Evaluating Spatiotemporal Changes and Predicting Land-Use Dynamics Using an Artificial Neural Network in Worcester City, Massachusetts

Abstract The land use and land cover (LULC) change has a remarkable impact on the urban ecosystems due to its effect on the environmental processes, natural resources, and urban sustainability. The aim of this study was to conduct the spatiotemporal analysis of LULC in Worcester City, Massachusetts, and to identify the dynamics of forest cover and justify its causes in the given time range of 2003–2023. LULC classification and forest change detection were done with high-resolution datasets provided by the National Land Cover Database (NLCD). The MLP-ANN outperformed logistic regression and achieved higher in overall accuracy percentages and location Kappa values. Furthermore, a Multi-Layer Perceptron Artificial Neural Network (MLP-ANN) model embedded in the MOLUSCE plug-in of QGIS was used in order to forecast future LULC scenarios in 2033, 2043, and 2053. The findings showed that the built-up areas moved up from 82.1% to 83.2%, and the forest cover reduced from 12.4% to 11.9% over the period studied. Water bodies were relatively stable, wetlands were slightly raised, whereas the grasslands and barren lands recorded slight reductions. Forest change detection showed a total of 0.81% loss of forest, 0.03% gain of forest, and 99.10% change in forest that remained the same. The model validation was high (Kappa = 0.96), which proved that it can be used in the future to simulate it. The future LULC projections showed further urbanization and slow forest cover loss due to urbanization, control of pests, control of invasive species, and variability in climatic conditions. These results emphasize the necessity of implementing constant LULC monitoring during urban planning and policy regimes to make sure that the development is in balance with the minimum ecological impact. Graphical abstract The graphical abstract summarizes the study by visualizing the NLCD-based land cover data for Worcester City, Massachusetts, highlighting the high-resolution datasets used for analyzing LULC changes from 2003 to 2023. It illustrates the spatiotemporal analytical workflow, including LULC classification and forest change detection, used to quantify landscape transitions. The application of the MLP-ANN model within the MOLUSCE plug-in is depicted to show how future LULC scenarios for 2033, 2043, and 2053 were simulated. Key results are presented through comparative maps and statistics showing increased built-up areas, declining forest cover, and strong model accuracy (Kappa = 0.96). The final component visually conveys that continued urban expansion and gradual forest loss are expected, emphasizing the importance of ongoing LULC monitoring and sustainable planning to minimize ecological impacts.