The Landsat program is one of the world’s greatest open data success stories. Landsat satellites have been orbiting the earth for decades, creating an irreplaceable archive for studying questions ranging from the retreat of the Aral Sea to water quality in Iowa.
It’s also commercially valuable. Although Landsat data isn’t very high-resolution, it’s used by companies (including Mapbox) for lower-zoomlevel satellite imagery. And because it’s calibrated to scientific levels of accuracy, companies launching a new generation of cheap, flexible satellites can sometimes use Landsat data to calibrate their own imagery. The list of Landsat’s uses goes on and on.
Landsat is such a success, in fact, that it’s is commonly credited as the reason the European Union’s own Sentinel satellites make their data freely available. Sentinel also happens to be similar to Landsat in terms of spectral bands and resolution, so the fact that both are open allows for the two programs’ data to be used synergistically by the world’s governments, researchers, businesses and citizens.
Doing so is surprisingly easy thanks to the ecosystem of redistribution that has sprung up around these data sources. Because of its open licensing, Landsat imagery can be accessed through the public dataset programs of platforms like Amazon Web Services and Google Cloud. This is really important: the Landsat dataset is huge, and loading it requires specialized skills. The fact that it’s available on these computing platforms significantly expands the number of people who can use it.
You can probably guess what’s coming next: Landsat’s openness is being threatened. The program is run by the U.S. Geological Survey (USGS), a part of the Department of the Interior (DOI). DOI leadership has asked USGS to explore charging for Landsat data. They’re running a study, and have also asked the Landsat Advisory Group (LAG) to review the situation. We think that charging for Landsat data is a bad idea, and we recently submitted a comment to the LAG explaining why.
The first thing to note is that this has been tried before without success. The Land Remote Sensing Commercialization Act of 1984 was specifically about monetizing Landsat data. The Land Remote Sensing Policy Act of 1992 was about undoing what had, by then, come to be understood as a mistake. Passing bills through Congress is difficult. Having to reverse them is a straight-up fiasco.
Another privatization push occurred in 2002. Details are scarce — this slide deck says that the identification of a private sector partner “was cancelled as one bidder dropped out and the other bidder did not generate sufficient capital to meet the Government’s terms”. You can track the start of this process in the 2002 Landsat program updates, then observe its total absence from the 2003 updates. Whatever the specifics were, things didn’t work out.
In fairness, DOI, USGS and the LAG are all keenly aware of this history. But there’s reason to think a push to charge for Landsat won’t work this time, either. For one thing, the EU’s Sentinel satellites now exist, offering a workable alternative for many Landsat users. But even more importantly, charging for Landsat data would disrupt today’s thriving ecosystem of use and make working with Landsat data much harder — and not just because the data would cost money. To explain why, we have to dig into some federal government administrative policy. Get ready for fun!
Right now, lots of users — maybe most? — get Landsat data through third parties like the aforementioned AWS or Google public data programs. It’s simpler and it avoids the need for everyone to buy their own truckload of hard drives. How would this be affected by a fee system? The answer might seem simple: charge AWS, Google and similar platforms a lot of money for Landsat access and count on them to pass the costs on to their customers. This is a classic strategy for governments that are trying to sell data. Public bureaucracies generally don’t have sales or marketing teams, so it can make sense to cut a few big deals with deep-pocketed resellers and count on the marketplace to distribute costs from there.
But this doesn’t work under federal data policy. OMB Circular A-130 says that “[a]gencies shall not, unless specifically authorized by statute, establish fees [for public information] that exceed the cost of dissemination to the public”. And Circular A-25 (which is about charging fees) says that “[c]harges will be made to the direct recipient of the special benefit even though all or part of the special benefits may then be passed to others” (in this case the “special benefit” would be access to Landsat data). Put these together and it’s clear: AWS, Google and similar parties could only be charged as much as it costs to get them the data — not the full value that downstream users enjoy. That would limit revenue so severely that monetization would be pointless.
So to make this worthwhile, the government would have to shut down the redistribution ecosystem. Everyone would have to get Landsat data directly from USGS . That would make it much more technically and financially imposing to work with Landsat data. Less good work would be done with it. The program’s value to society would be badly harmed. It’s a bad idea.
Besides which, charging users is complicated. Running the necessary accounting, billing and support systems isn’t exactly trivial. And the government would have to find some way of preventing Landsat data purchasers from simply giving the data away the data they bought. If that 1984 bill is any indication, that would require another act of Congress. And a change like that would impose an entire new set of complications and costs onto users.
It’s a lot of hassle for an uncertain payoff. The LAG itself has previously noted that “demand for landsat imagery is overwhelmingly derived from the research and public government sectors”.
Budgets are tight and it doesn’t hurt to ask questions. But we don’t see much reason to think that, in this case, the third time will be the charm. Landsat is one of the world’s greatest open data success stories, and it should stay that way.