Lack of dry deposition response
 Aerosols: relative response of gas-phase vs. aqueous phase to drought

Future projection: 1-6% increase for ground-level O3 and 1-16% for
PM2.5 in the US by 2100 (compared to the 2000s) due to increasing
drought alone
Reference:
Wang, Y., Y. Xie, W. Dong, Y. Ming, J. Wang, L. Shen, Adverse Effects of
increasing drought on air quality via natural processes, Atmos. Chem. Phys.
Discuss., doi:10.5194/acp-2017-234, 2017
Back-up slides
Model Evaluation: 1990-2015 (Mar to Oct)
mean distribution
Ozone (ppbv)
Obs.
Mod.
Mod. – Obs
Mod.
Mod. – Obs
PM2.5 (ug/m3)
Obs.
Model simulated ozone is within 10% of observed value over most regions
• ~ 10 ppb underestimate at California
• ~ 10 ppb overestimate at southeast US
Model simulated PM2.5 is ~20% lower than observations
• ~ 5 ug/m3 underestimation at southeast US
Model Evaluation
Correlation r between model and observations (1990-2014)
Ozone
PM2.5
• Grids with less than 5 years data are excluded
• Model simulated ozone correlates well with observation
(r > 0.7 over most regions)
• Model simulated PM2.5 show moderate correlation with
observation (r~0.3-0.7)
Pollution enhancements caused by drought
Drought effects on ozone
Comparison between model simulated and observed isoprene
Obs
Observation
(regrid to model)
Mod
Model collocates
with obs (N=2388)
Corr r