Effects of Weather Change on Agricultural, Food Production & the - - PowerPoint PPT Presentation

Effects of Weather Change on Agricultural, Food Production & the Developing World

Wolfram Schlenker

Columbia University and NBER

New School - November 18, 2013

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 1 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 2 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 3 / 35

Statistical Approaches

Statistical studies

Estimated using real-world data

Aggregate or field-level data

Limited number of variables

Usually precipitation and temperature (degree days) Agronomic models use many more variables

Estimated impacts of climate change

Predicted changes in yields/profits from panel Estimating adaptation using cross-section Studies regress trends in yields on trends in weather

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 4 / 35

Statistical Approaches

Statistical studies

Estimated using real-world data

Aggregate or field-level data

Limited number of variables

Usually precipitation and temperature (degree days) Agronomic models use many more variables

Estimated impacts of climate change

Predicted changes in yields/profits from panel Estimating adaptation using cross-section Studies regress trends in yields on trends in weather

What key variable to use?

Statistical studies have learned from crop models

Degree days (non-linear transformation) Degrees above (below) threshold

Statistical studies have identified key parameters

E.g., negative effects of extreme heat

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 4 / 35

Construction of Degree Days

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 5 / 35

Construction of Degree Days

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 5 / 35

Construction of Degree Days

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 5 / 35

Construction of Degree Days

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 5 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 6 / 35

Link between Temperature and US Yields

Statistical Analysis

Panel of county-level yields in Eastern United States Corn and Soybeans (two biggest staple commodities in US) Fine-scale weather (daily temperature / precip on 2.5mile grid) Years: 1950-2005

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 7 / 35

Link between Temperature and US Yields

Statistical Analysis

Panel of county-level yields in Eastern United States Corn and Soybeans (two biggest staple commodities in US) Fine-scale weather (daily temperature / precip on 2.5mile grid) Years: 1950-2005

Model accounts for

Amount of time spent in each 1◦C interval Quadratic in total precipitation State-specific quadratic time trends County fixed effects

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 7 / 35

Results: Effect of Weather on Yields

Panel of Corn and Soybean Yields

Schlenker & Roberts (PNAS 2009)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 8 / 35

Results: Source of Variation

Corn and Soybean Yields - Various Source of Identification

Schlenker & Roberts (PNAS 2009)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 9 / 35

Results: Climate Impacts

Climate Impacts - Uniform Scenarios

Schlenker & Roberts (PNAS 2009)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 10 / 35

Recent Example: 2012 Heat Wave / Drought

Berry, Roberts & Schlenker (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 11 / 35

Recent Example: 2012 Heat Wave / Drought

Berry, Roberts & Schlenker (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 12 / 35

Recent Example: 2012 Heat Wave / Drought

Berry, Roberts & Schlenker (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 13 / 35

Adaptation to Trends (1980-2005)

Burke & Emerick (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 14 / 35

Adaptation to Trends (1980-2005)

Burke & Emerick (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 15 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 16 / 35

Agronomic Evidence on Mechanism

Schlenker & Roberts (PNAS 2009)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 17 / 35

Agronomic Evidence on Mechanism

Biophysical evidence

Lobell, Hammer, McLean, Messina, Roberts, Schlenker (2013)

APSIM: biophysical model of crop growth Includes water balance, etc

Mechanism behind EDD (Extreme degree days)

Impacts water stress in two ways

Reducing soil water (evaporation) Increased demand for soil water to sustain carbon uptake

Precipitation only impacts soil moisture

Drought is a relative concept

Water requirements depend on temperature

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 17 / 35

Heat versus Water

Water versus Temperature: Chicago Marathon (2007) in Hot Weather

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 18 / 35

Heat versus Water

Water versus Temperature: Chicago Marathon (2007) Ran Out of Water

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 18 / 35

Heat versus Water

Lobell, Hammer, McLean, Messina, Roberts, & Schlenker (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 18 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 19 / 35

Statistical Study in Africa

Lobell, B¨ anzinger, Magorokosho, and Vivek (2011) Unique data set of fiel trials

123 research stations

CIMMYT

Testing for drought conditions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 20 / 35

Statistical Study in Africa

Lobell, B¨ anzinger, Magorokosho, and Vivek (2011) Unique data set of fiel trials

123 research stations

CIMMYT

Testing for drought conditions

Matched with closest weather station

Better than gridded weather data Authors split season into three phases (separate coefficients)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 20 / 35

