Advances in Remote Sensing and Machine Learning for Land Use/Land Cover Change Detection and Climate Model Bias Correction: Methods and Applications

Abstract

Ecosystems globally are increasingly threatened by the integrated impacts of human activities and climate change, especially in the Ethiopian Rift Valley, where data are scarce. Therefore, there is an urgent need for implementing advances in sustainable resource management. This study presented a systematic meta-analysis that synthesized more than 100 peer-reviewed studies (2004-2025) to critically evaluate how remote sensing (RS) and machine learning (ML) could be integrated for detecting LULC change and climate model bias. Using a PRISMA protocol, the methodological progression from traditional statistical approaches was evaluated through advanced deep-learning and hybrid methods. The results provided strong evidence that integrating RS and ML had significant transformative potential; however, there still remained a "gap" in their integration. Both fields advanced greatly, yet each existed as a largely independent entity. Recent studies addressed this disintegration, however operational frameworks were scarce. In order to close this gap, a novel integrated conceptual framework was proposed. This would dynamically integrate multi-sensor data fusion, ML-based LULC mapping, hybrid ML-statistical bias correction, and SWAT+ hydrological models into a single end-to-end processing pipeline designed specifically for basins that contained little or no data. Transfer learning and explainable artificial intelligence (XAI) could be used to solve the problem of applying data-hungry models in situations where very little or no data were available. The integration of XAI methods such as SHAP and LIME gave a promise for improving model interpretability in environmental applications. The review suggests that robust-policy-relevant environmental prediction depended on designing combined systems that were not only accurate but also interpretable and adaptable to local conditions.

Keywords: Climate Model Bias Correction; Data Integration; Explainable AI; Land Cover Change Detection; Machine Learning; Remote Sensing