search

orbisgis / geoclimate

Geospatial processing toolbox for environmental and climate studies
GNU Lesser General Public License v3.0
59 stars 16 forks source link

Issues related to "podium" buildings #770

Open j3r3m1 opened 1 year ago

j3r3m1 commented 1 year ago

In many big Asian cities, buildings might be of "podium type" (buildings are stacked). For example, this building is located at ground level and its height is 15 m and this building is right above (its base is located at 15 m and its top at 44 m).

In Hong-Kong, the building which is located on the ground has most of the time its tag "layer" equal to -1. From @ebocher and my understanding this should be set to 0 instead and the building right above to 1. This is a first issue but which is mostly related to OSM data which needs change.

The real issues are related to how to deal with this type of building:

  1. if we do not know the height of each building, how to make sure the algorithm used for building height estimation will be correct (the highest building remains the highest building) ?
  2. if we incorporate stacked buildings, it induces the modifications of all algorithms since most algorithms consider that the buildings starts from ground level.

For the first issue we might create a new building height indicator called "building_height" (?) which for podium building type would be the height of the stacked building from its base (for a normal building it would be from the ground). The RandomForest algo would then be use as dependent variable this "building_height" and we could add the zindex as an independent variable. Thus an extra calculation would be added to calculated "height_roof" for buildings having zindex > 0: it would be to sum their "building_height" to the one of the building below. Thus we make sure that the height from ground level for the upper building is higher than the one of the base building... And this "trick" is also profitable for the second issue: for some of the indicators (such as facade density), the "height_roof" might be replaced by the "building_height". "height_roof" remains the one to use for some other indicators (such as SVF). Some indicators such as "area_weighted_mean_height" within the RSU still needs some modifications since the area of the upper building needs to be removed to the area of the base building.

j3r3m1 commented 1 year ago

@franetibe this might be an interesting improvements to propose for future projects.

franetibe commented 1 year ago

Sure @j3r3m1 ! We may have "podium buildings" in France as well (e.g. Olympiades in Paris).

ebocher commented 1 year ago

You think about this one https://www.openstreetmap.org/relation/3283058#map=17/48.82495/2.36546

franetibe commented 1 year ago

Correct ! Plus other places in the conurbation.