Mapping of the Maize Area Using Remotely Detected Multispectral and Radar Images Based on a Random Forest Machine Learning Algorithm

Food security requires accurate mapping of the maize crop to ensure long-term sustainability. However, mapping maize crops using remote sensing images poses a significant challenge due to the ecological gradient and mixing between maize crop and nonmaize crop areas. This study aims to investigate the combined use of optical Sentinel-2 (S2) and Sentinel-1 (S1) radar images to map the maize crop in the South African Free State province. The random forest (RF) algorithm was used for classification and the feature variables used in mapping were ranked. The results indicate a high overall accuracy of 95%, indicating that the RF algorithm performed well in distinguishing maize and non-maize crops. Consequently, the red edge band was the most important feature in the classification due to its ability to identify biophysical variables in maize crops. The results of this study can help agronomists, economists, farmers, policy makers and the government come up with strategies to increase maize production, which is a crucial component of sustainable development and food security.