Tracking rapid mariculture expansion in Xuan Dai Bay, Viet Nam (2015 - 2024) with Sentinel-1 SAR time-series imagery
Xuan Truong Trinh (a*), Wataru Takeuchi (b), Masahiko Nagai (c)

a) Faculty of Engineering, Yamaguchi University, Japan
*trinh[at]yamaguchi-u.ac.jp
b) Institute of Industrial Sciences, The University of Tokyo, Japan
c) Yamaguchi University Center for Research and Application of Remote Sensing, Yamaguchi University, Japan

Abstract

Mariculture increasingly underpins global food security, yet its accelerated growth-especially in developing nations-often outpaces regulatory capacity and threatens coastal ecosystems. This study offers the first decadal, object-based time-series assessment of floating mariculture infrastructure in Xuan Dai Bay, central Viet Nam, derived from Sentinel-1 C-band synthetic-aperture radar (SAR) imagery acquired between 2015 and 2024. Multi-temporal Lee filtering and annual median compositing suppressed speckle and wave-induced noise, enhancing the backscatter signal of cage arrays. Image objects were generated with the Simple Non-Iterative Clustering (SNIC) algorithm and classified using a Random Forest model that fused backscatter statistics with geometric metrics. The approach achieved an overall accuracy of 70 %, with empty water surfaces mapped with 100 % producer accuracy, reflecting their consistently low backscatter. Commission errors were concentrated among cages, ponds and near-shore structures, indicating the value of auxiliary optical data or refined geometric descriptors. Time-series analysis reveals a 3.4-fold increase in cage area and a marked intensification between 2018 and 2022. Such expansion risks exacerbating eutrophication, habitat degradation and biodiversity loss. Our findings demonstrate the utility of freely available SAR archives for routine, large-scale surveillance of offshore aquaculture and highlight the pressing need for stronger management frameworks in Viet Nam. Future work should integrate higher-resolution SAR and multispectral sensors to resolve small cages and quantify water-quality impacts in situ, enabling more precise environmental assessments and evidence-based policy making.

Keywords: coastal management, environment quality, SAR, object-based image analysis, machine learning

Topic: Topic B: Applications of Remote Sensing