Spatiotemporal Prediction of Urban Expansion in Nusantara Capital Using GeoAI: Integration of Convolutional Neural Networks and Google Earth Engine for Smart and Sustainable Urban Planning

Nusantara Capital (IKN), the new administrative center of Indonesia, is envisioned as a smart and sustainable city that demands an adaptive, evidence-based spatial planning system. Rapid infrastructure development and environmental sensitivities in the region necessitate advanced monitoring and predictive modeling tools. This study presents a GeoAI (Geospatial Artificial Intelligence) framework for spatiotemporal mapping and prediction of land cover changes in the IKN region. We integrate Convolutional Neural Networks (CNN) and Google Earth Engine (GEE) to process and classify multitemporal Sentinel-1 SAR and Sentinel-2 optical imagery from 2019 to 2024. The proposed CNN model identifies six major land cover classes including built-up areas, primary forest, secondary forest, shrubland, agriculture, and water bodies with an overall accuracy of 87.9% and an average F1-score of 0.84, outperforming traditional machine learning classifiers. Spatiotemporal prediction indicates a 35.6% expansion of built-up land by 2030, primarily at the expense of secondary forests and shrublands along infrastructure corridors such as highways and administrative zones. Using integrated heatmap visualization, the model highlights areas with high probability of land conversion, enabling more proactive planning responses. This study demonstrates the effectiveness of GeoAI in supporting spatial decision-making, providing an open, scalable, and replicable workflow for urban monitoring and prediction. The results reinforce the importance of data-driven governance and spatial intelligence in realizing IKN’s vision as a smart and sustainable capital city.