Climate Analysis for Enhanced Resilience
Ben ZaitchikJohns Hopkins University
Projected Precipitation Changes
IPCC AR4
-15
-10
-5
0
5
10
15
B1A1BA2
16 Different Climate Models
% P
reci
pita
tion
Chan
ge
Change in Annual Precipitation:2080s – present, Blue Nile Headwaters
NASA’s Project Nile
GCM-based Precipitation Maps
• Spatially coarse• Highly uncertain• Underestimate extremes• Systematically wrong in some
regions
So: Are the GCMs Useless?
• No: there is useful information in the projections
• But: precipitation projections applied directly to food/water security planning can be worse than useless
1. Dynamical Downscaling
• Physically-based predictions
• Can handle non-stationarity
• Requires extensive evaluation of RCM and GCM
• Computer and time intensive
2. Statistical Downscaling
• Data-based and not computer intensive
• Does not rely directly on GCM atmospheric dynamics
• Requires 30+ year meteorological station records
• Assumes stationarity
1950 1960 1970 1980 1990 2000 2010
01
00
20
03
00
40
05
00
60
07
00
interpolated prate
Jul - NA m a.sl. 37.42 degE 11.6 degN
Time
Kg
/m^2
/s
3. Physiographic Interpolation
• Physically based
• Highly localized at low computational expense
• Can utilize satellite data
• Derived from a larger scale projection
• Requires calibration and evaluation
Summary
• GCM projections require interpretation and downscaling
• Dynamical downscaling is valuable but resource intensive, and it relies on GCM atmospheric boundary conditions
• Much can be achieved with statistical methods, but data and understanding are required
• Coordinated dynamical-statistical approaches are optimal