Evaluation of Land Cover Change and Projection of Future Data by Using Two Statistical Models

Land cover serves as a primary driver of ecological processes, undergoing continuous transformation through natural events and anthropogenic activities. Precise identification of these patterns is essential for effective land-use planning. By integrating historical land cover data with statistical and mathematical models, researchers generate future scenarios that guide local territorial management. This study evaluates two predictive approaches: Ordinary Least Squares (OLS) and Non-Spatial Markov Chain (MC) to project landscape transitions for the year 2035 in Sorocaba, SP, Brazil, utilizing an Artificial Intelligence (AI) platform. The results reveal continuous, differentiated growth in urban areas, a trend projected to persist. Conversely, the pasture category exhibits a marked and sustained loss of territory across all scenarios. The Remaining Natural Vegetation category shows historical fluctuations and yields divergent predictions: the OLS algorithm suggests stability, whereas the MC algorithm predicts an increase. Model performance evaluation, validated against 2025 data, demonstrates that the OLS algorithm achieves greater accuracy and higher similarity between mapped and calculated datasets. Ultimately, AI platforms streamline environmental modeling, and their rapid evolution ensures increasingly precise and consistent future spatial predictions.