Crop growth modelling and crop yield forecasting using satellite derived meteorological inputs

One of the key challenges for operational crop monitoring and yield forecasting using crop models is to find spatially representative meteorological input data. Currently, weather inputs are often interpolated from low density networks of weather stations or derived from output from coarse (0.5 degree) numerical weather forecasting systems. The current study investigated the possibilities of deriving basic meteorological inputs (temperature, radiation, evapotranspiration) from observations of MeteoSat. A time-series of 10 years of decadal satellite products was used to run theWOFOST crop growth model and to simulate crop yield indicators for Spain, Belgium and Poland. Results with regard to the performance of the system using satellite derived meteorological inputs were compared with more traditional methods (interpolation from weather stations). Our results demonstrate that the MeteoSat derived temperature and radiation products can used be as input in a mechanistic crop model. Although the year to year variability in the potential crop simulations is not reproduced, it should be noted that the CGMS database itself contains large uncertainty in radiation input and therefore is not an absolute reference for comparison. For water-limited production levels we concluded that the MeteoSat based simulation are unable to reproduce the drought stress which usually occurs under Mediteranean conditions. This is a result of the fact that the MeteoSat based reference evapotranspiration is on average 30 percent smaller compared to the standard Penman reference evapotranspiration. The absence of drought stress in MeteoSat based simulations lead us to conclude that the EARS potential evapotranspiration product is not directly suitable for use in the WOFOST model. A considerable recalibration of the evapotranspiration related components of the WOFOST model will be necessary before we can use the MeteoSat based reference evapotranspiration in the model