Integrating remote sensing and machine learning into environmental monitoring and assessment of land use change

Abstract

Addressing the increasing burden on land use requires effective policy for sustainable land use along with economic development. Analysis of local and global indicators based on land use maps could reveal information on the progress of sustainable development. This study proposes a method that reduces the time and cost of creating land use maps applicable for many purposes of environmental protection. Freely accessible existing data, Sentinel-2 satellite images, together with a machine learning algorithm, Random Forest, are integrated to generate an annual map, sufficient to meet the intended needs. The method is illustrated by a case study of Phuket in Thailand. An annual map for Phuket created using the proposed method was compared to the official map released by the Thai government for the year 2018. The two maps did not differ significantly, validating the efficacy of the proposed method. Annual maps were then produced for several years to assess the effect of land use change in the past 19 years on the environmental and sustainable management in Phuket. Although there was evidence of the efforts to develop Phuket island as a sustainable province such as the government policy to conserve green areas, land use change based analytical results indicated Phuket's urban development was not going in an environmentally sustainable direction.