Statistical Study in Africa

Lobell, B¨ anzinger, Magorokosho, and Vivek (2011) Unique data set of fiel trials

123 research stations

CIMMYT

Testing for drought conditions

Matched with closest weather station

Better than gridded weather data Authors split season into three phases (separate coefficients)

Major results

Find nonlinearity effect of temperature on yield Stronger under drought conditions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 20 / 35

Location of Field Trials

Lobell, B¨ anzinger, Magorokosho, & Vivek (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 21 / 35

Regression Coefficients for Temperature

Lobell, B¨ anzinger, Magorokosho, & Vivek (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 22 / 35

Simulating 1◦C Warming

Lobell, B¨ anzinger, Magorokosho, & Vivek (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 23 / 35

Geographic Distribution of Impacts

Lobell, B¨ anzinger, Magorokosho, & Vivek (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 24 / 35

Separating Impact Due to Temperature and Precipitation

Schlenker & Lobell (2010): Country-level panel for Subsaharan Africa

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 25 / 35

Separating Model and Climate Uncertainty

Schlenker & Lobell (2010): Country-level panel for Subsaharan Africa

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 26 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 27 / 35

Statistical Model

Do climate trends already have an effect on food production

Statistical model linking yields to weather

Predicted production under observed trend Predicted production if trend is removed

Difference in global production

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 28 / 35

Statistical Model

Do climate trends already have an effect on food production

Statistical model linking yields to weather

Predicted production under observed trend Predicted production if trend is removed

Difference in global production

Focus on four major staples

Maize, rice, soybeans, wheat Responsible for 75% of global caloric production

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 28 / 35

Statistical Model

Do climate trends already have an effect on food production

Statistical model linking yields to weather

Predicted production under observed trend Predicted production if trend is removed

Difference in global production

Focus on four major staples

Maize, rice, soybeans, wheat Responsible for 75% of global caloric production

Panel of country-level yields (FAO data)

Matched with weather data (University of Delaware) Averaged over area where crop is grown

Monfreda, Ramankutty & Foley 2008

Averaged over crop-specific growing season

Sacks, Deryng, Foley & Ramankutty (2010)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 28 / 35

Global Production: 1961-2010

Roberts & Schlenker (2013)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 29 / 35

Temperature Trend (1980-2008) in Historic Std. Deviation

Lobell, Schlenker & Costa-Roberts (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 30 / 35

Country-Crop Specific Temperature Trends (1960-1980)

Lobell, Schlenker & Costa-Roberts (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 31 / 35

Country-Crop Specific Temperature Trends (1980-2008)

Lobell, Schlenker & Costa-Roberts (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 32 / 35

Predicted Impact of Observed Trend

Combined Price Effect: 18.9% (no CO2 fertilization), 6.4% (including CO2 fertilization)

Lobell, Schlenker & Costa-Roberts (2011)

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 33 / 35

Outline

1

Motivation for Statistical Studies

2

Modeling US Yields

3

Water versus Temperature

4

Modeling Yields in Africa

5

Climate Change and Global Production Trends

6

Conclusions

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 34 / 35

Conclusions

Statistical studies of climate change

Driving force: extreme heat Large yield decline if maximum temperature rises a lot Impact depends on baseline and predicted increase

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 35 / 35

Conclusions

Statistical studies of climate change

Driving force: extreme heat Large yield decline if maximum temperature rises a lot Impact depends on baseline and predicted increase

Agronomic evidence

APSIM model Extreme heat has larger effects on yields than precipitation

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 35 / 35

Conclusions

Statistical studies of climate change

Driving force: extreme heat Large yield decline if maximum temperature rises a lot Impact depends on baseline and predicted increase

Agronomic evidence

APSIM model Extreme heat has larger effects on yields than precipitation

African agriculture

Most places already hot Further heating harmful due to temperature nonlinearity Comparable climate and model uncertainty

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 35 / 35

Conclusions

Statistical studies of climate change

Driving force: extreme heat Large yield decline if maximum temperature rises a lot Impact depends on baseline and predicted increase

Agronomic evidence

APSIM model Extreme heat has larger effects on yields than precipitation

African agriculture

Most places already hot Further heating harmful due to temperature nonlinearity Comparable climate and model uncertainty

Observed temperature trends 1980-2008

Already have effect on global food prices

Wolfram Schlenker (Columbia and NBER) Weather Extremes and Agriculture New School - November 18, 2013 35 / 35