Abstract

Critical watersheds that exceed their carrying capacity occur in many regions of the world; their formation is facilitated by a significant driving factor known as land use/land cover (LULC) changes. This study aims to identify the LULC changes in Cisadane Watershed, Indonesia, in 2010, 2015, 2021, and simulate future LULC for 2030 and 2050. Landsat 2010 and 2015 and Sentinel 2A images from 2020 were employed for deriving LULC maps using Random Forest. This study applied a Land change modeler (LCM) under the multi-layer perception Markov chain (MLP-MC) to predict the future LULC in three scenarios. The scenarios are business as usual (BAU), protecting paddy fields (PPF), and protecting forest areas (PFA). The results showed that all the LULC maps demonstrated excellent accuracy, indicated by >83% overall accuracy. Furthermore, BAU produces the worst effect of decreasing forest and paddy field areas. PPF tends to cause forest loss, while PFA is predicted to reduce the paddy fields. There is a trade-off between maintaining food security and conserving natural resources. The study reveals the importance of efficient land use planning in the future amidst increasing resource demand due to population growth while existing land resources are limited.

Keyword

• Cisadane watershed
• Land change modeler
• Random forest
• Remote sensing