Sharing Remote Sensing Know-How

Two years ago, when we had just set-up the first prototype of Sentinel Hub, I wanted to test the innovative custom band combination option and was immediately presented with a challenge — which bands exactly should I use to create an NDVI or a simple NIR false colour composite? It was easy to skim various scientific papers but “translating” these into something understandable to me, a remote sensing novice, was not an easy task. Later, when I managed to understand the basics, I tried finding some nice examples for our multi-temporal feature and stumbled across papers describing how to monitor burned area. It was somehow possible to match scientific part with our visualization engine, but results were not very good. I concluded that I misread something and that I should probably leave this to remote sensing experts.

Burned area in NE Attica, Greece, mapped using multi-temporal difference of Normalized Burn Ratio

Then questions from our users started coming in. Experienced ones wanted to double-check how we build our EO products. Or which colour scale/legend we are using for NDVI. Others had very similar problems as described above. It seems that the Sentinel-2 program activated a whole new group, non-experts, which is great by itself. Having freely available data available in good enough resolution and short revisit time looked promising for many businesses, especially in precision farming. Sentinel Hub services made it easy to integrate data in their apps, but the interpretation of data is still left to them. Many struggle.

It seems that the whole Earth observation world is changing (and Sentinels certainly have something to do with it). It is now easier than ever to monitor what is happening with the world, both what is happening now or months and years ago. It is possible to test various remote sensing methods globally. But it is still very difficult to get specific information from scientific papers. To replicate the process. Might it be that research is a bit over-protective?

Since our tools do not make much sense without all this knowledge, we have decided to try to make it easier to find it. We came across a great resource of structured basic remote sensing indices —Index DataBase, managed by Institute of Crop Science and Resource Conservation, University of Bonn. This should certainly be a starting point for anyone venturing in EO. We have sponsored INRES to develop REST API so that we can keep our indices up-to-date. I hope that others will be using this service as well so that database will grow.

Then we went one step further. We wanted to make it possible to get even more information about how to get some valuable data from satellite imagery. We have noticed that our users are sharing “scripts” over twitter and in blogs (we call them custom scripts as they represent a custom processing step in Sentinel Hub).

As we see remote sensing processing not unlike software development (even more so in this “new age of RS”, when more and more will be handled by statistical processing and machine learning, rather than visual inspection), we have chosen the most common platform for sharing of open source applications — GitHub — and established Custom scripts repository there (more user friendly view of the same here). We have migrated practically all of our EO products, along with some other popular methods. And we will be adding more as soon as we find some which look promising. GitHub supports versioning, so scripts can evolve through time, an important feature allowing all of us to add even imperfect scripts. We would be delighted if other people, either our users or not, would contribute their knowledge.

Adapting our tools to make the process a bit smoother was easy. It is now possible to pass custom script as either an URL or as base64 encoded parameter. We certainly hope that earth observation training will get even more interactive with these kind of toys.

Let this be our small contribution to the rapidly evolving remote sensing field so that even more people are able to monitor what is happening in their neighbourhood or on the other side of the world, using toolkits and knowledge that was developed by researches in the last few decades. It should also make it easier for our users to understand what is happening with the pixels coming from satellites to their applications behind the stage. And last but not least, we hope that more and more EO research will be easily replicated and tested in other areas and tweaked to fit specific needs.