Abstract:
Rice (Oryza sativa L.) is the second main food crop after wheat in Pakistan and a key source of export earnings. To ensure food security, timely and accurate estimation of micro/macro level estimation of crop yield is important. A study was conducted in district Sheikhupura, a major rice growing area of Pakistan, with the objective to use remote sensing and crop model for rice crop yield estimation. The results of these two techniques were validated with field survey data. Landsat 8 satellite imageries were downloaded covering different stages of rice crop growth (July to November 2015). Satellite data was pre-processed in ENVI software. Normalized Difference Vegetation Index (NDVI) of rice crop was calculated for all the six imageries (n=33) of the growing season and correlated with observed yield. The highest relationship (R2=0.788; p≤0.05) between NDVI and observed yield was recorded on September 29, 2015. The NDVI map of September 29, 2015 was used to generate rice yield map using a linear regression model. A best fit was found between observed and NDVI calculated yield (R2= 0.738; RMSE=178 kgha-1). The raster map of the crop yield was used for computing the total crop production of the entire district. A comparison of the calculated production and area under rice cultivation with those of the Crop Reporting Survey of Punjab found to be 14.6 % and 9% higher, respectively. The Decision Support System for Agrotechnology Transfer (DSSAT)-Crop Simulation Model (CSM) was also used for prediction of the final rice yield based on farmers’ field inputs and other data i.e. climate, soils and genetic coefficients of the selected crop etc. A best fit was found between observed and DSSAT predicted yield (R2= 0.928; RMSE=111 kgha-1). The results of the study showed that the DSSAT predicted the rice grain yield more accurately than the RS technique. However, DSSAT-CSM needs extensive data input and has constraints of spatial extrapolation.
