The hedge fund industry has always sought information advantages over competitors. For decades, this meant better models, faster data feeds, and more skilled analysts. Today, it also means looking down from orbit. Commercial satellite constellations operated by companies like Planet Labs, Maxar, and Satellogic can image any location on Earth's surface multiple times per day at sub-metre resolution. Combined with computer vision algorithms, this creates a stream of economically meaningful signals unavailable from any traditional data source.
Oil Storage and Crude Inventories
One of the most widely cited applications of satellite imagery analysis in finance is estimating crude oil inventory levels. Above-ground oil storage tanks have floating roofs that move up and down with the level of oil inside. The shadow cast by the roof edge changes length with the fill level — a signal that is directly visible in high-resolution satellite images.
Computer vision algorithms trained on thousands of labelled tank images can measure shadow lengths automatically, estimating fill levels and aggregating them across the hundreds of major storage facilities globally. This produces near-real-time estimates of global crude oil inventories — data that official sources like the US Energy Information Administration (EIA) publish only weekly, with a multi-week reporting lag. Traders who observe filling or draining trends in real time have a meaningful information advantage over those waiting for official reports.
Retail Activity and Consumer Spending
Car park monitoring is another productive application. Before companies report quarterly earnings, investors try to estimate sales figures by analysing proxy measures of activity. Satellite images of retail store car parks can be used to count vehicles — a reasonable proxy for customer traffic. Object detection models (typically variants of YOLO or Faster R-CNN) identify and count vehicles in each image, producing time series of traffic indices for individual stores and aggregated chains.
Studies comparing car park traffic indices with subsequent reported same-store sales have found statistically significant correlations — suggesting that satellite-derived traffic data genuinely predicts earnings surprises. This signal is particularly valuable for large-format retailers (discount stores, home improvement centres, electronics chains) with large, clearly demarcated car parks.
Agricultural Commodity Analysis
Commodity markets — wheat, corn, soybeans, cotton — are heavily influenced by crop yield expectations. Traditional crop forecasting relies on meteorological models and ground-level surveys, both of which have significant delays and uncertainties. Multispectral satellite imagery (which includes wavelengths beyond the visible spectrum) enables direct measurement of vegetation health through indices such as NDVI (Normalised Difference Vegetation Index). NDVI values, computed from red and near-infrared reflectance, are highly correlated with crop biomass and health.
CNN models trained on paired satellite and yield data learn to estimate crop yields directly from imagery, producing county-level or field-level yield forecasts updated at each satellite pass. Aggregated across major producing regions, these forecasts provide early warning of supply shocks that may move commodity prices. Trading signals derived from satellite-based crop monitoring have been particularly valuable during growing seasons with unusual weather patterns — droughts, floods, frosts — where traditional forecasting methods struggle.
Shipping and Trade Flows
Monitoring global shipping activity provides insights into trade flows, supply chain health, and demand for specific commodities. Satellite-tracked vessel positions (from AIS — Automatic Identification System — combined with optical imagery) can identify vessels docked at specific ports, loading or unloading cargo. Counting ships at Chinese ports, for instance, provides real-time estimates of import and export activity — a leading indicator for manufacturing data and trade statistics that are officially reported with multi-month lags.
Challenges: Data Volume and Legal Considerations
Processing petabytes of raw satellite imagery requires substantial cloud computing infrastructure and specialised ML pipelines. Change detection — identifying meaningful differences between images of the same location at different times — requires robust algorithms that can handle varying illumination, weather, and sensor characteristics. Data licensing and legal considerations also matter: some imagery products have restrictions on commercial use in trading applications, and firms must carefully review licensing agreements before deploying satellite-based signals in production strategies.