December 5-6, 2007
INDRU Kick-off Meeting
India International Center (IIC), New Delhi
Precipitation extremes in IPCC models
Masahiro Sugiyama
University of Tokyo, IR3S
[email protected]
Motivation:
Can we quantify damages from extremes?
If so, can we do better?
IPCC (2007, WG2, SPM)
Many economic assessments have relied on
changes in mean
—
—
—
Tol (2002) is one of comprehensive studies, but has a long list
of omitted costs, which includes “extreme weather”
Perhaps the closest example is studies on economics of
tropical cyclone. There are a few but only for USA (Nordhaus
2006; Sachs 2007, undergraduate thesis; Hallegatte 2007).
Their treatment of precipitation is rudimentary at best;
Interestingly, damage goes as sixth or higher power of winds
One reasons for lack of such studies is absence of
precipitation extreme projections, which this study tries to
address
Summary and implications
—
—
—
—
Precipitation extremes remain a difficult challenge for climate
models, although there are a handful of good models
Multi-model ensemble analysis shows that precipitation extremes
increase more than the corresponding mean, but not
necessarily at the rate expected from the Clausius-Clapeyron
relation; the reason for the divergence from the thermodynamic
constraint is not clear yet
Damage could increase even faster
At the local level, model discrepancies dominate and any
conclusive result is hard to draw; although it is desirable to utilize
climate extreme information for impact assessment, such analysis
is still full of uncertainty
Precipitation (even its mean) remains a challenge for
climate models
IPCC
(2007)
—
—
Precipitation changes for SRES A1B scenario for
(2080-2099) relative to (1980-1999)
Regions without stipples indicate that
more than 20% of models disagree on the sign of change
Global warming changes both mean and
variability of precipitation
—
—
—
—
Changes in precipitation variability (e.g., floods and droughts)
could matter more than mean rainfall
Some recent studies (e.g., Allen and Ingram 2002; Held and
Soden 2006) provide physical basis for understanding changes
in the global hydrological cycle, especially for mean
precipitation/evaporation
Can we predict changes in precipitation extremes on some
physical basis? At the local scale?
Here we focus on floods and do not consider droughts
Precipitation comes from water vapor in the
atmosphere; more moisture due to warmer climate
probably means more intense precipitation events
—
Under a constant relative humidity assumption, precipitable
water increases at the rate of saturation vapor pressure e*
(Clausius-Clapeyron relation)
*
1 de
(latent heat of vaporization )
=
~ 7 %/K
*
2
e dT
(gas constant )T
—
—
Since global-mean precipitation tends to increase 2%/K,
precipitation extremes are likely to increase faster than the
mean
Previous studies (Allen and Ingram 2002; Emori and Brown
2005; Pall et al. 2007) supported this view, but relied on a
limited set of models
Approach: Analysis of IPCC AR4 models
Acronym
Modeling center
bccr_bcm2_0
Bjerkness Center for Climate Research,
cccma_cgcm3_1_t63
Canadian Centre for Climate Modelling and Analysis, Canda
cnrm_cm3
Centre National de Recherches Meteologiques, Meteo-France
csiro_mk3_0
CSIRO Atmospheric Research, Australia
gfdl_cm2_0
Geophysical Fluid Dynamics Laboratory, USA
giss_aom
NASA/Goddard Institute for Space Studies, USA
ipsl_cm4
Institut Pierre Simon Laplace, France
miroc3_2_hires
CCSR/NIES/FRCGC, Japan
miroc3_2_medres
CCSR/NIES/FRCGC, Japan
mpi_echam5
Max Planck Institute, Germany
mri_cgcm2_3_2a
Meteorological Research Institute, Japan
ncar_pcm1
National Center for Atmospheric Research, USA
We can predict local temperature increase,
given a global temperature increase
—
SRES A1B and B1 scenarios
We can predict local water vapor increase,
given a local temperature increase
—
Consistent with the thermodynamic argument
Predicting precipitation at the local scale is hard
Allen-Ingram (2002) diagram:
Cumulative distribution function (CDF) of
precipitation derived from daily data
CDF(probability)
Switch axes
P
[mm/day]
P
[mm/day]
CDF
(probability)
Zoom
up
P
[mm/day]
CDF
(probability)
0
90
99
99.9
percentile
99.99
99.999
MIROC (CCSR/NIES/FRCGC) 3.2 hires
–
1-in-1-year precipitation event
~ 99.7th percentile
–
1-in-10-year precipitation event
~ 99.97th percentile
–
1-in-100-year precipitation event ~ 99.997th percentile
Error relative
to observation
~Clausius-Clapeyron
Fractional
changes of P
Ratio of percentage
increases to the mean
precipitation changes
Precipitation extremes increase
more than mean rainfall
Why do some models exhibit super-ClausiusClapeyron behavior?
—
Does the distribution of precipitable water change?
—
Dynamical feedback?
–
Gross moist stability changes if precipitable water
increases
MIROC(hires)
21C(A1B)
20C
The reason for super-ClausiusClapeyron relation is not simple
redistribution of precipitable water
Aggregate analysis is fine, but what about local
changes?
—
Precipitation is notorious for model discrepancies
—
Let’s check on that with actual data…
Issues and future research
—
Why do models deviate from thermodynamic constraints?
—
Disaggregate seasons, climate regimes
—
What does constant percentage increase mean for the
distribution? How does it relate with a Gamma distribution,
for instance?
Summary and implications
—
—
—
—
Precipitation extremes remain a difficult challenge for climate
models, although there are a handful of good models
Multi-model ensemble analysis shows that precipitation extremes
increase more than the corresponding mean, but not
necessarily at the rate expected from the Clausius-Clapeyron
relation; the reason for the divergence from the thermodynamic
constraint is not clear yet
Damage could increase even faster
At the local level, model discrepancies dominate and any
conclusive result is hard to draw; although it is desirable to utilize
climate extreme information for impact assessment, such analysis
is still full of uncertainty