Scholarlink

Riverbank erosion and accretion are prevalent phenomena along the Teesta floodplain in Bangladesh. The present chapter was designed to determine the erosion-accretion on ArcGIS 10.8 using Iso cluster unsupervised classifications. The estimation of land use and land cover (LULC) changes was done through the Support Vector Machine (SVM) classifier grounded on the cloud-based Google Earth Engine (GEE). The SVM approach has been observed to exhibit comparable performance to the most advanced machine learning algorithms when applied to the categorizations of datasets with many dimensions. The SVM is a supervised classifier gaining popularity in remote sensing (RS) applications. The variability of classification precision in the SVM classifier can be attributed to the selection of the internal function and its associated variables. Landsat 5 TM images for 1988, 1998, 2004, and 2008 and Landsat 8 OLI images for 2014, 2018, and 2022 were used to consider the significant flood events. The highest erosion (73.55 km2) occurred in 2004, whereas the most elevated deposition (76.51 km2) was found in 1998. The overall accuracy of 0.93 was higher than the Kappa accuracy of 0.92. The LULC transformation was very unusual due to the rapid change of physiography with unusual floods and irregular riverbank erosion in the study area. The results showed that the agricultural and vegetation area was converted to a built-up area, which was then converted to agricultural land from 1988 to 2022.

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

Riverbank Erosion-Accretion and Land Use Changes Using Geo-Spatial and Machine Learning Technique of Teesta River, Bangladesh

  • 章节

  • 发表日期：

    2025年

查看全文
  • Most. Mitu Akter,
  • N. M. Refat Nasher

摘要

Riverbank erosion and accretion are prevalent phenomena along the Teesta floodplain in Bangladesh. The present chapter was designed to determine the erosion-accretion on ArcGIS 10.8 using Iso cluster unsupervised classifications. The estimation of land use and land cover (LULC) changes was done through the Support Vector Machine (SVM) classifier grounded on the cloud-based Google Earth Engine (GEE). The SVM approach has been observed to exhibit comparable performance to the most advanced machine learning algorithms when applied to the categorizations of datasets with many dimensions. The SVM is a supervised classifier gaining popularity in remote sensing (RS) applications. The variability of classification precision in the SVM classifier can be attributed to the selection of the internal function and its associated variables. Landsat 5 TM images for 1988, 1998, 2004, and 2008 and Landsat 8 OLI images for 2014, 2018, and 2022 were used to consider the significant flood events. The highest erosion (73.55 km2) occurred in 2004, whereas the most elevated deposition (76.51 km2) was found in 1998. The overall accuracy of 0.93 was higher than the Kappa accuracy of 0.92. The LULC transformation was very unusual due to the rapid change of physiography with unusual floods and irregular riverbank erosion in the study area. The results showed that the agricultural and vegetation area was converted to a built-up area, which was then converted to agricultural land from 1988 to 2022.