We often get questions about satellites: how they work, who operates them, and what data they capture. While we’re always happy to nerd out about spectral bands and temporal resolution over the phone, we thought capturing that information here would be a helpful reference.
Let’s start with the basics: first, we deal with Earth Observation (EO) satellites, which are distinct from other types of satellites like GPS, internet, or telescopes. EO satellites gather data while orbiting the earth. Some of these are public satellites, where the data is made publicly available. These include satellites operated by NASA and the European Space Agency. Others are owned by commercial companies who sell access to that data. This includes companies such as Maxar (formerly DigitalGlobe), Airbus, and Planet, all of whom are Upstream Tech partners.
EO satellites allow us to look back in time using archive data — meaning that we can see how ground conditions or land cover have changed over time. Many public EO satellites were launched several decades ago, so there is a long archive of historic data to explore. Those launched more recently are often able to capture higher resolution data, as well as other novel and useful information about our planet.
We provide satellite data to our clients both as imagery, and also as analyses it to understand natural resource implications. We use computer code to transform the raw data from a satellite into images that convey information about specific ground conditions. Two such transformations that we use often in our work are the Normalized Difference Vegetation Index (NDVI), which shows vegetation vigor, and the Normalized Difference Water Index (NDWI), which shows surface water presence. When created as a timeseries, these allow us to visualize ecological changes on the landscape over time. The image below illustrates a gradient from true color to water (NDWI) to vegetation (NDVI).
As we develop new services, such as Property Monitoring™, a big question for us is determining which EO satellite data to incorporate and how to analyze this data based on the needs of our clients. We weigh three main satellite data characteristics: temporal, spatial, and spectral resolution. First, some quick definitions:
Each EO satellite captures different data along this spectrum and processes it into images. So when we’re partnering with an organization who’s trying to track or understand certain natural resource conditions on the ground, we start by thinking through which satellite(s) will provide the right data to answer that question. For example, thermal (infrared) data is one factor that informs our irrigation intensity analysis, radar can provide insight into ground texture or shallow water depth, and visible light information at high resolutions is well suited for monitoring nuanced changes in landscape conditions.
We are often faced with trade-offs between temporal, spatial, and spectral resolutions. For example, public satellites gather data ~weekly at 10–30 meter resolution, while higher spatial resolution imagery may be captured less often depending on the location on the earth. Due to both accessibility, frequent availability of data, and extendability to new areas we tend to use the public data for machine learning tasks, while reserving the commercial data for applications requiring more visual inspection. If imagery is needed at a specific time, we can also work with a commercial satellite provider to “task” a satellite, or to point the satellite in a particular direction to capture imagery of that site. If we are looking back in time to set baselines and analyze change over time, we also consider satellite launch date, which determines how far back in time its archives go. For example, Landsat 7 from NASA can go back to 2000, while Sentinel 2 from ESA was launched in 2016. With all of these variables in mind, we strategically choose which sources to pull from to provide our clients with the information they need.
The bottom line is this: we’re experts in all of this so you don’t have to be! Reach out to discuss how satellite data and machine learning can help your organization better manage and monitor natural resources on a changing planet.
