Sensitivity of crop models to the inaccuracy of meteorological observations

The accuracy of a crop model is judged mostly by how precise it is in estimating the production. It is especially significant in cases when crop models are used in decision support. The preciseness of a crop model is determined on one hand by the authenticity of the algorithms describing the processes of the real world, while on the other hand by the quality of its input data. The goal of this study was to test how sensitive crop models are to errors occurring in the observation of key weather elements. Air temperature, solar radiation, and precipitation are the key input data for most crop models. The level of error in air temperature, radiation, and precipitation measurements were reported to be ±0.2 °C, ±2% and ±3% respectively by the Hungarian Meteorological Service. The question we asked was: 'To what extent do the yield and biomass change due to this level of inaccuracy in weather input?' First the model was run with the original 20 year-long observed weather inputs. Then all the error combinations of the three elements were applied as inputs for each year. On an average we got 3.2% uncertainty for biomass and 6% uncertainty for yield. Certain combinations of errors resulted in more then 20% deviation compared to the yields gained using the base input data set. There was a combination of errors that caused 10% uncertainty even in the average of 20 years. The uncertainty caused by the incorrect weather data set was found to be greater in low yielding years than in high yielding years. The crop model proved to be most sensitive to errors of the measured temperature, and least sensitive to errors of the measured radiation.