Transcript ppt

A statistical method for calculating the
impact of climate change on future air quality
over the Northeast United States.
ozone episode
June 24, 2003
Collaborators: Cynthia Lin, Katharine Hayhoe, Edwin Maurer, Christian
Hogrefe, Pat Kinney, Daniel Jacob
Relationship of meteorology and ozone episodes
Northeast
Average num. of days
Number of summer days with 8-hour average
ozone > 84 ppbv over northeast U.S. sites
1988, hottest
on record
Probability
of ozone exceedance
vs. daily max. temperature
Curves include effects
of T-dependent biogenic
emissions, stagnation,
and chemistry.
Lin et al., 2001
Curves represent the total derivative d[O3]/dT, the sum of partial derivatives
(dO3/dxi)(dxi/dT), where x is the ensemble of ozone forcing variables that are
temperature-related.
A statistical method to study effects of climate change on air quality.
Idea: Use probabilities of ozone exceedance + daily GCM maximum temperatures to
predict number of exceedance days for each summer in future.
Step 1. Find exceedance probability for each model day’s maximum temperature
Step 2. Sum up probabilities for ensemble of summer days to get total exceedances
for that summer.
Summer Temperature Anomaly (oC)
10
CCSM 1
CCSM 2
CCSM 3
CCSM 4
ECHAM 1
ECHAM 2
ECHAM 3
8
6
PCM 1
PCM 2
PCM 3
HadCM3
GFDL 0
GFDL 1
GISS
1.
2.
+
4
Observed probability
of ozone exceedance
vs. daily max. temperature
2
0
-2
A1
-4
1900
1950
2000
2050
2100
Future temperature change over
Northeast 1900-2100, calculated by
many global climate models
Lin et al. 2001
= future smog episodes
Assumptions:
•
Emissions of ozone precursors remain constant over time.
•
For each region, the suite of conditions that lead to high ozone levels
do not vary over time.
Models and scenarios:
Statistically downscaled temperatures from the GFDL, PCM, Hadley
Centre models.
A1fi and B1 scenarios.
A1fi
Trends in CO2
emissions, 2000 to 2100
B1
Statistically downscaled temperatures: use observations to train GCM.
1. Interpolate observed 1961-1990 temperatures onto 1/8o grid. Calculate
monthly mean and variability at each gridpoint.
2. Apply bias correction to GCM temperatures and apply observed probability
density functions so that model matches observations.
T downscaled (x,t) = T GCM gridsquare (t) + a + b(x) + c(x,t)
3. Apply these same fixes to future GCM data: possible problem?
Obs 1990s
JJA daily
max Temp
Regional
model,
driven by
GCM met at
boundaries
Statistically
downscaled
GCM output
GCM
1990s
Hayhoe et al., 2006. Apologies for unclear temperature scales.
Frequency distributions of present-day daily maximum temperatures.
Hadley
Obs
“Historical” means present-day.
GCM,
Northeast U.S.
GFDL
GCM data for 1961-1990: PCM,
Hadley, and GFDL.
PCM
Observations are for 1980-1998.
Obs
Downscaled,
Northeast U.S.
models
Downscaling makes GCM output
look nearly perfect.
Mismatch at high end noted in
Hayhoe et al., 2006, but may also
be due to our spatial averaging. We
are investigating.
Summertime exceedances averaged over the Northeast, calculated using
daily max temperatures statistically downscaled from three GCMs.
Exceedance days averaged over the
Northeast:
- Increase by 10-30% by 2020s
- Double by 2050
- Increase beyond 2050 depending
on climate scenario and on model
A1Fi
CMAQ
B1
observations
Caveats:
- Assumption is that anthropogenic
emissions remain constant.
- Method does not capture high
ozone episodes during recent past:
we are investigating.
Use same technique for PM2.5?
Lin et al., 2007
Deterministic approach:
Statistical approach:
Advantages: day-by-day calculations,
allows for changes in anthropogenic
emissions of ozone precursors and for
detailed diagnosis of causes of ozone
change (transport, VOCs, boundary layer
height, clouds. . .)
GCM
met
met
BC
Advantages: fast, takes into account
many factors that affect ozone levels,
based on observations, allows for easy
model intercomparison across many
years, good for near-term outlook.
GCM
Regional
climate
model
Daily max Ts
fields
statistical downscaling
downscaled met
Global
chemistry
model
Regional chemistry
chem model (i.e., CMAQ)
Probabilities of ozone
exceedances
BC
Exceedance = f (maxT)
FUTURE AIR QUALITY
Frequency distributions for daily max temperatures:
comparison between future model and observed present-day
~2100 GCM
temperatures
A1 scenario,
Northeast U.S.
Frequency distributions are
for 2070-2099 model results.
Obs are from 1980-1999
present-day obs
~2100 downscaled
temperatures
Downscaled GCM output
does not capture the change
in temperature probability
distributions.
This could be a problem for
calculating air quality.
present-day obs
Advantage of method: Quick gauge of
ozone sensitivity to climate and of climate
penalty.
Limitations: Assumes constant emissions
of ozone precursors, and unvarying
temperature + met variable relationships
Also may not capture variability of
exceedances.
Emissions of carbon dioxide across the 21st century for
a range of scenarios.
Scenarios used
by UCS:
A1F1
B1
Unscaled GCM results are consistent with our previous research
with GISS GCM:
Hotter maximum temperatures
Triangles indicate
days of highest
pollutant
concentrations.
2050s
2000s
Reduced cloud cover
Our research shows that higher maximum temperatures are linked to longer
stagnation periods in future climate.
Reduced meridional temperature gradient
+ Increased eddy transport of latent heat
fewer cold fronts + more persistent heat waves