Tuesday, June 28, 2016

Tools for Teaching and Research Using Satellite Imagery

Google Earth Engine interface (source)
I've been processing satellite data for a long time and I've used lots of different tools to do it. I've used various digital image processing (DIP) software products (Idrisi is closest to my heart, but also ENVI and ERDAS), and a few programming languages (IDL, Python, R). It can be a challenge to keep up with all of the new tools that come online, and to decide which one to use when I'm developing a new study. It's especially hard to decide which ones to teach to my students, when everything can shift so rapidly.

When I teach Intro to Remote Sensing, I start the semester with Idrisi, which is a great teaching tool, especially for those with no programming experience. Then we move on to R, which is good for the statistical tools, and broad community support. But I rarely get beyond some examples in Idrisi, ERDAS, and ENVI, and a bit of R, before the semester is over. More and more, though, I think our teaching methods have to evolve from graphic user interfaces and commercial DIP products.

I'm really excited by the excellent Python teaching tools for working with remotely sensed data developed by P. Lewis at University College London. Leicester graduates tell me that Python is in high demand with employers. Indeed, it's what I tend to use the most these days in my own research, with gdal and numpy/scipy providing most of the geographic and array tools, and pandas for data analysis. It can be somewhat intimidating if you're new to programming, but let's be honest about what students are getting into here. There aren't many business that would hire someone to click a button 1000 times.

And now, with Google Earth Engine, there's JavaScript to add to the mix. Of course, after a while, picking up new programming languages becomes a lot easier. It might take me a few days or a week to create something, but it's really powerful and quite fast. I probably won't be teaching undergrads to code in Earth Engine any time soon*, but our PhDs and post-docs are becoming our resident experts, and teaching each other (and me!) as they go. Come to think of it, this social process is how I've learned most of the tools that I regularly use to analyse satellite data.

There are some great tutorials out there already for using Google Earth Engine, and if you're looking for more, you might start with these ones by Michael A. Menarguez at the University of Oklahoma. The best part about Google Earth Engine, I think, is the ability to share algorithms quickly and easily (also without having to download all the data). For example, check out this visualization of the Hansen forest loss data from Global Forest Watch. I'll be sure to post some of my own code here related to fire disturbance recovery cycles!

*Like anything, my enthusiasm for Google Earth Engine comes with a few caveats. It belongs to Google, and it might also disappear (like Google Maps Engine and my beloved Google Reader!). They have tons of data, but unfortunately not the raw MODIS imagery, which I use a lot.

No comments:

Post a Comment