Prediction and Simulation of Land Use and Land Cover Changes Using Open Source QGIS. A Case Study of Purwokerto, Central Java, Indonesia

Abstract

Population size multiplies along with the increasing need for residential space. As often occurs in developing cities like Purwokerto, population growth is associated with land use/land cover (LULC) change to accommodate housing demand both in the present and future. Therefore, this study was intended to map LULC changes in three different years: 2008, 2013, and 2018, and predict the change in 2023. For LULC data extraction, a pixel-based digital classification with a maximum likelihood algorithm was applied to Landsat images. In addition, the LULC change prediction was modeled with Modules for Land Use Change Simulations (MOLUSCE) from the QGIS plugins. It used two algorithms: artificial neural network (ANN) with a multilayer perceptron (MLP) and cellular automata (CA). The LULC classifications for 2008, 2013, and 2018 were 88%, 86%, and 88% accurate, while the prediction was 75.26% accurate, with a kappa of 0.634. Predictions and simulations indicate fluctuations in LULC change in the City of Purwokerto periodically, especially for built-up land, showing growth that continues to increase significantly.