Marcello Schiavina has just finalized a tif file with UN country boundaries. However, approximately 1600 grid cells are populated by the model, but outside UN country boundaries. This only affects the modelled results, not the input data. In addition, Marcello notes "about top left.. I saw your layers slightly shifted with respect to GHSL grids". I have asked him to confirm whether how much layers are shifted.
With unchanged country boundaries, we'd lose approximately 93,000 people and 8.5km2 of built-up land (worldwide) from our modelling results. This is not critical (in my opinion) but something that is better addressed anyway.
Locations of populated grid cell outside the country boundaries:
Zoomed in (red dot is populated but outside; brown is hasland; blue is coastline shapefile):
Running hypothesis is that GeoDMS executes poly2grid() on the topleft corner, while the GHSL team's approach takes the centre. Conceptually I think that the centre approach probably yields a more likely representation of the grid cell characteristics than the top left corner.
As a thought, is the easiest way not to move the grids and subgrids in domain and sub_domain to the right and down by half the grid cell size? If I am not mistaken this would approximate the centre check done in the GHSL and ensire consistency.
Marcello Schiavina has just finalized a tif file with UN country boundaries. However, approximately 1600 grid cells are populated by the model, but outside UN country boundaries. This only affects the modelled results, not the input data. In addition, Marcello notes "about top left.. I saw your layers slightly shifted with respect to GHSL grids". I have asked him to confirm whether how much layers are shifted.
With unchanged country boundaries, we'd lose approximately 93,000 people and 8.5km2 of built-up land (worldwide) from our modelling results. This is not critical (in my opinion) but something that is better addressed anyway.
Locations of populated grid cell outside the country boundaries:
Zoomed in (red dot is populated but outside; brown is hasland; blue is coastline shapefile):
Running hypothesis is that GeoDMS executes poly2grid() on the topleft corner, while the GHSL team's approach takes the centre. Conceptually I think that the centre approach probably yields a more likely representation of the grid cell characteristics than the top left corner.
As a thought, is the easiest way not to move the grids and subgrids in domain and sub_domain to the right and down by half the grid cell size? If I am not mistaken this would approximate the centre check done in the GHSL and ensire consistency.