Outline to Use SLFMI Raster to Assess Live Fuel Moisture in Shrubs
(Elevated and Extreme Fire Danger Rasters are Described in Step 4)
1) Have GIS software installed on computer. You can download free QGIS (Version 3) from this website here.
2) Download compressed folders below. Each folder include one SLFMI raster, as well as the corresponding Sentinel-2 SLC (scene classification) and TCI (true color imager) layer.
Sentinel-2 SLC Classification
(raster will be classified as grayscale when opening)
3) For accurate assessment of shrub pixels, analyze SLFMI pixels values for the same pixels over time along with high resolution background imagery (Google and Bing imagery can both be loaded for free as continuous background imagery in free QGIS). All raster have 20-meter resolution; as a result, there is likely going to be non-shrub values averaged into pixel values, which is a reason why a comparison of the same pixels over time needs to be made. It is important to know that the pixels have SLFMI values; shrub values range from about 2.0 to 8.0 (values also depend on type and density of vegetation, as well as the amount and color of soil that may visible). Pixel values much outside of the 2.0 to 8.0 range include other vegetation, shaded areas, or other surfaces. Avoid riparian vegetation and riparian areas in general, as well as any type of tree or tree shadow. Also, Sentinel-2 pixels may have about half a pixel horizontal positioning error, so keep at least a one pixel boundary between pixel you use and non-shrub surfaces. Also, you need to make sure to avoid a recently burned area, the imagery may not be recent enough to show a recent burned area.
Pixels analyzed need to be in sunlight at the time of Sentinel-2 imaging all season long so change over time for the same pixels can be assessed. Use both the TCI and SLC layer to check for clouds and shadows (clouds are not visible enough by just viewing the TCI). You can decipher shadows, for objects such as trees and clouds, by viewing the TCI and just using common sense. It is a good idea to avoid a surface within about 4 kilometers of a cloud, even if the SLC file shows the surface being clear.
In mountainous areas in particular (because of significantly changing and abrupt topography), sunlit surfaces are in large part a function of solar azimuth and solar elevation. It is more difficult than it would seem to distinguish sunlit vegetation in mountains from vegetation in shadow by solely viewing a satellite image. For areas in California, you can download a corresponding hillshade tile from this webiste to help locate surfaces in sunlight. Hillshade tiles that can be downloaded on this website have been designed for the corresponding SLFMI imagery - they are based on the average Sentinel-2 solar azimuth and the lowest Sentinel-2 solar elevation from the beginning of spring to the beginning of fall. California elevation rasters can be downloaded from this website.
Below, is an example of 20-meter pixel boundaries (orange) that represent sunlit shrub areas from spring to fall with Google high resolution imagery. Try to use pixels with the least amount of soil you can, though it may not be possible to find imagery of shrubs in a particular areas without some soil visible.
4) A way to use the data in QGIS (and similarly in other GIS software) to analyze and compare SLFMI values, is to have the high resolution imagery on the top and checked in the Layers window (so it is the visible layer). If you then click the Identify Features tool, whatever layer that is shaded in the Layers window (click on layer name to shade it), will have the value appear in the Identify Results window. You can have many SLFMI rasters in the Layers window under the visible high resolution imagery, then click on the same shrub area but have different SLFMI layers shaded to compare fuel moisture change over time.
SLFMI pixel values are in their native floating point format ranging from about 2.0 to 8.0. These values can be correlated to fuel moisture (FM) values for a particular area, but not for the whole scene. This is so because different pixels with vegetation that has the same fuel moisture can have different reflectance values (because the vegetation type, density, as well as amount and color soil visible can vary). For actual fuel moisture amounts to be mapped for a particular area, a predictive model would need to be made for a particular area, then an average or median SLFMI values could be used to predict a fuel moisture value there.
However, there is a way to estimate if fuel moisture has decreased to dangerous levels, or is within danger levels. To do this, determine dates that you know fuel moisture corresponds to elevated and extreme fire danger levels. Contact GIS Ag Maps and we will upload the Sentinel-2-based SLFMI raster to the website for you to download. For example, for Chamise Chaparral the elevated and extreme live fuel moisture levels can be interpreted as 100 and 77 percent, respectively (Dennison et al, 2008; PDF on this website).
Figure 2 (Dennison, et al. 2008)
In the United States, a date for a particular fuel moisture level can be determined or closely estimated from the following website: National Fuel Moisture Database (offsite page; open in new tab); elevated or extreme levels from the website can be use as baseline values. Simply compare raster values with the Identify Features tool, or subtract a baseline SLFMI raster from a SLFMI raster with the Raster Calculator, and if the new raster shows negative pixels values, values are lower than the baseline values and have decreases to elevated or extreme danger. An example of Elevated and Extreme Fire Danger rasters are included in Tile 18 downloads (keep in mind though that within a Sentinel-2 scene, FM values can vary; so an Elevated or Extreme Danger Level in one area of the imagery may not be in another).
Reference
Dennison P.E., Moritz M.A., Taylor R.S. 2008. Evaluating predictive models of critical live fuel moisture in the Santa Monica Mountains, California. International Journal of Wildland Fire 17, 18–27.