Thank you to anonymous referee #1 for the comments.



Concerning the discussion about WUDAPT and GeoClimate comparison, as the referee assumed, it is not the purpose of this manuscript. The GeoClimate LCZ algorithm has been available for the community since several years so it was needed to:

• describe clearly what was the methodology used to determine the LCZ of a given area

• evaluate what would be the difference in using the worldwide available OSM dataset instead of the French BDTopo one.



Some preliminary works (such as in Blond et al. (2023)) have been achieved to compare GeoClimate to WUDAPT on some French cities but they have not been published yet. The main rough observation is that GeoClimate seems more appropriate than WUDAPT for urban areas but less appropriate for rural areas.



Blond, N., Breton, F., Micolier, A., and Mendez, M.: A modeling approach to address building energy consumption and thermal comfort under urban climate change , EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-14169, https://doi.org/10.5194/egusphere-egu23-14169, 2023.

Thank you to Jan Geletic for his review and comments. As an answer to your general comment, we would like to highlight that our main contribution to the field is having a generic vector-based method to calculate the LCZ and that this algorithm is available by anyone. Then any dataset can be used to calculate the LCZ using this method. The main goal of the manuscript is then to describe the detailled methodology used by the GeoClimate LCZ algorithm. Then the application of the algorithm is illustrated using BDT and OSM datasets. Our manuscript thus falls into the GMD manuscript type called "Model description papers" (https://www.geoscientific-model-development.net/about/manuscript_types.html).

Below are answers to Jan Geletic three major comments.

1. For many European and North America cities, we expect a better identification of urban LCZ types using GeoClimate and OSM than using LCZ generator since the degree of completeness of OSM is rather good for building geometry and building types in these regions and also OK concerning building height (see Bernard et al., 2022). However, LCZ generator is probably more appropriate for rural areas since many small tree vegetation patches or types might be missing in OSM. Moreover, there is still a big lack of OSM data such as in many African, Asian or South American countries. As described in Bocher et al. (2021), GeoClimate algorithms can roughly be distinguished in two steps: first data are imported into a well defined and generic data model; second all indicators and classifications are performed using the data previously included in the generic model. Thus GeoClimate might still be used if local vector dataset exist. However, the first step will still be needed to be performed.

2. As discussed in the manuscript section 2.5, the main expectations we may have about the differences between OSM and BDT data are that:

   • BDT building height is more accurate than the OSM one since for most of the OSM buildings, this information is simply estimated using a random forest model using BDT as real value for the training (cf. Bernard and al., 2022).
   • OSM data has a generally higher land coverage which is mainly due to a better representation of impervious area and vegetation within cities and also a higher building coverage. This information has been verified in our study and is discussed section 3.3. However, the statistics given have been inverted and will be corrected in a next version (37% in BDT against 55% in OSM).

   We have not verified the data to manually defined samples in this study since the comparison of the two datasets was not the objective of this manuscript (thus the datasets are compared relatively to each other).

   What does the reviewer mean by algorithm accuracy ? Accuracy of the LCZ algorithm in general ? This accuracy is not discussed in the article since it always mean that a reference data should be found and that this reference data may also have its shortcomings. As answered to the #Anonymous referee 1, in the near future, we aim at comparing the GeoClimate tool results to manual classifications or remote sensing methods such as LczGenerator. If by accuracy Jan Geletic means the accuracy of the UCP used for LCZ classification, there are limitations which depends on each UCP. For example for the SVF calculation, more information about the accuracy can be found in Bernard et al. (2018) but the algorithm is limited in a way that it only consider flat roofs and no trees. The accuracy of each UCP is then really UCP dependent.
3. As previously discussed, there is indeed almost no OSM data in some part of the world and thus OSM is not applicable there yet. However, as previously explained also, GeoClimate is a two steps approach where the first step concerns data import into the generic GeoClimate data model. Thus some other datasets can be merged during this first step (even though it might mean a considerable amount of work depending on data completeness) and then be used in the presented GeoClimate algorithm (second step of the GeoClimate processing chain). But the first objective of this article is to share the methodology used by GeoClimate to identify the LCZ (so the second GeoClimate step), not to show combination of data that can be used in GeoClimate (first GeoClimate step). However we have illustrated the behavior of the algorithm depending on two different datasets. In the future, a potential work could indeed be to try to merge different datasets to obtain the best land coverage and thus LCZ classification but this has to be done in a future work, this one being the description of the LCZ methodology used by GeoClimate.