Satellite radar and AIS reveal a 97% decline in shipping traffic through the Strait of Hormuz

Dear Editor,

The Strait of Hormuz serves as a critical choke point linking the Persian Gulf to global markets, accounting for approximately 20% of global maritime crude oil shipments.1 On February 28, 2026, a joint US-Israeli strike against Iran prompted Iran to announce an immediate and temporary blockade of the strait.2 This escalation increased navigation safety risks and contributed to volatility in global oil prices.

In response to this crisis, this study conducts a rapid event-driven quantitative assessment of maritime traffic dynamics by integrating Sentinel-1 radar imagery with Automatic Identification System (AIS) data over the week preceding and the week following the February 28 event (February 22–March 7, 2026). The Sentinel-1 radar satellite offers all-weather imaging capabilities at large spatial scales, enabling the detection of vessel spatial distributions regardless of cloud cover.3 In comparison, AIS can provide real-time positions, identity details, and behavioral patterns of vessels.4 This study contributes to maritime situational awareness by the combined use of satellite radar imagery and AIS data during the geopolitical incident.

We developed an automated method for rapidly assessing navigation dynamics, which consists of two modules. The first module is threshold-based ship detection using Sentinel-1 radar imagery. Specifically, we acquired 258 scenes of 10-m-spatial-resolution Sentinel-1 ground range detected (GRD) products in interferometric wide (IW) swath mode from the Copernicus Programme. All data were preprocessed on the Google Earth Engine (GEE) platform following standard protocols, i.e., orbit file correction, border and thermal noise removal, radiometric calibration, and terrain correction. The backscatter coefficients (σ0) recorded in Sentinel-1 were then converted to decibels (dB). Potential ship candidates were identified by selecting pixels exceeding an empirical threshold of −15 dB and were further refined to exclude land artifacts using a high-precision ocean mask.5 A connectivity analysis was then applied to eliminate isolated noise clusters (<3 pixels). To rule out data artifacts, we performed manual visual inspections. The ship detection method achieved an F1-score of 96.0%, with a precision of 94.2% and a recall of 97.9% (Figure 1E), on 18 test images covering congested port and open-water conditions. Finally, to mitigate temporal biases arising from uneven satellite coverage, we introduced the observed average count (OAC) metric, defined as the mean number of ships detected per satellite image within each grid (0.25° × 0.25° grid is shown in Figure 1A1).