FORECASTING MAIZE YIELD USING REMOTE SENSING
Abstract
The study aimed at evaluating the relationship between NDVI and maize yield at Makande prison farm in Thyolo district. The data used in the study were remotely sensed images, and crop yield statistics of years 2019 to 2021. Images used were obtained using the Sentinel-2 satellite constellation, which is part of the Copernicus program, and were acquired using QGIS and were processed. NDVI was used with an empirical regression model to estimate maize yield. The predicted yield was correlated with the actual yield using linear regression. A correlation coefficient of 0.99, R2= 98.8%, and root mean error square and mean absolute error of 0.1779 and 0.119 in that order were established. The findings illustrated a very strong correlation between NDVI and maize yield. The study recommends that the use of more satellite images and yield data for not less than 6 years will help to increase the prediction accuracy.
Research summary
The study aimed at evaluating the relationship between NDVI and maize yield at Makande prison farm in Thyolo district. Images used were obtained using the Sentinel-2 satellite constellation, which is part of the Copernicus program, and were acquired using QGIS and were processed. A correlation coefficient of 0.99, R2= 98.8%, and root mean error square and mean absolute error of 0.1779 and 0.119 in that order were established. The study recommends that the use of more satellite images and yield data for not less than 6 years will help to increase the prediction accuracy.