Scholarlink

<p>Mountain ecosystems, particularly in the Western Himalayas, are highly sensitive to climate variability and anthropogenic pressures. This study investigates climate-driven land use and land cover (LULC) changes across the Lahaul region over a 30-year period (1994–2024), integrating multi-temporal remote sensing data with climatic trends and machine learning classification techniques. Using Random Forest (RF) and Support Vector Machine (SVM) classifiers, our results show that mean annual temperatures rose from 0.53&#xa0;°C to 0.67&#xa0;°C, maximum temperatures increased from 10.81&#xa0;°C to 12.37&#xa0;°C, and average daily precipitation grew from 2.30&#xa0;mm to 2.97&#xa0;mm. Between 1994 and 2024, forest cover in Lahaul declined significantly from 206.99 sq km to 132.52 sq km (RF), while SVM estimates showed a similar decrease from 210.66 sq km to 140.19 sq km, indicating ecosystem stress due to climatic changes. Concurrently, scrub and grassland areas declined from 1503.65 sq km to 1404.41 sq km (RF) and from 1424.52 sq km to 1447.04 sq km (SVM), reflecting land pressure and classification variation. These trends underscore landscape degradation driven by rising temperatures and changing land use dynamics, while urban settlements expanded substantially, correlating with tourism and the cultivation of climate-resilient crops. Surface albedo also increased from 0.21% to 0.43%, suggesting land degradation or reduced vegetation density. Machine learning classifiers improved LULC detection: RF accuracy increased from 79.32% (1994) to 94.33% (2024), with Kappa coefficients rising from 0.79 to 0.93. Producer accuracy for settlements and agriculture also reached 96.55% in 2024. These findings underscore the urgent need for climate-informed land use policies, sustainable tourism planning, and adaptive resource management. AI-based geospatial analytics offer, scalable tools to monitor high-mountain environments and guide evidence-based decision-making in ecologically fragile regions like the Western Himalayas.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Climate-driven landscape transitions and emerging ecological instability in the Western Himalayas revealed by machine learning (1994–2024)

  • 期刊论文

  • 发表日期：

    2025年12月10日

查看全文
  • Akash Kashyap,
  • Amit Kumar

摘要

Mountain ecosystems, particularly in the Western Himalayas, are highly sensitive to climate variability and anthropogenic pressures. This study investigates climate-driven land use and land cover (LULC) changes across the Lahaul region over a 30-year period (1994–2024), integrating multi-temporal remote sensing data with climatic trends and machine learning classification techniques. Using Random Forest (RF) and Support Vector Machine (SVM) classifiers, our results show that mean annual temperatures rose from 0.53 °C to 0.67 °C, maximum temperatures increased from 10.81 °C to 12.37 °C, and average daily precipitation grew from 2.30 mm to 2.97 mm. Between 1994 and 2024, forest cover in Lahaul declined significantly from 206.99 sq km to 132.52 sq km (RF), while SVM estimates showed a similar decrease from 210.66 sq km to 140.19 sq km, indicating ecosystem stress due to climatic changes. Concurrently, scrub and grassland areas declined from 1503.65 sq km to 1404.41 sq km (RF) and from 1424.52 sq km to 1447.04 sq km (SVM), reflecting land pressure and classification variation. These trends underscore landscape degradation driven by rising temperatures and changing land use dynamics, while urban settlements expanded substantially, correlating with tourism and the cultivation of climate-resilient crops. Surface albedo also increased from 0.21% to 0.43%, suggesting land degradation or reduced vegetation density. Machine learning classifiers improved LULC detection: RF accuracy increased from 79.32% (1994) to 94.33% (2024), with Kappa coefficients rising from 0.79 to 0.93. Producer accuracy for settlements and agriculture also reached 96.55% in 2024. These findings underscore the urgent need for climate-informed land use policies, sustainable tourism planning, and adaptive resource management. AI-based geospatial analytics offer, scalable tools to monitor high-mountain environments and guide evidence-based decision-making in ecologically fragile regions like the Western Himalayas.