Enhancing performance of multi-temporal tropical river landform classification through downscaling approaches

Multi-temporal remote sensing imagery has the potential to classify river landforms reconstruct evolutionary trajectory of morphologies. Whilst open-access archives high spatial resolution are increasingly available from satellite sensors, such as Sentinel-2, there remains a fundamental challenge maximising utility information in each band whilst maintaining sufficiently fine identify landforms. Although image fusion and downscaling methods on Sentinel-2 have been investigated for many years, is need assess their performance multi-temporal object-based landform classification. This investigation first compared three methods: area point regression kriging (ATPRK), super-resolution based Sen2Res, nearest neighbour resampling. We assessed by accuracy, precision, recall F1-score. ATPRK was optimal approach, achieving an overall accuracy 0.861. successively engaged set experiments determine training model, exploring single multi-date scenarios. find that not only does with better quality improve classification performance, but datasets establishing machine learning models should be considered contributing higher accuracy. paper presents workflow automated recognition could applied other tropical rivers similar hydro-geomorphological characteristics.