Multi-Season Land Cover Change Modeling of Pantabangan-Carranglan Watershed Using Sentinel-1 and Sentinel-2 Imagery

Abstract. Understanding land cover change is vital for sustainable management, particularly in diverse and ecologically significant landscapes like the Pantabangan-Carranglan Watershed (PCW) in the Philippines. This study employed multi-seasonal Synthetic Aperture Radar (SAR) and optical imagery from Sentinel satellites to enhance land cover classification and predict future changes in PCW. Data preprocessing and combination were performed using the Sentinel Application Platform (SNAP) software, resulting in multi-season datasets that accounted for the area’s distinct climatic patterns. Classification was conducted using the Random Forest algorithm, generating land cover maps for 2017, 2020, and 2023, followed by change detection and prediction using Artificial Neural Network (ANN) for years 2023 and 2026. Results indicated a general increase in forest cover, with notable gains observed from non-forest vegetation and bare soil classes, suggesting ecological succession. Increases in forest cover of about 89 km2, 34 km2, and 28 km2 were observed for 2017-2020, 2020-2023, and 2023-2026 analysis, respectively. Generally, classification accuracy (total accuracy) remained acceptable (77%-85%), and ANN-based predictions showed limitations which were affected by input data misclassifications. In general, the multi-season/combined season model emerged as the most effective for change detection, outperforming the mono-season approach with an average total accuracy of 83.73%. For the predicted future land cover based on the best performing model, the total accuracy for 2023 was at 79%. Despite the challenges, the study underscores the potential of integrating the Sentinel-1 and Sentinel-2 data for land cover monitoring, offering insights into landscape dynamics and conservation strategies. Future work should focus on refining methodologies to improve differentiation between classes particularly bare soil and built-up.