EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Performance of water indices using large-scale sentinel-2 data in Google Earth Engine Computing

Mathias Tesfaye Abebe1 and Lutz Breuer1,2
Mathias Tesfaye Abebe and Lutz Breuer
  • 1Institute for Landscape Ecology and Resources Management (ILR), Research Centre for BioSystems, Land Use and Nutrition (iFZ), Justus Liebig University Giessen, Giessen, Germany (
  • 2Centre for International Development and Environmental Research (ZEU), Justus Liebig University Giessen, Giessen, Germany

Evaluating the performance of water indices and quantifying the spatial distribution of water-related ecosystems are important for monitoring surface water resources of our study area since there is a limited study available to compute water indices using high-resolution and multi-temporal sentinel-2 data on a large scale. In addition, a comparative performance analysis of water indices methods using the aforementioned dataset on a country scale, showing their strengths and weaknesses, was missing too. To address these problems, this paper evaluated the performance of water indices for surface water extraction in Ethiopia. For this purpose, high spatial and multi-temporal resolution large-scale sentinel-2 data were employed and processed using the Google Earth Engine cloud computing system. In this study, seven indices, namely water index (WI) and automatic water extraction index (AWEI) with shadow and no shadow, normalized difference water index (NDWI), modified normalized difference water index (MNDWI), sentinel water index (SWI), and land surface water index (LSWI) were evaluated with overall accuracy, producer’s accuracy, user’s accuracy, and Kappa coefficient. The result revealed that the WI and AWEIshadow were the most accurate to extract the surface water compared to other indices in qualitative and quantitative evaluation of accuracy indicators obtained with a kappa coefficient of 0.96 and 0.95, respectively, and with overall accuracy for both in 0.98. In addition, the AWEIshadow index was also relatively better at suppressing shadow and urban areas. The accuracy difference between LSWI and other indices was significant which performed the worst with overall accuracy and kappa coefficients of 0.82 and 0.31, respectively. Using best-performing indices of WI and AWEIshadow, 82650 and 86530 square km of surface water fractions were extracted, respectively. Therefore, our result confirmed that WI and AWEIshadow indices generated better water extraction outputs using a high spatial and multi-temporal resolution of sentinel-2 data under a wide range of environmental conditions and water body types on the country scale.

How to cite: Abebe, M. T. and Breuer, L.: Performance of water indices using large-scale sentinel-2 data in Google Earth Engine Computing, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-2395,, 2023.