Port electrification and cold-ironing uptake proxied by nighttime emissions
Sentinel-5P TROPOMI measures SO2 and NO2 columns over port areas at 3.5 km pixel resolution. Cross-referenced with AIS dwell times, falling per-vessel emissions during port calls proxy cold-ironing uptake, with honest caveats about fuel-switching ambiguity.
Sensors
- Sentinel-5P TROPOMI: Measures tropospheric SO2 and NO2 column densities at 3.5 × 5.5 km pixel resolution (improved from the original 7 × 3.5 km after August 2019 upgrade). Daily global coverage. Detection of shipping-related SO2 plumes has been demonstrated over busy sea lanes and port clusters in peer-reviewed literature, though individual small ports may fall below the noise floor under light traffic.
- Spire Global AIS: Spaceborne AIS from Spire's LEMUR-2 constellation provides near-global vessel position and identity data, including berth arrival and departure timestamps. Combined with TROPOMI overpasses, it enables per-vessel dwell-time attribution to observed emission columns. Coverage gaps exist in very high latitudes and in regions with deliberate AIS suppression.
- Sentinel-1 SAR: C-band synthetic aperture radar at 5 × 20 m (IW mode) to 20 × 22 m (EW mode) provides cloud-independent vessel detection and berth occupancy counts. Used here to ground-truth AIS-reported vessel presence and flag dark vessels that are physically berthed but not transmitting, ensuring the emission denominator is accurate.
- VIIRS Day/Night Band (Suomi-NPP / NOAA-20): 500 m resolution nighttime light imagery provides a complementary proxy for port activity intensity. Unusual reductions in light output at a berth are not a reliable cold-ironing signal on their own, but persistent changes corroborate emission trends and can flag operational shutdowns that explain anomalous dips in TROPOMI columns.
What auxiliary engines actually emit, and why it shows up from orbit
A large container ship at berth running its auxiliary diesel generators to power hotel loads, refrigerated containers and cargo handling equipment burns fuel continuously. At typical loads, marine diesel engines emit SO2 in proportion to the sulphur content of the fuel and NO2 as a combustion byproduct. Aggregate these across a busy port with dozens of vessels and the signal is substantial enough to appear in TROPOMI's tropospheric column retrievals.
TROPOMI's SO2 product uses the DOAS (differential optical absorption spectroscopy) method applied to UV backscatter in the 312 to 326 nm range. NO2 is retrieved in the visible around 405 to 465 nm. The 3.5 km pixel footprint means a single large terminal, say one occupying a 2 km berth frontage, sits within one or two pixels. That is sufficient for trend analysis across many overpasses but not for attributing emissions to a single vessel on a single day.
The logic of the proxy: dwell time divided into emission column
The analytical approach is straightforward in concept and genuinely difficult in execution. AIS records give the arrival and departure time of each vessel at each berth. TROPOMI passes over once per day, typically in the mid-morning local time. By accumulating many overpass observations across many vessel calls, and normalising observed SO2 or NO2 column anomalies by the number and size of vessels present at the time of the overpass, a per-vessel-call emission rate can be estimated.
If that rate falls over months or years, something has changed. The vessel may be using lower-sulphur fuel, the port may have deployed shore power infrastructure, or the fleet mix may have shifted toward newer, cleaner engines. This is where the method hits its principal limit: TROPOMI cannot distinguish between a ship that has plugged into shore power and one that has simply switched from heavy fuel oil to VLSFO (very low sulphur fuel oil, 0.1% S cap under IMO 2020 rules). Both produce lower SO2. Distinguishing them requires additional data such as port authority records, fuel delivery manifests, or the presence of shore-power metering infrastructure visible in very-high-resolution optical imagery.
The NO2 signal is somewhat more informative on this point. Shore power eliminates onboard combustion entirely, so NO2 should also fall. A vessel on VLSFO still combusts and still emits NO2, just less SO2. A joint SO2-plus-NO2 reduction that tracks the arrival of documented shore-power infrastructure is a stronger, though still indirect, signal of cold ironing.
Noise sources that will frustrate a naive analysis
Port areas are surrounded by industrial activity, road freight, and in many cases, power stations. TROPOMI columns over a port reflect all of these sources simultaneously. Wind direction at overpass time determines whether ship plumes are advected over the pixel containing the berth or carried offshore. A consistent easterly can make a busy port look clean and push the signal over residential areas downwind.
Cloud cover is the bluntest problem. TROPOMI retrievals are flagged invalid under thick cloud, and many of the world's busiest ports, including those in northern Europe and East Asia, experience substantial cloud cover in winter months. A port that electrifies in November may not accumulate enough clear-sky overpasses to confirm the trend until the following spring. Analysts should expect to work with multi-month aggregation windows rather than event-level detection.
Vessel size matters too. A single very large crude carrier at berth dwarfs the emission contribution of several small feeder vessels. Failing to weight by vessel engine power or gross tonnage when building the per-vessel metric will produce a noisy and potentially misleading trend.
