Comparison of Geostatistical Method and Remote Sensing in Estimating Rice (Oryza sativa L.) Grain Yield in Guilan Province

Introduction Food security has been the most important human concern on the planet. Food security and future changes in food prices and crop development undeniably depend on the average yield of crops. Rice is the second largest crop in terms of area under cultivation and provides food for more than half of the world's population. Due to the importance of rice production, it is important to monitor its production on a large scale. Remote sensing and geostatistical methods play an important role in spatial and temporal evaluation of climatic, soil, living factors and management methods. The study of the amount of yield obtained in province of Guilan can provide appropriate information in advancing/pursuing goals such as the study of the amount of yield gap in this region; therefore, this study was conducted to evaluate the yield of rice using two methods of geostatistics and remote sensing approach in 2016 and 2017 in Guilan province.Materials and Methods In this study, different methods of preparing vegetation map to determine rice yield map were evaluated. To compare methods based on satellite-imagery and geostatistical procedure-models to estimate the rice yield of rice cultivated lands in Guilan province, a study was conducted in the 2016 and 2017. For field operations, 320 fields were surveyed to record grain yield (of total 238,000 hectares of rice-grown fields) during the physiological maturity stage. In this study, 33 statistical procedure-models were used to interpolate the amount of grain yield and then the accuracy of interpolation methods were evaluated using various statistical criteria. Satellite-imagery based methods using Landsat 8 satellite Operational Land Imager (OLI) sensor images related to the dates of June 18, August 9, August 21 in 2016, and July 23, August 8 in 2017 and the images of the Sentinel-2 satellite on June 30 and September 13, 2017, were used. Eight satellite-derived vegetation indices were calculated and the relationship between them and the yield variable was extracted using the regression relationship, and the yield map was prepared and evaluated. By fitting the peak logistics model between yield values and vegetation indices and selecting the superior index, the yield map was prepared with the help of remote sensing and the obtained yield maps were compared with different statistical criteria.Results and Discussion The results of the evaluation of geostatistical interpolation methods showed the superiority of the ordinary stable kriging procedure-model over other models. In this study, the RVI vegetation index was selected as the superior index to predict actual yield throughout the Guilan Province. Comparison of geostatistical procedure-models and satellite-imagery oriented models based on the determination coefficient of regressed equations and root mean square error (RMSE) to estimate grain yield in Guilan province showed that both procedures had acceptable accuracy, however, due to the ability of remote sensing to distinguish the pointwise optical reflection of phenomena and to predict yield with high spatial resolution, this method achieved higher accuracy in yield estimating.Conclusion Comparison of geostatistical and remote sensing methods in predicting farm yield indicated that the remote sensing method was more accurate. Early harvest forecasting based on the information extracted from this study showed that the use of image / images obtained in June and August and OLI sensor of Landsat-8 satellite in Guilan province can be used as a basis for forecasting the yield of this plant in the coming years. Such studies in Guilan province, by considering the share of Guilan province in the country's rice supply, will play an effective role in managerial decisions regarding rice supply and demand at the macro level. The results also showed that sampling in the study area based on a regular (systematic) spatial pattern can increase the accuracy of geostatistical methods in estimation of regional yield. The results of this project can provide suitable basic information for other studies such as yield gap, the reasons behind it, the relationship between land suitability and the obtained yield and forecasting and estimating yield in different time periods.