Skip to content

Step 4. Compute forest mask

Step 4 : Creating a forest mask, which defines the areas of interest

The harmonic model is adjusted for all pixels, but in our case only coniferous forests are of interest. The export of dieback detection maps requires to filter out irrelevant areas.

This steps aims at computing and writing a binary raster, then used as a 'forest mask'. This can be done either by rasterizing a vector file, or by producing this binary mask directly. The final mask must have the same dimensions, spatial resolution, origin, and projection as Sentinel-2 images. It is also possible to run fordead without any forest mask. Specific options allow taking advantage of land cover or forest spatial database available in France (BDFORET from IGN and Land cover from THEIA).

Here, we used a shapefile identifying forested areas in the example dataset, which was rasterized as a binary raster.

Comprehensive documentation can be found here.

Study area with area of interest Resulting mask
study_area_interest forest_mask
Running this step using a script

Run the following instructions to perform this processing step:

from fordead.steps.step4_compute_forest_mask import compute_forest_mask

                    forest_mask_source = "vector", 
                    vector_path = "<MyWorkingDirectory>/vector/area_interest.shp")
Running this step from the command prompt

This processing step can also be performed from a terminal:

fordead forest_mask -o <output directory> -f vector --vector_path <MyWorkingDirectory>/vector/area_interest.shp

The outputs of this step are stored in data_directory:

  • In the folder ForestMask, the binary raster Forest_Mask.tif with the value 1 for pixels of interest, the value 0 elsewhere.

NOTE : Though this step is presented as the fourth, it can actually be used at any point, even on its own in which case the parameter path_example_raster is needed to give a raster from which to copy the extent, resolution, etc...