An efficient posterior probability-based image fusion change detection model for the estimation of seasonal agricultural changes using microwave and optical datasets

Abstract

Detecting seasonal variations is an important application of remote sensing to understand temporal change patterns of agricultural land, soil productivity, and crop yield predictions. It also provides valuable insights for farmers and policymakers to make informed decisions. Remote sensing is one of the most effective and cost-efficient methods for monitoring agricultural land on a global scale. While optical sensors are commonly used to observe seasonal vegetation trends, their effectiveness is significantly limited by cloud cover and atmospheric disturbances. Whereas microwave sensors can penetrate the clouds and provide structural and moisture-related information, they lack spectral sensitivity to key vegetation indices derived from optical data. This article develops a novel posterior probability-based fusion change detection (PFCD) model by integrating the posterior probability space into image fusion and change detection, enabling the accurate estimation of seasonal agricultural changes. To validate the proposed model, a case study was conducted in a part of Punjab, India, for seasonal agricultural changes during the 2023–24 period, utilizing optical-based multispectral imager (MSI) from Sentinel-2 and microwave-based synthetic aperture radar (SAR) from Sentinel-1. The experiments confirmed that PFCD had achieved an accuracy of 92.73–96.41 %, with a kappa value of 0.89–0.95 for thematic maps and an accuracy of 90.21–93.05 %, with a kappa value of 0.89–0.93 for change maps. A cloud cover analysis further validated the model’s robustness, demonstrating its effectiveness in accurately estimating land surface changes during cloudy periods without compromising spectral or spatial detail.