Dear Dr. Majasalmi,
thank you for submitting your carefully formulated author responses! I think it is very important to address the 4 major points of reviewer 1 in the revised manuscript as you rightly have announced. It is also important to use the latest version of CCI_LC product as you agreed in your response to reviewer 2. In my view it is ok not to add subfigures in Fig. 4 (response to reviewer 2).
I also agree with you that it is important to provide the information where the MS-NFI maps can be requested. Would it be possible to provide them open access, given that you get a data DOI before?
I have a few comments to make:
Response to Reviewer 1:
1. Please also double-check that the revised manuscript text adequately explains this point:
R1: page 6, lines 1-6 if you classify based on V and H, why do you originally cluster with CL and LAI? Especially since you correct the original clusters with V and H and MD?
Your response: "The LUT values represent medoids of the four-dimensional (i.e. V-H-CL-LAI) clusters of the twelve forest classes (3×species group × 4×subgroup). Related LC-product is needed to apply the LUT to obtain Fennoscandic maps of V, H, CL and LAI. The LC-map to apply the LUT was prepared using MS-NFI maps, and MS-NFI maps were classified based on V and H to simplify the classification task (p5, r6-7).
o The clusters are not corrected using MD, but it is used to determine cluster-membership (i.e. into which group an observation with certain V and H combination belongs to based on smallest median MD (i.e. for each subgroup a gridded representation median MD values was calculated, and the Fig. 2. was formed by assigning each grid cell location for the subgroup which had the smallest median MD) (p5, r7-13)"
2. It will be important to provide information on the algorithms used in addition to the R routines, cf. Reviewer #3 comment:
R3: Method lacks in depth explanations and references are missing. For example, on several occasions, authors are quoting R routines with their cryptic abbreviations, but not providing any information or reference what is the basis for the algorithm used in that code.
Your response: "We can revise description of the methods and add citations to used algorithms/methods."
3. in your revision of the introduction as well as Material and Methods, please make sure that the land-cover classes are clearly separated from PFTs in their definition.
4. Please make sure this point is adequately addressed (i.e. clearly described and discussed) in the revised manuscript.
R3: "This is for at least two reasons: 1) regional climate models operate on 1-10 km resolution, and 2) LC classes need to be converted into PFTs (by cross-walking procedure). The question is how different would these 2 PFT maps converted from enhanced and original LC maps appear after aggregation to coarser resolution and cross-walking. Therefore,I suggest authors rewrite the article and clearly explain what they have done or to perform an analysis as I have suggested above. The latter approach would certainly serve to link the climate modelling and forest observation community, as they seem to aspire in the manuscript."
Your response: "As our enhanced LC-product has an attached LUT of the key forest structural attributes, no cross-walking procedure is needed, i.e., our LUT may be applied to forest classified pixels directly. Cross-walking is only needed to convert the ESA LC-product to a PFT-map. The influence of data aggregation to coarser resolution on raster values will be analyzed and discussed in the revised manuscript. We agree that that our paper needs substantial revision, and that the suggested corrections and analysis steps will increase the impact of our paper."

Best wishes,
Kirsten Thonicke.