The algorithm used for TSU creation has been recently updated to better catch big area of low and high vegetation (both were mixed previously) and also to account for rail linear (https://github.com/orbisgis/geoclimate/pull/822).
There will also be soon modifications to identify roads as unique TSU (the main line was previously used for segmentation).
These modifications might have an impact on the building height estimation. Thus it seems necessary to update (and it might be a good time also to improve - cf #770) the random forest model.
The algorithm used for TSU creation has been recently updated to better catch big area of low and high vegetation (both were mixed previously) and also to account for rail linear (https://github.com/orbisgis/geoclimate/pull/822).
There will also be soon modifications to identify roads as unique TSU (the main line was previously used for segmentation).
These modifications might have an impact on the building height estimation. Thus it seems necessary to update (and it might be a good time also to improve - cf #770) the random forest model.