<p>Understanding long-term land-use and land-cover (LULC) change processes is important for achieving sustainability in urban development and environmental management. This research assessed the spatio-temporal patterns of LULC change within the coastal cities of Jiangsu Province, China; specifically, Yancheng, Taizhou, and Nantong during the time periods of 2004, 2014, and 2024, through the application of a multi-temporal Landsat remote sensing data set. A supervised Random Forest-based classification was employed to generate LULC maps, followed by change detection and transition analysis. Furthermore, the Cellular Automata-Markov (CA–Markov) model was applied to simulate and predict future LULC changes. The classification accuracy assessments indicated high overall accuracies of 89.60%, 90.76%, and 92.33% and Kappa coefficients of 0.878, 0.892, and 0.910 for 2004, 2014, and 2024, respectively. However, the dominant land use category was crops, which declined by 1,055.10 km² between 2004 and 2024. Conversely, Buildup areas increased significantly at an annual rate of 1,086.89 km². It was also found that the greatest land conversion occurred from cropland to built-up areas, resulting in a loss of 1,957.11 square kilometers of cropland. An evaluation of the model’s performance compared the predicted and observed LULC maps for 2024 and demonstrated a high degree of conformity, supporting the use of the CA–Markov method as a reliable tool for modeling future land-use trends. The projected results for 2034 indicate that urbanization will continue at an average annual rate of 138.41 km², while cropland and wetland areas will decrease at rates of 180.72 and 229.40 km² per annum, respectively. Overall, the results indicate that the primary driver of land transformation remains urban growth, underscoring the importance of implementing sustainable land-use strategies to preserve agricultural and ecological environments.</p>