Large-scale Simulation of Rice yield and Water Productivity Using CERES-Rice Model

In order to evaluate the performance of the CERES-Rice model in rice biomass and grain yield estimation in large scale، this study was carried out in paddy fields of Guilan، Iran. First the model was calibrated and evaluated using data from a research at Rice Research Institute of Iran، then the model accuracy was assessed in 26000 ha of Some Sara region. 110 common farmers fields were chosen، the soil، water، crop samples and other needed information were taken. The model was ran for each field. The results showed that the Error of yield and biomass estimation in research station were below 10 percent. The results in large scale showed that the model estimation of yield and biomass could be categorized in three different parts. While in fields with average biomass and yield observation the setimation error was negligible. Underestimation was observed in the fields with the yield above the average of the region while in high yielding fields overestimation was observed. The Normalized Root Mean Square Error was about 22. 7 and 21% for biomass and yield prediction. Although in some cases the error was about 50%، in 75% of the region it was in range of 0-10%، and in 95 % of the region it was in range of 0-20%. Water productivity based on water used in most of the fields was in the range of 0. 5- 0. 6 kg/m3 and the water productivity based on transpiration was 0. 85-0. 9 kg/ m3. The model showed a satisfactory accuracy to estimate both yield and biomass in large scale paddy fields.