Yield & Climate Variability: Learning from Time Series & GCM
Presented by Amer Ahmed at the AGRODEP Workshop on Analytical Tools for Climate Change Analysis
June 6-7, 2011 • Dakar, Senegal
For more information on the workshop or to see the latest version of this presentation visit: http://www.agrodep.org/first-annual-workshop
Yield & Climate Variability: Learning from Time Series & GCM
1. Yield & Climate
Variability:
Learning from
Time Series & GCM
Presented by:
Amer Ahmed, World Bank
www.agrodep.org
AGRODEP Workshop on Analytical Tools for Climate
Change Analysis
June 6-7, 2011 • Dakar, Senegal
Please check the latest version of this presentation on:
http://agrodep.cgxchange.org/first-annual-workshop
2. Yield & Climate Variability:
Learning from Time Series & GCM
Amer Ahmed
World Bank
June 7, 2011
AGRODEP Members’ Meeting and Workshop
Dakar, Senegal
3. Outline
• Statistical models of crop response & climate
– Focus on time series and panel data approaches
– Ricardian approach of Mendelsohn et al. not
covered
• Illustrations & insights from recent research
– Long run forecasts
– Volatilities and extremes
• Implications of methodology
2
4. General:
Statistical Modeling Literature
• Strengths:
– Abstracts away from biophysical processes
– Statistical measures of accuracy (e.g. model fit)
– Does not need calibration
• International time series production/yield data
• Several publications in recent years (e.g. Lobell et al.,
2008, 2011 in Science)
– Generally negative impacts in LDCs, and tropics
– Temperature often more important than precipitation
3
5. Climate and CO2 Changes
Agricultural productivity
Average Global Yields vs. Temperatures, 1961-2002
Yield Change (%)
Temperature Change (ºC)
4
Lobell and Field (2007)
6. Africa : Schlenker-Lobell (2010, ERL)
• Data:
– for estimation: 1961-2002 data from FAO, National
Centers of Environmental Prediction, CRU;
– for prediction: climate data from 16 GCMs (2046–2065)
• Model
– Four specifications: Average weather; Quadratic in
average weather; Degree days; Degree day categories
• Predicted impacts distribution
5
7. In all cases,
except cassava
– 95%
probability
that damages
> 7%
– 5% probability
that damages
> 27%
6
8. Changing Climate Volatility
• Extreme outcomes may be particularly
important for agriculture (White et al, 2006;
Mendelsohn et al, 2007)
• Climate volatility is already changing
(Easterling et al, 2000)
– Higher temperature and precipitation extremes in
the future (IPCC, 2007)
7
9. US Maize Yield Response to Temperature
8
Schlenker and Roberts (2008)
11. Illustration: Sensitivity of Tanzanian Grain
to Climate Volatility
• Ahmed, Diffenbaugh, Hertel, Lobell, Ramankutty,
Rios, & Rowhani (GEC, 2011)
• Econometric estimation to explain interannual
change in yields using panel data from 17
administrative regions: 1992-2005
– Maize, rice, and sorghum yields (tonnes/ha)
– Temperature (growing season mean in degrees C)
– Precipitation (growing season mean in mm/month)
• Use yield equation to translate historical and future
climate into output changes
10
13. Distribution of Interannual % Changes in
Tanzanian Grains Yield due to Climate
Mass of distribution
shifting left median yield change
(more/larger outcomes
with % decline in yield
from previous year)
Outcomes within box represent 75% of all predicted outcomes
12
Ahmed et al. (2011)
14. Take Homes
• Panel data approach can be powerful tool
– Temperature more important than precipitation
• African staples yields likely to decline
• Climate change includes change in volatility
– Extreme outcomes can be more important than
mean changes
• Need to account for heterogeneity &
uncertainty due to GCMs
– Bounded envelopes
– Bias-correction
13