Land use change prediction with deep learning approach using artificial neural network-cellular automata (ANN-CA)

Abstract Land use prediction in future with spatiotemporal modelling are essential for monitoring and improving land use management. The aim of this study was to identify land uses in Bangkalan Regency, Indonesia, from 2009 to 2023 and predict their future changes in 2030 by ANN-CA. Data from Landsat level 2 was analyzed for land use, road and river network as well as road and river distance. In addition, NASA DEM was analyzed for land slope. Distance and slope were recognized as driving factors. Land use from 2009 to 2023 was processed and analyzed by employing GEE platform. While the land use prediction in 2030 used MOLUSCE plugin and ANN-CA model in QGIS that considered driving factors of land use changes. The classified images were evaluated with an Overall and Kappa accuracy. The result of land use identification of Bangkalan Regency from 2009 to 2023, showed that agriculture land and water body area decreased while forest, built-up area and bare land increased with the Kappa accuracy was moderate (0.59) for 2009 and strong (0.83) for 2016 and 2023. The prediction for 2030 indicated an increase in agricultural and built-up areas by 2.96% and 3.11%, respectively, while forest, water bodies and bare land are expected to decrease by -3.89%, 1.44% and 0.73%. This study may provide valuable insights for future land use management, planning strategies and policymakers to mitigate urban sprawl and promote sustainable land use practices in Bangkalan.