Getting Clear on Satellite Imagery - Climate TRACE

From Climate TRACE’s inception, using satellites to track human-caused greenhouse gas (GHG) emissions to their sources has been a cornerstone of our technology-centric approach. But why satellites? How do they play a role? And which satellites are we using?

A foundation of Earth observation

Earth observation (EO)is the gathering of information about the Earth’s physical, chemical, and biological systems” via remote sensing technologies. Remote sensing detects and monitors things from a distance by observing reflected and emitted radiation. It can include sensors mounted to aircraft or even to land-based objects, but satellites are especially powerful — they provide global coverage with an all-encompassing view that gives emissions sources no place to hide.

The launch of the modern era of satellite-based Earth observation

Satellites play a central role in Climate TRACE (and other forms of EO) thanks to the recent evolution and proliferation of their technology. It took five decades for the world to see 1,000 active satellites in orbit. Largely within the past decade, that number has rocketed to more than 4,500 active satellites as of earlier this year, including more than 1,000 new satellites launched in 2020 alone.

Moreover, the rise of cloud computing and freely available software and data processing have put the images from those satellites’ sensors at researchers’ proverbial fingertips. Today we’re able to “just grab the data” like never before for ingestion into the AI/ML algorithms of coalitions like Climate TRACE.

Beyond that, at least five additional major variables have evolved to the point that initiatives like Climate TRACE become possible: 1) What the satellites can see. 2) How well they see it. 3) How often they see it. 4) What it costs to deploy the system. And 5) Who maintains the system. Let’s take a closer look at each.

1. WHAT satellites see (spectral resolution)

The overwhelming majority of satellites with sensors aren’t directly sampling the atmosphere. Instead, they have sensors (think “fancy cameras”) pointed toward Earth that see through the atmosphere and sample the Earth’s land surface — and translate that into images that AI/ML algorithms can process to identify GHG emissions-causing assets and activities happening there.

Satellite sensors “see” across various wavelengths of the spectrum, including but not limited to ultraviolet, visible, and infrared. Often, they observe multiple wavelengths simultaneously, which is known as multispectral resolution. For example, Climate TRACE uses data from multispectral sensors mounted on satellites such as Landsat-8 & 9, Sentinel-2A/B, Terra, Aqua, Suomi NPP, and PlanetScope to identify and measure features and activities related to GHG emissions.

The European Space Agency’s Sentinel-2 has multispectral resolution spanning 13 wavelength bands at three different spatial resolutions. Image: ESA

2. HOW WELL satellites see (spatial resolution)

What satellites see is one thing, but how well they see it is quite another. The latter issue is largely about spatial resolution, or how much detail a satellite’s sensor can see within a given area of the Earth’s surface. Think of it like the number of megapixels the camera on your smartphone can capture when you take a picture. Lower megapixels result in lower-resolution images that can appear pixelated and quickly lose important details when you zoom in. Higher megapixels generate much larger file sizes but include much finer details and allow you to zoom in and “inspect” specific regions of the broader image.

For satellites, this comes across as different levels of spatial resolution. In some cases, coarse spatial resolution might be 300 meters (or greater) per pixel. Low resolution is ~60 meters per pixel. Medium resolution has more detail, resolving the Earth at 10 to 30 meters per pixel. And high resolution can “see” the Earth’s surface at 30 centimeters to 5 meters per pixel! Here at Climate TRACE, we typically use imagery with spatial resolution in the range of 3 to 500 meters to identify power plant plumes, sludge ponds at animal feedlots, landfills, crop and forest areas, and even heat-intensive units at steel mills.

Landsat 8 images of Reykjavik, Iceland, at three different spatial resolutions. Image: NASA Earth Observatory

3. HOW OFTEN satellites see (temporal resolution)

Only a small fraction of active satellites are geostationary, meaning they remain in a fixed position in the sky over a particular point on the Earth. The rest orbit the Earth at various altitudes, angles, and speeds… which brings us to temporal resolution, or how often satellites “see” a point on the Earth’s surface. Temporal resolution usually refers to a satellite’s revisit or frequency cycle — the time it takes for the satellite to complete its orbit and observe the same location again.

