A Review on Future Land Use/ Land Cover (LULC) Prediction Techniques: Global Methodological Shifts and Publication Trends

Abstract

Future Land Use/Land Cover (LULC) prediction is essential for sustainable urban planning, environmental management, and climate resilience. This review aims to systematically examine global methodological shifts, model performance, software ecosystems, and publication trends in future LULC prediction research from 2015 to 2025. Specifically, it seeks to identify the methodologies used for LULC modeling, evaluate their strengths and weaknesses, determine the predominantly used classifiers, assess available software and data portals, and analyze the temporal, geographic, and publisher-wise distribution of related studies. The review synthesizes traditional approaches such as Cellular Automata–Markov (CA–MC), Logistic Regression, and Agent-Based Models alongside machine learning techniques including Random Forest (RF), Support Vector Machines (SVM), and Artificial Neural Networks (ANN), as well as deep learning architectures such as CNN and LSTM. A comparative assessment highlights trade-offs in accuracy, interpretability, computational demand, and spatial–temporal modeling capability. Bibliometric analysis indicates a rapid increase in publications after 2020, reflecting the growing integration of hybrid and data-driven frameworks supported by cloud-based geospatial platforms and open satellite datasets. The study provides consolidated understandings into evolving research directions and supports informed model selection for future LULC change prediction.