What the method can realistically deliver to a port authority or regulator
Used honestly, this approach gives a port authority or national maritime regulator a low-cost, independent check on whether aggregate at-berth emissions are moving in the expected direction following infrastructure investment. It does not replace stack monitoring or fuel sampling. It is useful for screening a portfolio of ports to prioritise inspection resources, or for providing a publicly defensible, satellite-derived trend line to accompany a sustainability report.
For regulators tracking IMO 2020 compliance across a national fleet, the combination of TROPOMI SO2 anomalies and AIS vessel identity provides a ranked list of vessel types and operators associated with persistently elevated port emissions. That list is an audit trigger, not a conviction. The quantitative output is a relative index, not a calibrated mass-flux estimate.
Satellize runs this kind of multi-sensor atmospheric and AIS fusion analysis on open constellations for government clients. The method is the same one applied, at a different sensor scale, to agricultural monitoring work such as the Tonga crop-estimation programme: accumulate many imperfect observations, model the confounders, and report confidence intervals rather than false precision.
Archive depth and what historical baselines are available
Sentinel-5P was launched in October 2017 and TROPOMI data are publicly available from the Copernicus Data Space Ecosystem from mid-2018 onward. That gives roughly six years of archive at the time of writing, spanning the IMO 2020 sulphur cap introduction in January 2020. A pre-2020 versus post-2020 comparison of SO2 columns over a port is therefore feasible with existing data, and provides a natural experiment against which cold-ironing infrastructure rollout can be assessed.
Spaceborne AIS from Spire covers roughly the same period at useful density. Sentinel-1 SAR archive goes back to 2014 for some regions. The combination means a new client does not need to wait for data collection: a baseline analysis can begin immediately using the public archive.
Typical figures
| TROPOMI spatial resolution (SO2, NO2) | 3.5 × 5.5 km per pixel (post-August 2019); earlier data at 7 × 3.5 km |
| TROPOMI revisit | Daily global coverage; one overpass per day per location |
| TROPOMI SO2 detection sensitivity | Approximately 0.5 DU (Dobson units) for individual pixels; ship lane signals typically 1 to 3 DU above background in published studies |
| AIS positional latency (Spire) | Near-real-time to 30-minute delay depending on satellite pass geometry; historical archive available |
| Sentinel-1 SAR resolution | 5 × 20 m (IW mode); 20 × 22 m (EW mode); 6 to 12 day repeat at mid-latitudes |
| Spectral bands used | TROPOMI UV (312 to 326 nm, SO2); visible (405 to 465 nm, NO2); Sentinel-1 C-band (5.405 GHz) |
| Minimum detectable port cluster | Ports with consistently 5 or more large vessels at berth during overpass are typically detectable above background; smaller terminals require multi-month aggregation |
| Archive depth | TROPOMI: mid-2018 to present; Sentinel-1: 2014 to present (region-dependent); Spire AIS: 2016 to present |
| Cloud limitation | TROPOMI retrievals invalid under thick cloud; effective clear-sky sampling rate varies from roughly 30% (northern Europe, winter) to over 80% (arid coastal regions) |
| Delivery formats | NetCDF (TROPOMI L2 products); GeoTIFF aggregations; CSV vessel-call tables; GIS-ready polygon layers |
Analytics Satellize can run
| Per-port SO2 and NO2 emission index | TROPOMI L2 column retrieval aggregated over port polygon, normalised by AIS-derived vessel count at overpass time; multi-month rolling mean | Monthly time-series chart and GIS layer per port; delivered as PDF report and GeoTIFF |
| Vessel-class emission attribution | AIS vessel type and gross tonnage joined to TROPOMI overpass observations; regression of column anomaly against fleet composition at each overpass | Ranked table of vessel types associated with above-average at-berth SO2; CSV feed suitable for regulatory audit prioritisation |
| Cold-ironing uptake trend signal | Joint SO2 and NO2 column trend analysis before and after documented shore-power infrastructure commissioning dates; change-point detection on monthly aggregates | Before-and-after trend report with confidence intervals; flagged as indicative only where fuel-switching cannot be excluded |
| Berth occupancy ground-truth layer | Sentinel-1 SAR vessel detection (CFAR algorithm) fused with AIS to identify dark or misreporting vessels; corrects emission denominator for vessels not appearing in AIS | Berth occupancy raster at 10 m, updated on each Sentinel-1 pass; integrated into emission index calculation |
| Regulatory screening portfolio | Multi-port comparative ranking by per-vessel SO2 index across a defined set of national or regional ports; standardised against vessel size distribution | Quarterly comparative dashboard; ports ranked by emission intensity with trend direction; delivered as interactive HTML or static PDF |
| IMO 2020 compliance signal | Pre- and post-January-2020 SO2 column comparison over port areas, controlling for traffic volume changes using AIS; identifies ports or operators with anomalously slow SO2 reduction | One-off or annual compliance signal report; includes methodology note on ambiguity between fuel-switching and shore-power adoption |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.