Some satellites revisit the same spot every 1 to 2 days. Others like Landsat have a 16-day revisit cycle. Longer revisit cycles create the challenge of what happened at that location in between satellite visits. For most Climate TRACE work, we optimize various satellites with different temporal resolutions in order to identify the activity causing GHG emissions.

(Note: because satellite-mounted cameras can adjust their direction, it’s often possible to image an area of the Earth’s surface more often than the satellite’s revisit cycle, by re-pointing the camera toward the desired spot even when the satellite isn’t over the same location in orbit.)

4. What are COSTS to deploy a satellite network

Falling costs to design, build, and launch satellites has fueled a thriving satellite industry in the last ten years and the recent massive proliferation of satellites in orbit. How?

For one, government agencies and international collaborations remain the tip of the spear for cutting-edge satellite technology and science, but the satellite industry has also democratized as private-sector commercial players have added diversity, innovation, and competition to a sector previously dominated by a few government programs. As a result, the costs to launch a satellite have fallen dramatically from $8,000–$30,000 per pound of payload to about $1,200 per pound.

For another, the satellites themselves have gotten smaller. A new class of CubeSats (also known as nanosatellites) measure just 10 centimeters square! Naturally, smaller satellites weigh less and thus are even cheaper to launch into orbit.

5. WHO maintains the system

As profound as how many satellites are now in orbit is also who is launching them: satellites are no longer the sole purview of the public sector (i.e., governments) and the military. The commercial/private sector and even some organizations from civil society and multinational scientific collaborations are launching satellites, too. (For the record: No, Climate TRACE does not launch our own satellites. We have a global network of satellite imagery data partners that contribute to the coalition’s work.)

Climate TRACE uses a combination of public and private satellite data from different sources. This includes freely available data from the European Space Agency (Sentinel-1 and -2 missions), NASA, and the U.S. Geological Survey (Landsat, Terra and Aqua and Suomi NPP missions). Other satellite data includes commercial data from Planet Labs.

What satellites (and associated data) is Climate TRACE not using?

Finally, while there are some satellite sensors that measure GHGs directly, such as Sentinel-5P TROPOspheric Monitoring Instrument (TROPOMI) and Orbiting Carbon Observatory-2 and -3 (OCO-2 and -3), and GOSAT-1 and 2 (Greenhouse gases Observing SATellite). These sensors cannot attribute GHG concentrations to specific sources, such as a power plant or a rice paddy field. For this reason, we use the satellites described in section 1 above to develop proxy and secondary measurements related to sector-specific emissions.

Connecting the dots… and the satellite imagery

Don’t be fooled by sci-fi TV shows and movies that show satellites zooming in to infinite levels of fine-detailed, live-streamed views in real-time. That type of imagery of the entire Earth would require immense amounts of transmission bandwidth, processing power, and data storage.

In practice, Climate TRACE researchers are constantly navigating a balancing act of spectral vs. spatial vs. temporal resolution—choosing the right combination of specs for the task at hand. For example, the VIIRS sensor mounted on Suomi NPP has relatively coarse spatial resolutions (1 kilometer, 750 meters, and 375 meters) but its spectral coverage spanning the visible and infrared wavelengths support the efficient detection of forest fires across large swaths of land area.

By stitching together complementary imagery from different satellites and cross-referencing additional non-satellite data in order to improve GHG emissions estimates, we’re able to create a more-continuous time series of imagery, fill in gaps, overcome challenges (e.g., what to do when clouds obscure a location), and paint a fuller picture of tracking human-caused GHG emissions.

Ann Marie Gardner is an editor, journalist, and Climate TRACE coalition member. The image at the top of the article is a stylized composite of this image and this image, both from NASA.