Evaluation of Land Use Land Cover Changes in Nan Province, Thailand, Using Multi-Sensor Satellite Data and Google Earth Engine
Jiratiwan Kruasilp,Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170, Thailand,Sura Pattanakiat,Thamarat Phutthai,Poonperm Vardhanabindu,Pisut Nakmuenwai,Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170, Thailand,Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170, Thailand,Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170, Thailand,Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170, Thailand
DOI: https://doi.org/10.32526/ennrj/21/202200200
2023-02-23
Environment and Natural Resources Journal
Abstract:Land use and land cover (LULC) conversion has become a chronic problem in Nan province. The primary factors of changes are lacking arable land, agricultural practices, and agriculture expansion. This study evaluated the usefulness of multi-sensor Landsat-5 (LS5), Landsat-8 (LS8), Sentinel-1 (S1), and Sentinel-2 (S2) satellite data for monitoring changes in LULC in Nan province, Thailand during a 30-year period (1990-2019), using a random forest (RF) model and the cloud-based Google Earth Engine (GEE) platform. Information of established land management policies was also used to describe the LULC changes. The median composite of the input variables selection from multi-sensor data were used to generate datasets. A total of 36 datasets showed the overall accuracy (OA) ranged from 51.70% to 96.95%. Sentinel-2 satellite images combined with the Modified Soil-Adjusted Vegetation Index (MSAVI) and topographic variables provided the highest OA (96.95%). Combination of optical (i.e., S2 and LS8) and S1 Synthetic Aperture Radar (SAR) data expressed better classification accuracy than individual S1 data. Forest cover decreased continuously during five consecutive periods. Coverage of maize and Pará rubber trees rapidly expanded in 2010-2014. These changes indicate an adverse consequence of the established economic development promoted by industrial and export agriculture. The findings strongly support the use of the RF technique, GEE platform and multi-sensor satellite data to enhance LULC classification accuracy in mountainous area. This study recommended that certain informative and science-based evidence will encourage local policymakers to identify priority areas for land management and natural resource conservation.