Mapping Burned Forest Areas in Western Yunnan, China, Using Multi-Source Optical Imagery Integrated with Simple Non-Iterative Clustering Segmentation and Random Forest Algorithms in Google Earth Engine DOI Creative Commons
Yue Chen, Weili Kou, Wenna Miao

и другие.

Remote Sensing, Год журнала: 2025, Номер 17(5), С. 741 - 741

Опубликована: Фев. 20, 2025

This study aimed to accurately map burned forest areas and analyze the spatial distribution of fires under complex terrain conditions. integrates Landsat 8, Sentinel-2, MODIS data in western Yunnan. A machine learning workflow was developed on Google Earth Engine by combining Dynamic World land cover with official fire records, utilizing a logistic regression-based feature selection strategy an enhanced SNIC segmentation GEOBIA framework. The performance four classifiers (RF, SVM, KNN, CART) burn detection evaluated through comparative analysis their spectral–spatial discrimination capabilities. results indicated that RF classifier achieved highest performance, overall accuracy 96.32% Kappa coefficient 0.951. Spatial further revealed regions at medium altitudes (800–1600 m) moderate slopes (15–25°) are more prone fires. demonstrates robust approach for generating accurate large-scale maps provides valuable insights effective management areas.

Язык: Английский

Mapping Burned Forest Areas in Western Yunnan, China, Using Multi-Source Optical Imagery Integrated with Simple Non-Iterative Clustering Segmentation and Random Forest Algorithms in Google Earth Engine DOI Creative Commons
Yue Chen, Weili Kou, Wenna Miao

и другие.

Remote Sensing, Год журнала: 2025, Номер 17(5), С. 741 - 741

Опубликована: Фев. 20, 2025

This study aimed to accurately map burned forest areas and analyze the spatial distribution of fires under complex terrain conditions. integrates Landsat 8, Sentinel-2, MODIS data in western Yunnan. A machine learning workflow was developed on Google Earth Engine by combining Dynamic World land cover with official fire records, utilizing a logistic regression-based feature selection strategy an enhanced SNIC segmentation GEOBIA framework. The performance four classifiers (RF, SVM, KNN, CART) burn detection evaluated through comparative analysis their spectral–spatial discrimination capabilities. results indicated that RF classifier achieved highest performance, overall accuracy 96.32% Kappa coefficient 0.951. Spatial further revealed regions at medium altitudes (800–1600 m) moderate slopes (15–25°) are more prone fires. demonstrates robust approach for generating accurate large-scale maps provides valuable insights effective management areas.

Язык: Английский

Процитировано

0