Scholarlink

<p>Accurate land-use and land-cover (LULC) classification in mountainous regions is essential for sustainable environmental management. However, challenges such as spectral confusion, complex topography, and sensor-related noise hinder classification accuracy. This study presents a novel framework, Multi-Index Fusion for Enhanced LULC Sampling (MIFE-LULC), developed to improve classification performance in the Lidder and Kuthar watersheds of the Himalayan Jhelum Basin between 2013 and 2023. The approach integrates multi-temporal Landsat 8 and 9 satellite imagery with topographic variables, analyzed using machine learning classifiers: Support Vector Machine (SVM), Random Forest (RF), and Classification and Regression Trees (CART), within the Google Earth Engine (GEE) platform. The MIFE-LULC method leverages four targeted spectral indices; Normalized Difference Vegetation Index (NDVI), Dry Built-up and Soil Index (DBSI), Modified Normalized Difference Snow Index (MNDSI), and Enhanced Built-up and Bareness Index (EBBI) to enhance class separability among spectrally similar classes. Among the classifiers, SVM demonstrated the highest performance, achieving an overall accuracy of 99% and a kappa of 0.99. Over the decade, urban areas expanded by 146.9%, and plantation cover increased by 121%. In contrast, cropland and water bodies declined by 58.2% and 77.2%, respectively. These transitions highlight growing anthropogenic pressures and climate-related stressors, raising concerns about ecological vulnerability. The findings underscore the effectiveness and scalability of the MIFE-LULC framework for high-precision LULC classification in complex terrains. The resulting datasets and trend analyses offer critical inputs for land-use planning, disaster risk reduction, and progress monitoring toward Sustainable Development Goals (SDGs).</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

Integrating spectral index fusion with machine learning for improved LULC classification and temporal analysis: a case study from the Northwestern Himalayas

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • Muzamil Hassan Lone,
  • Amit B. Mahindrakar,
  • K. Kumar

摘要

Accurate land-use and land-cover (LULC) classification in mountainous regions is essential for sustainable environmental management. However, challenges such as spectral confusion, complex topography, and sensor-related noise hinder classification accuracy. This study presents a novel framework, Multi-Index Fusion for Enhanced LULC Sampling (MIFE-LULC), developed to improve classification performance in the Lidder and Kuthar watersheds of the Himalayan Jhelum Basin between 2013 and 2023. The approach integrates multi-temporal Landsat 8 and 9 satellite imagery with topographic variables, analyzed using machine learning classifiers: Support Vector Machine (SVM), Random Forest (RF), and Classification and Regression Trees (CART), within the Google Earth Engine (GEE) platform. The MIFE-LULC method leverages four targeted spectral indices; Normalized Difference Vegetation Index (NDVI), Dry Built-up and Soil Index (DBSI), Modified Normalized Difference Snow Index (MNDSI), and Enhanced Built-up and Bareness Index (EBBI) to enhance class separability among spectrally similar classes. Among the classifiers, SVM demonstrated the highest performance, achieving an overall accuracy of 99% and a kappa of 0.99. Over the decade, urban areas expanded by 146.9%, and plantation cover increased by 121%. In contrast, cropland and water bodies declined by 58.2% and 77.2%, respectively. These transitions highlight growing anthropogenic pressures and climate-related stressors, raising concerns about ecological vulnerability. The findings underscore the effectiveness and scalability of the MIFE-LULC framework for high-precision LULC classification in complex terrains. The resulting datasets and trend analyses offer critical inputs for land-use planning, disaster risk reduction, and progress monitoring toward Sustainable Development Goals (SDGs).