Object DetectionCoastal MonitoringChange DetectionTime-Series SatellitePatent

Aquaculture pond detection and change analysis

Monash University -- Filed Patent IDS000010594, Issued June 2025 · Lead Developer

Illegal aquaculture expansion -- primarily shrimp and fish ponds -- along Southeast Asian coastlines is a leading driver of mangrove deforestation. Indonesia and Vietnam have lost significant coastal mangrove cover in the past two decades. Regulatory monitoring relied on periodic manual surveys, creating enforcement gaps measured in years between detection and intervention.

Pond detection from optical satellite imagery is non-trivial due to spectral similarity between aquaculture ponds and natural water bodies (rivers, tidal flats, estuaries). Tidal variation causes the same geographic location to appear as water or land depending on acquisition timing. Cloud cover over tropical coastlines further degrades single-date classification reliability. The core challenge is temporal: distinguishing stable natural water bodies from newly constructed ponds requires multi-temporal analysis across image stacks, not single-image classification.

Developed a multi-temporal detection pipeline on Sentinel-2 imagery combining: (1) spectral indices (NDWI, MNDWI, AWEI) for initial water body candidate extraction, (2) geometric regularity filters to distinguish rectangular anthropogenic pond structures from organic natural water body shapes, and (3) temporal differencing across image stacks to identify newly constructed ponds within user-defined monitoring windows.

Pipeline validated across Indonesia (Sumatra, Kalimantan coastlines) and Vietnam (Mekong Delta, Ca Mau province) -- geographies with distinct spectral and tidal characteristics, testing cross-regional generalization.

Outputs feed a coastal land-use change analysis linking pond expansion rates to mangrove coverage loss, providing spatial evidence for policy reporting on illegal conversion.

Supports government and NGO monitoring of illegal coastal land conversion at satellite speed versus field survey cadence. Results can be delivered within hours of a new satellite pass, enabling near-real-time enforcement alerts for regulatory agencies managing coastal biodiversity and carbon stock in Indonesia and Vietnam.

PythonGoogle Earth EngineNDWIMNDWIAWEISentinel-2 (multi-temporal)