Predicting national wheat yields using a crop simulation and trend models
The European Union's Directorate General for Agriculture requires timely and accurate estimates of the total wheat yield at national level. In this study, four dimple prediction models were tested to assess their operational usefulness. These models consist of a trend function and a function which accounts for weather influences.Two of them first predict the national yield per hectare which is then multiplied by an area estimate, the other two predict the national yield directly. Input variables are crop growth simulation results, planted area and a trend function. As trend functions, a linear time trend and a trend based on the nitrogen fertilizer application per hectare are tested. Total national wheat yield for twelve European countries for 10 yrs were predicted. The results were evaluated against observed official national yield and yield statistics, using as criterion the relative root mean square error (RRMSE) and the root mean square error (RMSE). Area and nitrogen application estimates come available a few months after the end of growing season. Therefore the models were also tested using estimated area and nitrogen application value available as input. Crop growth simulation results proved to be more useful for predicting national yield volumes than national yield, suggesting that both area and crop area and crop growth simulation results account for the annual variation of the yield volume.
Bibliographic Reference: Article: Agricultural and Forest Meteorology - An International Journal, 88 (1997), 199-214
Record Number: 199910224 / Last updated on: 1999-03-12
Original language: en
Available languages: en