Saturday, June 11, 2016

SNOW??? SNOW!!!

edit: I posted this the morning of 6/11, but by early afternoon, I had already made some tweaks. So, I'll add some minor edits to the original post below...

SNOW!!!

Inside voice, Adam.

In recent weeks, I had spent some time mucking around with Nasa's MODIS satellite imagery to try to deduce the size and extent of the Sierra snowpack. This is exactly the kind of problem that piques my interest. Usually, when I encounter these kinds of problems, I throw a few hours at it, do some really interesting stuff, but leave it incomplete and undocumented, because I saw something else shiny.

Well, this time I pledged to tug it a little further along, so I have. In typical Sierra Mapper style, this feature has been thrown in with abandon, mashed into a hole in existing code.

I'm still not sure what you're even talking about.

You'll see a new button at the bottom of the calculated route page, with a cryptic and exciting label of Snow?!? on it:

Like any good GUI designer, I placed the button for the exciting new feature buried at the bottom of the laundry-list of existing buttons

When you click on the Snow?!? button, you'll be whisked off to another screen. The screen is pretty plain, but shows three elevation profiles for your route. In atypical Sierra Mapper fashion, the elevation profiles are not very edit: somewhat fancy.

The red colored highlighted sections on the profiles denote sections that I declare are snow-covered. That declaration is rooted in image processing and threshold analysis of MODIS satellite imagery, geo-referencing of the resulting snow-cover array, and interpolation of that snow-cover array along your route. The top profile is from the most recent imagery, the one below it from slightly older imagery, and the one below that from a even older.  (Edit: now there's just one profile that is a rainbow of color and information). The point is to look at snow coverage on the route, and look at the dates, and think wistfully about your trip, convincing yourself that that there will or won't be snow, depending on your preference.

You should see something like this. See how there's more on the route that's red on 5/22 (bottom) than there is on 6/03 (middle) and 6/09 (top)? Neato. The red text indicates th
e fraction of the route that's covered in snow.(Edit: Old!)

Edit: New! Way much more fancier-ish, and easier to interpret to boot.

Caveats
Well, there are quite a few of these. First and foremost, I haven't calibrated this approach with even a modicum of rigor, so I'm not sure how sensitive it is to snow cover. In other words, I don't know if 50% snow cover will trigger my analysis to report an area as snow-covered, or 80% or 10%. Ultimately, it simply has to do with the reflected light back to the satellite and the brightness (and whiteness) of the corresponding pixel in the satellite image. Since that will be affected by tree cover, by time of day, and by color temperature and brightness of the satellite image itself, there are a lot of variables that undoubtedly have an impact. Still, the approach finds that the entire Sierra is covered in January, and that none of it is covered in September, so it certainly works--but again, there is the question of sensitivity.

I hope to write more about the process in the coming weeks, and add both features and some polish. Still, I'm haphazardly throwing this feature up now in order to get it out while it was still relevant--adding this in August of a low-snow year would be a lot less interesting.

Lastly, as usual but especially since this was so hastily put together, please report any issues you encounter!





3 comments:

  1. This is a great way to visualize the data. If groundtruthing or other verification is possible, it could have implications for water management planning in watersheds dependent on Sierra runoff. I work on water policy in LA, specifically on augmenting local supplies through stormwater capture and rainwater harvesting as a way of reducing imports -- and I can see your data visualizations coming in handy in showing why local water must be prioritized over relying on Sierra imports.

    I came across this very recent research and thought you might like to see it: http://onlinelibrary.wiley.com/doi/10.1002/2016GL068520/abstract

    ReplyDelete
  2. Excellent work as usual. Thank you for your work! I understand NASA uses LIDAR to measure the Sierra snowpack. It might be possible to obtain this data to help calibrate the snow pack measurement software you developed.
    - Daniel

    ReplyDelete
  3. This is pretty much the coolest thing ever.

    ReplyDelete