Assessing factors driving land use/land cover classes using machine learning, case study of Lombok Island-Indonesia

Land use/land cover (LULC) mapping is essential for identifying and monitoring changes in the natural environment and forest ecosystems. The use of machine learning algorithms for modeling and mapping spatial data is becoming widespread. The aim of this study is to evaluate the factors affecting LULC classes with machine learning algorithms, to compare the accuracy of the algorithms used and to determine the machine learning model that can best classify LULC. It is also aimed to determine the land use/land cover changes in the study area. The study area is Lombok Island located in West Nusa Tenggara, Indonesia. In this study, Landsat satellite imagery bands, spectral indices, topographical, biophysical, socio-economic and climatic variables were analyzed with the help of Random Forest, Gradient Boosting, Decision Tree and Naïve Bayes machine learning algorithms. When the accuracy of the models produced by machine learning was evaluated, the Random Forest algorithm achieved the highest accuracy (82%), followed by Gradient Boosting (80%), Decision Tree (73%) and Naïve Bayes (61%). In the accuracy assessments of the LULC classes produced using machine learning algorithms with the ground truthing points, the model produced by the Random Forest algorithm was the most successful machine learning model with an overall accuracy of 88%. The most effective variables for land use/land cover prediction in Lombok Island are distance to settlements, temperature and distance to roads. Significant forest loss (28 095 ha, 24% of the total forest area) from 2013 to 2022 due to urbanization, population growth and infrastructure development. The distribution of LULC classes in Lombok showed that dry land agriculture (27.56%) and paddy fields (24.22%) cover the largest areas, followed by primary forest (13%) and secondary forest (7.68%). The findings emphasize the importance of socio-economic factors in land use planning and the need for sustainable conservation policies to protect ecosystems and biodiversity in Lombok.