Satellite data integration for advanced analysis

As part of the Using Satellite Data for Analysis in Starboard series, this article outlines how Starboard integrates various satellite acquisitions across sensors, time, and geography, and how you can use the full range of available metadata for advanced analysis.

Different satellite sensing technologies offer complementary strengths, making their combined use essential for reliable maritime intelligence. This multi-sensor approach enables cross-validation, but managing such a complex layered strategy becomes increasingly challenging when multiple data feeds are ingested for the same focus area within a short time frame.

A single vessel can have several satellites targeting it at once

Starboard addresses this challenge by linking different data sources into visual and chronological timelines that update dynamically based on the user-selected time range and area of interest.

There are two main ways that analysts can sort through acquisition timelines in Starboard, either by Area of Interest, or by Vessel of Interst.

Area of Interest: Analysts can quickly click through all of the available acquisitions for an Area of Interest in the Added satellite data panel. This enables a seamless transition between available acquisitions along the selected timeline, while also preserving the underlying correlations with AIS data.

Starboard displaying 12 different sources of satellite collections over an area of interest on a single day.

Vessel of Interest: Each satellite detection that is correlated with an AIS vessel is added to that vessel’s track history and shown as a clickable icon along the selected AIS track on the map. Clicking an icon (either on the map or from the track history) opens the corresponding detection and displays the associated satellite footprint for contextual analysis. There are different icons for each of the three main satellite types:

A vessel detected in high resolution imagery has two previous detections from different satellites on the same day, highlighted in yellow boxes along its selected track.

While the growing number of commercial satellites in Low Earth Orbit (LEO) is reducing tasking delays and improving revisit rates, processing delays remain a significant bottleneck for effective operational use. Depending on the provider’s infrastructure, time-zone, and support capacity, total end-to-end latency before data becomes available to Starboard can range from hours to days.

Starboard’s cloud-based infrastructure minimises processing times and costs on our end, allowing us to provide the full range of satellite data by default rather than limiting delivery to image chips. This concept has already been introduced in the previous article of this series, and is expanded here to cover advanced RF metadata analysis and imagery case studies using Starboard, while highlighting the critical role of metadata integration in minimising the risk of erroneous intelligence reporting.

MDA Platforms which are trying to scale back on processing costs will often withhold crucial RF metadata from the end user such as signal properties, probability ellipses, and satellite footprint extents.

For signals intelligence practitioners, these datasets are integral to analytical accuracy and usability. For this reason, Starboard integrates a comprehensive range of the RF metadata to support advanced analysis.

An uncorrelated VHF detection is represented by a red probability ellipse. The associated RF signal properties are displayed in the selected detection panel on the left.

In this example, a VHF radio detection has an ellipse of mathematical probability with a major axis of 85 nm. This means that the detected vessel could feasibly be anywhere within the red probability ellipse, which covers a total calculated area of around 1,000 nm². As the ellipse overlaps with an EEZ boundary, more advanced analytical techniques and cross-validation strategies are required to determine if this detection is of interest to the relevant authorities.

This could include comparing the frequency of the VHF signal to known maritime radio channels, correlating the location with non-AIS data sources, or assessing nearby AIS-off events for potential candidates.

There are several factors which contribute to the size and shape of each probability ellipse. The following is a non-exhaustive list of factors that can influence the accuracy of each RF geo-location:

Some MDA platforms display only a centre-point estimate rather than the full probability ellipse, and omit signal properties altogether. Without access to the complete metadata, this can lead to misclassified detections, incorrect vessel correlations, and flawed intelligence reporting.

Subsequent articles in this series will continue to explore additional case studies for analysing RF data in Starboard.

Earth observation satellites can capture imagery spanning thousands of miles of ocean in a single frame. Processing this data is challenging, as it involves managing downlink limitations, storage capacity, and display methods, all within operational timeframes suitable for intelligence use.

To reduce costs and processing demands, many MDA platforms deliver only an image chip to the end user, which is a small cropped section of the full image containing just the detected vessel. This limits the analyst’s ability to assess the broader scene and most importantly – to identify false detections. This is particularly relevant for low resolution imagery, where detection algorithms are less reliable. Consider the following Sentinel-2 image detections against the background AIS.

This Sentinel-2 image contains numerous false negatives (AIS signals without detection boxes), requiring manual inspection of the wider image.

Subsequent articles in this series will continue to explore additional case studies for analysing SAR and EO data in Starboard.