Port Cargo Throughput and Vessel Congestion Analysis

Measure real cargo throughput and anchorage congestion at major global ports. Klarety processes Sentinel-2 optical imagery to count vessels at berth and in anchorage, estimate container stack density, and track berth utilization week-over-week. Logistics planners use reports to anticipate delays and optimize scheduling.

Analyze port activity

Port throughput intelligence without AIS reliance

Klarety AI

Congestion reports from optical satellite passes

Klarety agents count vessels at berth and anchorage from Sentinel-2 and return weekly throughput and utilization reports.

Sentinel-2

Vessel counting and container stack density

Agents classify vessel presence by size class, compute anchorage occupancy, and estimate container yard density from optical imagery.

GIS Output

Port activity layers for freight GIS platforms

Export vessel count polygons and berth utilization grids for port operations and freight visibility system integration.

Optical vessel counting and berth utilization scoring

Klarety agents detect vessel presence in Sentinel-2 imagery by identifying bright linear targets against port water, classify size by pixel footprint, and count vessels per berth zone and anchorage polygon. Berth utilization is scored as occupied berths divided by total berths. Output is a weekly utilization table and vessel count GeoJSON.

Klarety AIAnalyze port activity at your terminal

analysis/singapore_port_congestion.pyAgent code

python

import eeimport numpy as np
ee.Initialize()
# Port of Singapore anchorage areaport_aoi = ee.Geometry.Rectangle([103.6, 1.1, 104.1, 1.5])berth_aoi = ee.Geometry.Rectangle([103.75, 1.25, 104.0, 1.42])
def count_vessels(aoi, start, end):    """Count vessel-like bright objects in water from Sentinel-2."""    s2 = (ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')          .filterBounds(aoi)          .filterDate(start, end)          .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 10))          .sort('system:time_start', False)          .first())
    b4 = s2.select('B4')    b8 = s2.select('B8')    b3 = s2.select('B3')
    # Water mask: NDWI > 0    ndwi = s2.normalizedDifference(['B3','B8'])    water = ndwi.gt(0.05)
    # Vessels: bright in visible bands but not vegetation    vessel_candidates = b4.gt(1800).And(b3.gt(1600)).And(water)
    # Morphological opening: remove small noise, keep vessel-size objects    vessel_clean = vessel_candidates.focal_mode(2, 'circle').focal_max(2, 'circle')
    count = vessel_clean.connectedPixelCount(256).gt(4)    vessel_count = count.reduceRegion(        ee.Reducer.sum(), aoi, 10, maxPixels=1e9    ).get('B4').getInfo()    return int((vessel_count or 0) / 6)  # ~6 pixels per average vessel at 10m
weeks = [    ('2026-03-01','2026-03-07','Wk1'),    ('2026-03-08','2026-03-14','Wk2'),    ('2026-03-15','2026-03-21','Wk3'),    ('2026-03-22','2026-03-28','Wk4'),]
print(f"{'Week':<6} {'Anchorage':<12} {'Berth Zone':<12} {'Utilization'}")for s, e, label in weeks:    anc = count_vessels(port_aoi, s, e)    bth = count_vessels(berth_aoi, s, e)    util = min(bth / 42 * 100, 100)  # 42 berths at Tanjong Pagar    print(f"{label:<6} {anc:<12} {bth:<12} {util:.1f}%")

output/singapore_port_report.mdOutput report