Multisensor Visual Image Design Based on Machine Learning to Evaluate Urban Land Transformation and Ecological Environment Variation

The accurate detection and monitoring of seasonal lake dynamics and rangeland variations in high-altitude regions present significant challenges, particularly during snow-covered periods. This study introduces an integrated multisensor approach combining Synthetic Aperture Radar (SAR) and optical remote sensing and visual image design for monitoring seasonal changes in Saif ul Malook Lake and surrounding rangelands, Pakistan. We leveraged Google Earth Engine’s cloud computing capabilities to process and analyze Sentinel-1 SAR and Landsat imagery, implementing Random Forest classification (achieving 92% accuracy during melt season) for land use/land cover mapping, including rangeland delineation using visual image design. The methodology incorporated multiple water indices (Automated Water Extraction Index, Modified Normalized Difference Water Index, and Normalized Difference Water Index) and topographic parameters derived from digital elevation models. Additionally, visual image design was applied to improve the clarity of multisensor observations, enabling more intuitive detection of seasonal transitions in lake and rangeland conditions. This enhancement supported better interpretation and strengthened the overall monitoring framework. During the melt season (October–November), optical indices successfully detected the lake extent (2.8 km²) with high accuracy (>95%). However, their performance significantly decreased during snow-covered periods (January–February), with accuracy dropping to approximately 60%. SAR-based detection maintained consistent performance across seasons, successfully identifying lake extent even under snow cover. Land use classification revealed significant seasonal variations, with vegetation cover and rangeland areas decreasing from 45% to 15% during snow-covered periods, while snow/ice coverage expanded to 65% of the study area. The integration of Topographic Wetness Index and stream flow analysis provided crucial context for understanding the lake’s hydrological connectivity and its impact on adjacent rangeland ecosystems. This study demonstrates the effectiveness of combining SAR and optical remote sensing and visual image design for year-round lake and rangeland monitoring, particularly in challenging high-altitude environments. The findings highlight the importance of multisensor approaches and machine learning techniques for accurate lake detection and rangeland assessment under varying seasonal conditions, contributing to improved understanding of high-altitude lake dynamics and rangeland responses to environmental change.