Skip to content

Mapping fuel accumulation patterns

We've written a package named sclero to apply fuel accumulation curves to raster data. It's organized under forestobs-new/accumulation. To install it, navigate to that directory and run pip install -e .

The main event is the script sclero-accumulate.py. It produces a 4-band raster file encoding current fuel loads at four vertical strata: litter, litter + near-surface, elevated and bark.

sclero-accumulate.py [-h] -o output_file [--fuel-parameters csv_file] [--max-time years] [--dont-scale]
  [--density] [--from-formations] [--classes raster] [--fcover raster] [--time-since-fire raster]
  [--ladderfuels raster] [--canopycover raster] [--res xres yres] [--bounds xmin ymin xmax ymax]
  [--crs projection] [--dstnodata nodata_val] [--dtype data_type]

Computes current fuel loads (tons) using data on vegetation type, cover, and time since fire.

optional arguments:
  -h, --help            show this help message and exit
  -o output_file        Output file path
  --fuel-parameters csv_file
                        CSV file with fuel accumulation curve parameters
  --max-time years      Time since fire value to assign to unburned locations
  --dont-scale          Don't apply cover-based fuel accumulation scaling
  --density             Calculate tons/ha instead of absolute fuel load per
                        pixel
  --from-formations     Use formation-based accumulation curves. Make sure
                        --classes is pointing to a formations raster
  --classes raster      Path to the class or formation raster (1-band categorical)
  --fcover raster       Path to the fractional cover raster (3+band 0-100
                        continuous)
  --time-since-fire raster
                        Path to the time-since-fire raster (1-band days-since-
                        last-fire)
  --ladderfuels raster  Path to the ladder fuels raster for scaling elevated
                        fuels (1-band continuous)
  --canopycover raster  Path to the canopy cover raster for scaling bark fuels
                        (1-band continuous)
  --res xres yres       Output resolution
  --bounds xmin ymin xmax ymax
                        Output bounding extent
  --crs projection      Output coordinate reference system
  --dstnodata nodata_val
                        Output nodata value
  --dtype data_type     Output numpy data type

File paths

This script reads fuel accumulation curve parameters, stored in sclero/config/accumulation-params.csv, and applies them to maps of vegetation classes and fire history.

GCS paths to a series of default rasters are specified in the sclero config, but these can be specified at runtime with --class-path, --fcover-path, and --time-since-fire-path.

Configuring the output extent

You can specify a series of output raster configurations like the resolution, extent, projection, nodata value & data type.

It may be helpful to only apply the model in a specific area instead of across the full pilot extent, and you'd do this by configuring the --bounds {xmin} {ymin} {xmax} {ymax} option.

Scaling by vegetation density

Accumulation curves are parameterized on a class-by-class basis, so applying them to categorical raster data produces blocky, uniform outputs.

The geographic distributions of fuel loads are typically much more heterogeneous than this method can estimate, however, so we developed a simple method to scale the default fuel density outputs based on vegetation density. We make the following assumptions:

  • Fuel accumulation curves were parameterized with samples from representative field plots in different vegetation formations and classes. These were designed to represent the average conditions for each class.
  • Average != uniform, however, and there is spatial variation in how fuel is spatially distributed across the landscape.
  • Fuel accumulates faster in areas with higher-than-average vegetation cover, and slower in areas with lower-than-average cover.

Our method uses average values of vegetation cover—calculated separately for each class—to tune estimates of fuel density numbers based on the local vegetation density. We do this with our fractional cover raster, which you can specify with --fcover-path {path-to-new-dataset}.

This method adds more spatial texture to the output fuel maps while preserving the landscape-scale fuel densities. It can be turned off at runtime with --dont-scale.

Scaling by area

Accumulation models report fuel density in tons/hectare. Since we map things like vegetation cover at scales finer than 1 hectare, we need to be able to rescale the units of our data based on the size of our pixels.

sclero-accumulate.py converts data from tons/hectare to tons based on the resolution of the output file. So the value encoded in each pixel is the total mass of fuel in that pixel.

If you resample these data, however, make sure you do so with sum resampling. This will aggregate the total fuel loads to the new resolution, which is critical for preserving the per-pixel unit allocation.

Using formations-based accumulation curves

The default script behavior is to assume the --classes raster uses a fuel class map (which includes ~120 distinct classes.) You can instead set --from-formations to average the fuel accumulation parameters from all subclasses within a formation, and apply those fuel curves instead (there are ~17 total formations).

You'll also need to pass the path to the formations raster using --classes {path_to_raster}. By default, if the --from-formations flag is set and--classes is still the default value (as in, not specified by the user), it will automatically use the default formation raster path. The default paths for these datasets are defined in sclero/config/__init__.py.

Back to top