Arid Zone Research ›› 2022, Vol. 39 ›› Issue (3): 774-786.doi: 10.13866/j.azr.2022.03.11

Previous Articles     Next Articles

An automatic method for delineating lake surfaces in Qaidam Basin using Landsat images

WEN Guangchao(),LI Xing,WU Bingjie,WANG Xiaohe,XIE Hongbo   

  1. Institute of Resource and Environment, Henan Polytechnic University, Jiaozuo 454000, Henan, China
  • Received:2021-08-30 Revised:2021-10-06 Online:2022-05-15 Published:2022-05-30

Abstract:

Lakes in the Qaidam Basin maintain regional ecological balance, supply production and domestic water consumption, and protect the ecological environment. These lakes have undergone a series of changes in response to enhanced regional climate change and human activities. To analyze characteristics of lake changes and investigate casual factors, an automatic method for lake surface delineation based on Landsat remote sensing images with a different series of sensors (the Lake Water Differential Model) was proposed by understanding TOA (top-of-atmosphere) reflectance differences between water and non-water surfaces in the Koruk Lake Basin. This model is applied to extract lake information at varying times or locations. Model performance is evaluated using overall accuracy, the Kappa coefficient, and user’s accuracy, and the derived lake surfaces were compared to those from the NDWI (Normalized Difference Water Index) and MNDWI (Modified Normalized Difference Water Index) methods. These results show the following: (1) target and non-target objects are distinguished by a TOA reflectance difference. (2) Based on a stable threshold, the Lake Water Differential Model can delineate lake surfaces. Compared with the NDWI, MNDWI, and other water body information extraction methods, the Lake Water Differential Model can more effectively suppress interference factors, such as surface rivers, ice, snow, shadows, swamps, and wetlands. In areas of model application, our proposed method can achieve a performance of 99.48% of the average OA, 99.66% of user accuracy, and 0.9877 of Kappa. (3) Input data of the model can be either TOA or Landsat (level-2) surface reflectance data. (4) The model is suitable for extracting lake water information in a large area of Qaidam Basin and provides technical support for the study of lake surface dynamics.

Key words: Landsat, lake, TOA reflectance, lake water differential model, Qaidam Basin