Tuesday, May 13, 2014

New Fire Activity Layer

As the summer fire season approaches, I decided to take another look at integrating current fire activity with CalTopo.  There's a wealth of government supplied fire information out there, and I've long thought about integrating this into a custom source that plays well with CalTopo's other map layers.  When the MODIS layer I'd previously been using went down and stayed offline for several months, it gave me the nudge I needed.

The old MODIS layer has now been replaced by a "Recent Fire Activity" layer which pulls data from several different sources.

The first of these sources is, no surprise, MODIS detections.  The MODIS system consists of two satellite imagers (Aqua and Terra) that take daily multispectral imagery of the planet; the IR band can be used determine surface temperature and, from that, detect wildfires.  Wide swaths of imagery are taken at approximately 10:30AM and 1:30PM local time; while they don't follow timezones exactly, the East Coast is imaged 3 hours before the West Coast.

The Forest Service's Remote Sensing Applications Center processes MODIS images into wildfire hotspots and makes the locations available at http://activefiremaps.fs.fed.us/gisdata.php.  While there are other sources available that have better global coverage or a faster update time, the RSAC dataset has additional metadata I couldn't find elsewhere, like the actual observed surface temperature.

Edit: Shortly after this post, Southern California saw a bout of fire activity, unfortunately giving me an opportunity to put the RSAC data through its paces.  Due to the time lag between satellite passes and their shapefile updates, I've switched to a new NASA Source which still includes temperature observations.

Starting at a wide zoom, these detections appear as multicolored Xs and *s on the map.  I wanted to convey two different attributes, severity and recency, so I needed to vary both the color and symbology.  Government maps use a red->orange->yellow color gradient to indicate the age of MODIS detections, and I decided to follow that for consistency.  Red spots were detected in the last 12 hours, orange in the last 24 and yellow in the last 36.

The asterisks represent hotter, more likely hotspots, while the Xs represent colder, less probable spots.  Zoom in a little further and each spot is marked with its average temperature.  In busy locations this can get a little crowded:

Each MODIS pixel is at least 1km square, sometimes larger, and the surface temperature is an average across the entire pixel.  So a 122 degree spot could be small, intense hotspot or a large, smoldering fire.  As the temperature drops, the certainty that there's a wildfire there drops; based on this and presumably other factors I don't understand, the RSAC assigns a confidence number to each detection.

At high zoom levels, detections are represented as a circle with additional information.  The X and Y dimensions of the MODIS pixel are averaged to get the circle's radius, so the circle shows approximate coverage size.  The detection date - which may be later than when the hotspot actually started - is shown in Mountain Daylight Time (MDT), regardless of actual timezone.  Temperature and confidence are given on the next line, followed by the satellite that made the detection and the center that processed it (RSAC or GSFC, for the Goddard Space Flight Center).

Another data source I'm pulling in is smoke polygons from NOAA's Office of Satellite and Product Operations.  Color (gray->black) represents smoke severity, while crosshatching is used for current smoke and empty polygons show where the smoke was several hours previously.

The final source is large fire perimeters from GeoMAC.  Unlike the previous two products, which are based on satellite data, fire perimeters are manually reported by an incident's management team.  As a result, they seem slow to update and I'm on the hunt for a more reliable source.  In the meantime, I'm marking perimeters with the last date at which they were updated, in addition to the incident name, so that you know how recent/reliable the data is:

Because fire season is just ramping up, I haven't had the opportunity to test this layer as thoroughly as I'd like.  Normally I wouldn't add it to CalTopo until I'd had the chance to vet it more thoroughly, but since the MODIS layer already on the site was still having issues, it made sense to sub this in as quickly as possible.  Expect things to change, and please send me any feedback or bug reports.


  1. Hey.. This might be a little out of context - so Sorry I am posting this here. Can you tell me how I can change units of elevation profile? For example, when I am trying to access the Adirondacks area (link - http://caltopo.com/map.html#ll=44.14651,-73.77525&z=15&b=t) I see everything in meters whereas in other areas I do see elevation in Feet. Any idea why? Thanks

    1. Unfortunately, for whatever reason that's how some of the USGS maps are drawn. There's nothing I can do about it as they're part of the map scan and not some separate layer. However:

      - In areas where USFS maps exist, those should all be in feet. Unfortunately they don't have coverage for this spot.

      - Google's Terrain layer shows the elevation in feet.

      - You can enable the "40' Contour" layer checkbox to get a contour layer in feet. Hard to read on top of the USGS map, but readable on top of aerial imagery.

      - You can get the the cursor elevation to update in realtime (http://caltopo.blogspot.com/2014/04/cursor-elevation.html).

  2. This comment has been removed by the author.

  3. What does % conf. mean?

    Also, what does the circle mean? Is this the extent of actively burning fire, or is it a projection of potential fire?

    1. % conf is the percent confidence their algorithm has that the hotspot is actually a fire. The circle is the size of the hotspot detection - basically it means there is a fire somewhere inside that circle, but not necessary covering the entire circle. See http://caltopo.blogspot.com/2017/09/caltopo-guide-to-wildland-fire.html