Model for Estimating Rapeseed Production Based on Remote Sensing

Abstract

The study analyzed in dynamics a rapeseed culture based on remote sensing. Satellite images from the Landsat 8 system were used. The images were taken at 8 different time points, t1 (March), t2, t3 (April), t4, t5 (May), t6, t7 (June), t7 (July). Based on spectral information, NDVI and MSAVI indices were calculated. The variation of the values of NDVI and MSAVI indices in relation to time was described by polynomial models of 2nd degree, in conditions of statistical accuracy (R2=0.798, p=0.0182, F=9.9134 for NDVI, respectively R2=0.851, p=0.00856, F=14.286 for MSAVI). In the case of both indices, a deviation of the values of the moment t4 (flowering period) was found in relation to the theoretical model obtained. The regression analysis facilitated the obtaining of a production prediction equation (YP) based on the values of the NDVI and MSAVI indices, in statistical accuracy conditions, R2=0.994, p=0.0105, F=94.1279. The prediction error (PE) was calculated compared to the actual production obtained, and the moments t1, t2 and t3 facilitated the estimation of the production under the most accurate conditions (PE values were minimal). 3D and isoquants models were obtained and they have described the variation in estimated YP output in relation to NDVI and MSAVI. A high level of similarity regarding PE was recorded in the case of t1, t2 (SDI=17.098), respectively t2 and t4 (SDI=18.098). According to the PCA, PC1 explained 67.745% of the variance, and PC2 explained 32.037% of the variance.