Parlung Zangbo Mountain Glacier Dataset (2015) Supporting Machine Learning by Random Forest Algorithm for Identifying Glacier Pixels

ZHANG Jingxiao1JIA Li1Massimo MENENTI1HU Guangcheng1

1 State Key Laboratory of Remote Sensing Science，Institute of Remote Sensing and Digital Earth，Chinese Academy of Sciences，Beijing 100101，China

DOI:10.3974/geodb.2019.02.08.V1

Published:Feb. 2019

Visitors：7444       Data Files Downloaded：96      
Data Downloaded：67.70 MB      Citations：

Key Words：

machine leaning,debris-covered glacier,Parlung Zangbo,Qinghai-Tibetan Plateau,2015

Abstract：

By integrating the multiple temporal Landsat images and ASTER GDEM V2 (30 m), SRTM DEM (30 m) and TanDEM-X DEM (90 m), the Parlung Zangbo glacier dataset supports the machine learning with the Random Forest Algorithm for classification, which generates 100 binary decision trees to train the classifier. Each decision tree is individually constructed and trained based on random input training samples and random input features. The forest chooses the most popular class as the final result to identify the glacier pixels. The Parlung Zangbo basin is located in the southeastern Tibetan Plateau, a special region of the most important and concentrated regions of maritime glacier distribution in China, its geo-location is 28.5°N–29.8°N, 95.5°E–97.5°E. Two classes of the pixels had been identified, they are: (1) non-or-partially-debris-covered glacier includes snow and clean ice which is free of debris cover or parts of the glacier barely covered by debris in the ablation zone; (2) fully-debris-covered glacier represents the glacier parts with extensive amount of debris covering clean ice. The dataset consists of three data file groups: (1) geographic extent of the study region in .shp data format; (2) glacier distribution data in .tif format, including total glacier distribution data file, the non-or-partially-debris-covered glacier data file and fully-debris-covered glacier data file; (3) 382 validation samples data file in .shp format. The dataset is consisted of 29 data files and archived in .shp and .tif data format with data size of 145 MB (Compressed to one single file with data size of 722 KB).

Foundation Item：

Chinese Academy of Sciences (XDA19030203, 131C11KYSB20160061); National Natural Science Foundation of China (91737205)

Data Citation：

ZHANG Jingxiao, JIA Li, Massimo MENENTI, HU Guangcheng. Parlung Zangbo Mountain Glacier Dataset (2015) Supporting Machine Learning by Random Forest Algorithm for Identifying Glacier Pixels[J/DB/OL]. Digital Journal of Global Change Data Repository, 2019. https://doi.org/10.3974/geodb.2019.02.08.V1.

Data Product: