Improvement of glacial lakes detection under shadow environment using ASTER data in Himalayas, Nepal

The detection of glacial lake change in the Himalayas, Nepal is extremely significant since the glacial lake change is one of the crucial indicators of global climate change in this area, where is the most sensitive area of the global climate changes. In the Himalayas, some of glacial lakes are covered by the dark mountains’ shadow because of their location. Therefore, these lakes can not be detected by conventional method such as Normalized Difference Water Index (NDWI), because the reflectance feature of shadowed glacial lake is different comparing to the ones which are located in the open flat area. The shadow causes two major problems: 1) glacial lakes which are covered by shadow completely result in underestimation of the number of glacial lakes; 2) glacial lakes which are partly identified are considered to undervalue the area of glacial lakes. The aim of this study is to develop a new model, named Detection of Shadowed Glacial Lakes (DSGL) model, to identify glacial lakes under the shadow environment by using Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) data in the Himalayas, Nepal. The DSGL model is based on integration of two different modifications of NDWI, namely NDWIₛ model and NDWIₛₕₑ model. NDWIₛ is defined as integration of the NDWI and slope analysis and used for detecting non-shadowed lake in the mountain area. The NDWIₛₕₑ is proposed as a new methodology to overcome the weakness of NDWIₛ on identifying shadowed lakes in highly elevated mountainous area such as the Himalayas. The first step of the NDWIₛₕₑ is to enhance the data from ASTER 1B using the histogram equalization (HE) method, and its outcome product is named ASTERₕₑ. We used the ASTERₕₑ for calculating the NDWIₕₑ and the NDWIₛₕₑ. Integrated with terrain analysis using Digital Elevation Model (DEM) data, the NDWIₛₕₑ can be used to identify the shadowed glacial lakes in the Himalayas. NDWIₛ value of 0.41 is used to identify the glacier lake (NDWIₛ ≥ 0.41), and 0.3 of NDWIₛₕₑ is used to identify the shadowed glacier lake (NDWIₛₕₑ ≤ 0.3). The DSGL model was proved to be able to classify the glacial lakes more accurately, while the NDWI model had tendency to underestimate the presence of actual glacial lakes. Correct classification rate regarding the products from NDWI model and DSGL model were 57% and 99%, respectively. The results of this paper demonstrated that the DSGL model is promising to detect glacial lakes in the shadowed environment at high mountains.