Scholarlink

<p>Mountainous areas are undergoing rapid Land Cover (LC) changes due to climate change and human activity. However, the shortage of reliable training samples still hinders historical LC change monitoring in mountainous areas. Sample migration has recently been introduced to address this obstacle, but its efficiency in mountainous regions has not been well-documented yet. Additionally, previous research has solely utilized Spectral Angle Distance (SAD) and Euclidean Distance (ED) as metrics for similarity checking. In this study, we used a sample migration technique, Landsat imagery, and a cloud computing platform to investigate historical LC changes over the Alborz Mountain range of Iran between 1988 and 2023 at five-year intervals. Initially, we generated the LC map for the reference year (2023) using Landsat data and high-quality reference samples. We then assessed five similarity checking algorithms, namely SAD, ED, Census (CE), Kullback–Leibler (KL), and Multivariate Alteration Detection (MAD), to migrate high-quality samples from the reference year to the target years. Finally, we generated a set of eight 30-m LC maps for the Alborz Mountain at 5-year time intervals. Our analysis showed that the combination of KL and CE yields superior accuracy for migrating high-quality samples. This is because CE considers neighboring pixel information and KL effectively handles non-linear imaging differences. The produced LC maps based on the proposed sample migration technique achieved high overall accuracy, ranging from 88 to 90%. We also observed severe LC changes in the Alborz Mountains, with a dramatic decline in forest areas and a sharp increase in cropland and rangeland.</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

Evaluation of similarity-checking methods for reference samples migration to monitor land cover changes in the complex terrain of the Alborz Mountains, Iran

  • 期刊论文

  • 发表日期：

    2025年3月8日

查看全文
  • Meysam Moharrami,
  • Sara Attarchi,
  • Richard Gloaguen,
  • Seyed Kazem Alavipanah

摘要

Mountainous areas are undergoing rapid Land Cover (LC) changes due to climate change and human activity. However, the shortage of reliable training samples still hinders historical LC change monitoring in mountainous areas. Sample migration has recently been introduced to address this obstacle, but its efficiency in mountainous regions has not been well-documented yet. Additionally, previous research has solely utilized Spectral Angle Distance (SAD) and Euclidean Distance (ED) as metrics for similarity checking. In this study, we used a sample migration technique, Landsat imagery, and a cloud computing platform to investigate historical LC changes over the Alborz Mountain range of Iran between 1988 and 2023 at five-year intervals. Initially, we generated the LC map for the reference year (2023) using Landsat data and high-quality reference samples. We then assessed five similarity checking algorithms, namely SAD, ED, Census (CE), Kullback–Leibler (KL), and Multivariate Alteration Detection (MAD), to migrate high-quality samples from the reference year to the target years. Finally, we generated a set of eight 30-m LC maps for the Alborz Mountain at 5-year time intervals. Our analysis showed that the combination of KL and CE yields superior accuracy for migrating high-quality samples. This is because CE considers neighboring pixel information and KL effectively handles non-linear imaging differences. The produced LC maps based on the proposed sample migration technique achieved high overall accuracy, ranging from 88 to 90%. We also observed severe LC changes in the Alborz Mountains, with a dramatic decline in forest areas and a sharp increase in cropland and rangeland.