d. w. shin, s. cocke, y.-k. lim, t. e. larow, g. a. baigorria, and j. j. obrien center for...

D. W. Shin, S. Cocke, Y.-K. Lim, T. E. LaRow, G. A. Baigorria, and J. J. O’Brien

Center for Ocean-Atmospheric Prediction StudiesFlorida State University, Tallahassee, FL, USA

Agricultural&Biological Engineering Department, Univ. of FloridaMarch 6, 2008 at CPASW

Interannual Crop Yield Simulations over the Southeast US using Global and Regional

Climate Model Products

Outline

1. Background

2. The FSU/COAPS Climate Modeling System and The DSSAT Crop Model

3. Ensemble Runs

4. The FSU/COAPS GCM results

5. The FSU/COAPS RCM results

6. Station Level results

7. Crop Model results

8. Future Directions

Background

RISA http://www.climate.noaa.gov/cpo_pa/risa/Regional Integrated Sciences and Assessments

http://secc.coaps.fsu.edu

http://AgClimate.org

FSUNRSM(20km)

OASIS Coupler

FSU/COAPS Climate Modeling System

Regional Biosphere

FSUGSMT63 (200km)

Global Biosphere

OCEANHOPE-OM1, HOPE-G,

HYCOM, MICOM

Crop Model

Crop Model

DSSAT (Crop Model)

• DSSAT: Decision Support System for Agrotechnology Transfer

• DSSAT: a microcomputer software program combining crop soil and weather data bases and programs to manage them, with crop models and application programs, to simulate multi-year outcomes of crop management strategies.

• DSSAT allows users to ask "what if" questions and simulate results by conducting, in minutes on a desktop computer, experiments which would consume a significant part of an agronomist's career.

Linking Climate Models to Crop Models

• Grand idea is to be able to make forecast before season regarding crop situations and perhaps suggest “best management” practices for that year

• At present, we are looking into peanut or corn yields in some selected stations in southeast USA

The regional model was centered over the southeast U.S. and run at 20 km resolution, roughly resolving the county scale. Outputs from the model such as max/min surface temperature, precipitation and shortwave radiation at the surface is used as inputs into the crop model to determine crop yields.

Using the FSU/COAPS GSM & RSM system, warm season (March-September, 7 month simulation) and cold season (October-march, 6 month simulation) ensemble simulations are performed for the period of 19 yrs (1987-2005) to characterized uncertainty in the forecast. Twenty member ensembles of the regional model are generated using different initial conditions and model configurations (i.e., the ensemble methods based on different convective schemes).

Ensemble runs

GSM Results

PRECIPITATION: Temporal correlation (1987-2005)

PrecipitationPrecipitation

DEMETER MMEPAPCC MMEP

JJA

DJF

MME Hindcast Skill: Temporal Correlation/ 1981-2001MME Hindcast Skill: Temporal Correlation/ 1981-2001(Lee et al. 2007)(Lee et al. 2007)

2m Temperature: Temporal correlation (1987-2005)

2m Air Temperature2m Air TemperatureDEMETER MMEPAPCC MMEP

JJA

DJF

MME Hindcast Skill: Temporal Correlation/ 1981-2001MME Hindcast Skill: Temporal Correlation/ 1981-2001(Lee et al. 2007)(Lee et al. 2007)

Saha et al (2006)

PRECIPITATION: Temporal correlation (1987-2005)FSU/COAPS (1987-2005) CFS (1981-2003)

FSU/COAPS (1987-2005) CFS (1981-2003)Saha et al (2006)

2m Temperature: Temporal correlation (1987-2005)

RSM Results

Downscaling (Regional model)

FSU Regional Model

20 km

MAXIMUM TEMPERATURE19 year (1987-2005) ave (oC) (model – obs)

Ensemble Mean

MINIMUM TEMPERATURE19 year (1987-2005) ave (oC) (model – obs)

Ensemble Mean

PRECIPITATION19 year (1987-2005) ave (mm/day) (model – obs)

Ensemble Mean

Tmax: Temporal correlation (1987-2005)

Tmin: Temporal correlation (1987-2005)

PRECIPITATION: Temporal correlation (1987-2005)

Station Level Results

Station Level (Tifton, GA, 1987)

Tmax

Station Level (Tifton, GA, 1987)

Tmin

Station Level (Tifton, GA, 1987)

Prcp

Crop Model Results

Observed weather

Raw ensemble member 1….

Raw ensemble member 20

Raw daily seasonal-climateHindcast

Bias-corrected ensemble member 1….

Bias-corrected ensemble member 20

Bias-corrected daily seasonal-climate Hindcast

Bias-correction

Raw crop-yield ensem. member 1….

Raw crop-yield ensem. member 20

Crop yield Hindcast

CERES-Maize

Bias-corrected crop-yield ens. member 1….

Bias-corrected crop-yield ens. member 20

Crop yield Hindcast

CERES-Maize

CERES-Maize

Crop yield using observed weather

Experimental Design

PEANUT YIELDS

(1994-2003)

Site specific soil profiles (U.S. Soil Conservation Service data)

Rainfed conditions

Identical planting date for each year: April 25

Maize Yield

No bias-correction!

Peanut Yield

No bias-correction!

Baigorria et al. (2007) see member 2&6

Tifton, GA

Maize Yield global (green) vs. regional (red) model

Future Directions

1. More sites and other crops

2. A posteriori bias correction: precipitation

3. How can we use a climate ensemble forecast to issue an ACCEPTABLE probabilistic crop yield forecasts?

4. Dynamical vs. Statistical approaches

• Schoof et al. (2007); Lim et al. (2007)

5. CFS Statistical downscaling

6. A coupled version of atmospheric and crop models

- nonlinear seasonal weather-yield interactions