Mapping inland pond aquaculture for the coastal zone of Asia: an object-based, multi-sensor approach using Sentinel-1 and Sentinel-2 time-series

Abstract

Asia is the world's largest regional aquaculture producer, accounting for 88 percent (75 million tons) of the total global production, and has been the main driver of global aquaculture growth in recent years. The five largest aquaculture producing countries all come from Asia: China, India, Indonesia, Vietnam and Bangladesh. The farming of fish, shrimp, and mollusks in land-based pond aquaculture systems contributed most to Asia's dominant role in the global aquaculture sector, serving as the primary source of protein for millions of people. Aquaculture expanded rapidly since the 1990s in low-lying areas with flat topography along the coasts of Asia, particularly in Southeast Asia and East Asia. As a result of the rapid global growth of aquaculture in recent years, the mapping and monitoring of aquaculture are a focus in coastal research and plays an important role in global food security and the achievement of the UN Sustainable Development Goals. We present a novel continental scale mapping approach that uses multi-sensor Earth observation time series data to extract pond aquaculture within the entire Asian coastal zone, defined as the onshore area up to 200km from the coastline. With free and open access to the rapidly growing volume of high-resolution C-band SAR and multispectral satellite data from the Copernicus Sentinel missions as well as machine learning algorithms and cloud computing services, we automatically detected and extracted pond aquaculture on a single pond unit level. For this purpose, we processed more than 25,000 Sentinel-1 dual-polarized GRDH images, generated a temporal median image and applied image segmentation using histogram-based thresholding. The derived object-based pond units were enriched with multispectral time series information derived from Sentinel-2 L2A data, topographical terrain information, geometric features and Open Street Map data in order to detect coastal pond aquaculture and separate them from other natural or artificial water bodies. In total, we mapped more than 3.4 million aquaculture ponds with a total area of 2 million ha with a mean average overall accuracy of 0.91 and carried out spatial and statistical data analyses in order to investigate the spatial distribution and to identify production hotspots in various administrative units at regional, national, and sub-national scale.