Crop Classification in Uzbekistan Using Random Forest: Integrating Sentinel-1 SAR and Sentinel-2 Optical Data with Ground-Truth Validation

This study investigates the effectiveness of Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 optical imagery for crop classification using the Random Forest (RF) algorithm. We aim to assess the potential of combining these two datasets for improved classification accuracy of crop types and land cover. Sentinel-1 SAR data, which provides valuable information on surface roughness and backscatter, was applied to classify crops in an agricultural area. However, results showed that while Sentinel-1 was effective for some crop types, it struggled to distinguish rice and maize from cotton, which exhibited similar backscatter characteristics. In contrast, Sentinel-2 optical data, leveraging its rich spectral bands, showed a significant improvement in class separability, particularly for crops like cotton, fallow, and other. Combining both Sentinel-1 and Sentinel-2 data resulted in a notable enhancement in classification performance, with higher overall accuracy compared to the use of each sensor individually. The RF classifier, applied to the multi-sensor data, demonstrated robust performance with an overall accuracy of 0.98 and a Kappa coefficient of 0.96. This study highlights the complementary nature of SAR and optical data and their potential for enhancing crop classification accuracy. The findings underscore the importance of using multi-sensor datasets for accurate agricultural monitoring, offering valuable insights for land management, crop monitoring, and decision-making in precision agriculture.