IMPROVING AGRICULTURAL MANAGEMENT IN A LARGE SCALE PADDY FIELD BY USING REMOTELY SENSED DATA IN THE CERES-RICE MODEL

摘要

Compared to small-scale, some constraints in large-scale rice fi eld caused crop growth modelsfail to reach an acceptable estimate of the yield. This study was conducted to investigatethe possibility of enhancing the accuracy of CERES-Rice model prediction in a large-scalethrough the use of Landsat 5 satellite imagery (termed as assimilation). Firstly, the modelwas calibrated by data taken from a local research. The model accuracy was then evaluatedin 110 paddy fi elds over 26000 ha (method A). Then, the model was recalibrated by paddyyield estimated from Landsat 5 image (method B). The two methods were compared basedon their results. The results revealed that RMSEn in simulating grain yield in small-scalefi eld experiment on Hashemi cultivar for calibration and validation of the model were 9 and8%, respectively (R2=0.7), which indicated model’s high accuracy in yield prediction. WhileRMSEn in simulating grain yield in large-scale (methods A) was 22% (R2=0.54), the use ofLandsat image in the assimilation method (method B) increased its accuracy dramatically toRMSEn of 12.7% (R2=0.72).