Comparison of three water indices for tropical aquaculture ponds extraction using Google Earth Engine

Abstract

Information on the spatial distribution of aquaculture ponds, especially the inland brackish aquaculture, is crucial for effective and sustainable aquaculture management. Google Earth Engine (GEE) has been utilized to quickly map aquaculture ponds in different parts of the world, but the application is still limited in tropical regions. Selection of an optimal water index is essential to accurately map the aquaculture ponds from the Landsat 8 satellite images that are available in GEE. This study aims to evaluate the capability of three different water indices, namely Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI) and Automated Water Extraction Index (AWEI), in mapping of the aquaculture ponds in Sungai Udang, Pulau Pinang, Malaysia. The results show that MNDWI is the best index for aquaculture ponds extraction in Sungai Udang, with an accuracy of 81.87% and Kappa coefficient of 0.61. Meanwhile, the accuracy of NDWI and AWEI as compared to the digitized aquaculture ponds are 58.21 and 61.60%, and Kappa coefficient of 0.33 and 0.36, respectively. Then, MNDWI was applied to calculate the spatial changes of aquaculture ponds from 2014 to 2020. The result indicates that the area of aquaculture ponds has expanded by 26.16% since the past seven years.