BUG: Small fires that get grouped with other fires may not show up at all

[jordan-evens]

It's probably that there's a tiny polygon that doesn't get rasterized to a whole cell?
The root of the problem is that it converts the whole group into a raster and somehow the points aren't showing up
Looks like it just doesn't convert the tiny fire polygon into a cell. It might be doing cell majority/centroid or something instead of it just touching a cell

Probably want to burn individual start points into the raster as well as the rasterized polygons

[Den-Boychuk]

This symptom seems related to the rasterizing algorithm issue. NIP014 has been 2 ha for 3 days but is rasterized as 1 ha. (The pop up is not hiding another raster.)

NIP014 is part of a group. (In the map below, NIP014's raster is hidden by the flame symbol.)

[jordan-evens]

Since all the polygons are in the group, and the small ones just aren't getting burned in, I think the easiest way to accomplish this without changing how FS runs or most of the python code works in to break multipolygons apart in Rasterize() and then burn the centroid of each on into the raster (if it lands inside the polygon).

[jordan-evens]

Actually it seems like we could just add an option:

        # HACK: any cell that's touched should be burned so small polygons still do something
        gdal.RasterizeLayer(output, [1], lyr, burn_values=[burnVal], options=["ALL_TOUCHED=TRUE"])

So fires might get burned in a little bigger than they are but any cell that's touched by the polygons would get included in the rasrter. It's a lot simpler than what I was thinking but would over-estimate area. Going to try that and we can do the other thing if it's too much.

[jordan-evens]

Old behaviour

ALL_TOUCHED=TRUE

So definitely quite a bit more gets burned into the raster. I think a fire being missing is worse than it being burned in a little bit larger, but is it worth making it more complicated to try to do something else or is the larger perimeter fine?

[jordan-evens]

Seems like it might be easier with rio.features.rasterize()

@infantej2 Can you take a look at something like

--- a/src/py/firestarr/gis.py
+++ b/src/py/firestarr/gis.py
@@ -288,6 +288,13 @@ def Rasterize(file_lyr, raster, reference, datatype=gdal.GDT_Byte, creation_opti
         band = output.GetRasterBand(1)
         band.SetNoDataValue(0)
         gdal.RasterizeLayer(output, [1], lyr, burn_values=[burnVal])
+        # HACK: burn centroids in separately so small fires don't get missed when grouped
+        df_lyr = gdf_from_file(file_lyr)
+        df_explode = df_lyr.explode().to_crs(pyproj.CRS(ref_raster.GetProjectionRef()))
+        for p in df_explode.centroid:
+            # ensure point is inside polygon before burning
+            if df_explode.contains(p).any():
+                print(p)
+                # BURN INTO RASTER HERE
         # Close datasets
         del band
         del output

using:
https://rasterio.readthedocs.io/en/stable/api/rasterio.features.html

?