NWC REU 1999
May 24 - July 30



Using Long-range Seasonal Forecasts to Improve Wheat Prediction of Oklahoma Wheat Yield

Rebecca House and Scott Greene



Accurately predicted crop yield is economically valuable for farmers and for the government, who can better prepare for high or low yields given a crop forecast. The crop-environment resource synthesis (CERES) wheat model provides yield predictions given environmental variables, typically using climatological means of temperature and precipitation data. This model excludes anomalous weather regimes, leading to possible error in the forecasted yield. In Oklahoma, long-term records of daily weather are available for each county in the state. For this study, three-month seasonal forecasts were obtained from the Climate Prediction Center. To match the forecast information and to better model weather for an upcoming growing season, the climate history was divided into above normal, normal, and below normal temperature and precipitation regimes. The weights assigned to each of these categories were adjusted using probabilities from the long-range forecasts to generate a weighted climate history. Coupling this enhanced forecast with observed weather data provided a more accurate model of potential weather in years that depart from the normal. This model of weather was used in conjunction with the CERES model to predict wheat yield for five locations in Oklahoma. Incorporation of the long-range forecast showed little difference in the mean predicted yield. However, results indicated that enhanced climate forecasts can improve the prediction of wheat yield by decreasing random error in the predictions.

Paper available upon request.