Seeing What AIS Cannot: Integrating Satellite Vessel Tracking with Administrative Vessel Registers to Produce Experimental Fishing-Effort Statistics for Indonesia
abstract
Satellite-derived vessel tracking is routinely presented as a ready-made replacement for conventional fisheries statistics. This note argues, and quantifies, the opposite: automatic identification system (AIS) data is a powerful but structurally partial observer, and its value to an official statistics system is realised only when read against the administrative records describing the population it fails to see. We build a reproducible pipeline that aggregates Global Fishing Watch (GFW) gridded apparent fishing effort into Indonesia’s own fisheries management areas (WPP-NRI), joins it to synthetic aperture radar (SAR) vessel detections and to the Ministry of Marine Affairs and Fisheries (KKP) vessel register, and publishes the result as a public, zone-based experimental statistic at https://ais.angkabatam.id. Three independent sources converge. The KKP register shows 98.57 per cent of Indonesia’s 795,736 registered fishing vessels sit below the 30 gross tonnage (GT) threshold at which vessel monitoring is compulsory. AIS observed 21,510 vessel presences across the eleven zones in 2024, equivalent to 2.7 per cent of the registered fleet. Of 654,931 radar detections, 52.2 per cent could not be matched to a broadcasting vessel — an indicator of maritime traffic generally rather than of fishing specifically, since only 2.8 per cent of matched detections are fishing vessels. A gear concordance shows 26 licensed Indonesian gear types, covering 510,076 registered units, with no counterpart in the GFW vocabulary at all. We conclude that AIS-derived effort is publishable as an experimental statistic about the industrial and foreign-flagged segment specifically, not about the fishery, and set out the provider-side changes that would strengthen the integration.
Live dashboard: ais.angkabatam.id. It's open access with no login, and every zone-level figure in the note can be checked there.
What we built
A reproducible extract–aggregate–serve pipeline (Laravel 10 on PostgreSQL/PostGIS) that:
- Pulls Global Fishing Watch apparent fishing effort (gridded, monthly, by flag) and Sentinel-1 SAR vessel detections for Indonesia's eleven WPP-NRI fisheries management areas
- Attributes every grid cell to a zone by point-in-polygon against the official KKP geometry, splitting domestic (IDN) from foreign effort
- Detects hotspots with
ST_ClusterDBSCAN(eps 0.25°, at least 4 cells), backed by a full parameter sweep - Joins the KKP vessel register, gear register and capture-production series as population bounds rather than record linkage
- Serves stored aggregates only: the public dashboard never calls an external API, and every run is logged so each figure traces back to a retrieval
Key findings
- 98.57% of Indonesia's 795,736 registered fishing vessels are below the 30 GT threshold at which vessel monitoring becomes compulsory
- 2.70% of the registered fleet was observed by AIS in 2024 (21,510 vessel presences)
- 52.2% of 654,931 SAR radar detections matched no broadcasting vessel. Only 2.8% of matched detections are fishing vessels, so this measures non-broadcasting maritime traffic in general, not "dark fishing"
- 64.1% of the fleet (510,076 units across 26 licensed gear types) uses gear with no counterpart in the GFW vocabulary, led by handlines and drifting gillnets
- 84.4% of observed effort in WPP 711 (Natuna, South China Sea) is foreign-flagged, a zone-level figure no conventional Indonesian source currently reports
Why it matters
AIS is not a sample of the fishing fleet. It is a census of the vessels large enough to carry a transponder. Read against the administrative register, it becomes a defensible experimental statistic about the industrial and foreign-flagged segment, and the gap itself becomes an indicator of how far the monitoring system reaches.
The note closes with six concrete requests to KKP as data provider, starting with publishing capture production by WPP zone.