Added fix for false classification with duplicate tiles

discretizer/discretize.py

@@ -521,6 +521,11 @@ class Discretizer(Quadtile, Config):
                 columns=["id", "completeness", "geometry"]
             )
 
+        # Drop duplicate tiles from sea- and inland-water, so dissolving the combined tiles
+        # later by ID doesn't falsely sum up classification values
+        seawater_per_tile_gdf = seawater_per_tile_gdf.drop_duplicates(["id"])
+        inlandwater_per_tile_gdf = inlandwater_per_tile_gdf.drop_duplicates(["id"])
+
         # Concatenate inland-water and sea-water tiles
         water_per_tile_gdf = seawater_per_tile_gdf.append(inlandwater_per_tile_gdf)
 
@@ -531,7 +536,6 @@ class Discretizer(Quadtile, Config):
             water_per_tile_gdf = gpd.GeoDataFrame(columns=["id", "geometry", "completeness"])
 
         # Sum up sea-water and inland-water classification and aggregate geometries
-        # _combined_water = water_per_tile_gdf.groupby(["id"]).agg(sum)
         _combined_water = water_per_tile_gdf.dissolve(by="id", aggfunc={"completeness": "sum"})
         _len = len(set(_combined_water.index.tolist()))
         self.log.debug(f"<[{pid}]> Dissolved into {_len} water polygon geometries.")
@@ -646,6 +650,13 @@ class Discretizer(Quadtile, Config):
             water_per_tile_gdf.set_index("id")["completeness"] + self.classification.VALUE_LAND
         )
 
+        # Check that classification is within the constraints of the bitmask
+        if not all(0 <= t <= 7 for t in tile_polygons_gdf["completeness"].tolist()):
+            _gdf = tile_polygons_gdf[tile_polygons_gdf["completeness"].between(0, 7)]
+            raise ValueError(
+                f"<[{pid}]> Coast-tile classification has to be between `000` and `111` (7). {_gdf}"
+            )
+
         # Add water-polygon column to the tiles data set
         # We call the water-polygon field `water_geom`; bear in mind the land-polygon field
         # is called `geometry`