Land Cover Change Detection around Upstream of Serayu Watershed using Machine Learning

Abstract

Water resources are important to be managed in a sustainable manner. Changes to the land cover such as from forest to agriculture will affect the water quality and quantity of the whole watershed system. The Serayu watershed in Central Java is considered one of the critical watersheds in Indonesia due to the high erosion and sedimentation rate from the conversion of forested land to horticulture land. This paper presents a study of land cover change using the Sentine1-2 satellite images from 2018- 2019 to 2020-2021 using the machine learning method. The Sentine1-2 images have a temporal resolution of 10 days which is necessary because of the high cloud cover in the study area. Image classification using the Light Gradient Boosting yields an overall accuracy from the training and testing dataset of 1.0 and 0.929 for images 2018 – 2019 and 1.0 and 0.915 for images 2020 – 2021. Field verification upstream of the Serayu watershed shows a good agreement with the classification results, where discrepancies are mainly due to land clearing of the agriculture plots.