The manuscript by Olivia-Guerra et al. investigates the impact of assimilating satellite-retrieved land surface temperature (LST) data into a land surface model, here 'ORCHIDEE'. The authors investigate two different optimization methods for the assimilation, a gradient descent method and a genetic algorithm. The employed methods and experiments are well presented and the results are conclusive: assimilating 3-hourly CCI-LST data into ORCHIDEE and employing GA optimization, which proved to be the consistently superior method, lead to a median reduction of 20% in LST and turbulent fluxes (data from 34 sites and for a 12 year validation period). The manuscript is well structured and written and should be published after the minor points listed below have been addressed.

Specific points:

line 102: 'The CCI-LST observations  have an associated ...'

line 143: '..., each with increasing grid spacing ...'

lines 152-154: consider reformulating these two sentences - they are difficult to understand.

line 180: ' ... assimilations initiated with ...'

line 213: 'Based on this test, ... to represent the parameters.' 

line 214: '.... to each site allows to identify sensitive ...'

line 217: 'For the 34 sites, ...'

line 219: '... in the sensitivity analysis: model parameter name, description and default values ...'

line 220: 'The 11 sensitive parameters to be optimized for the ES-Abr site in the twin DA ...'

line 235-236: A set of ... to be optimized is then used as prior data in the optimization.

Equation 2: in my opinion there is a bracket around the terms in front of the '100' missing: currently the equation does not provide values in percent.

Caption Table 2: 'Example of typical ORCHIDEE parameters optimized in the DA experiments and used to determine the optimum strategy.

line 279: '... into which specific error components are improving ...'

Figures 8 and 9: in my opinion the y-axes of the sub-plots should be labelled 'RMSE' instead of 'MSE'; similarly, 'MSD' on the x-axes should be 'RMSD'. In the caption it should then also say 'root mean square difference' (RMSD), for which the definition could be repeated here for clarity.

The article ‘Assimilating ESA-CCI Land Surface Temperature into the ORCHIDEE Land Surface Model: Insights from a multi-site study across Europe’, by Olivera-Guerra et al, studies the potential of the LST-CCI fusion product (delivered by ESA) to constrain the ORCHIDEE surface model. This model is used as a surface parameterisation in the LMDZ/IPSL GCM, with the aim of simulating the energy fluxes between the Earth's surface and the near atmosphere (key variables of the atmospherically lower boundary conditions for GCM). The study focuses on the impacts of assimilation on energy fluxes, without addressing the issue of seasonal vegetation development or carbon fluxes (i.e. processes simulated by the ORCHIDEE model). The assimilation approach (ORCHIDAS), already tested several times with ORCHIDEE, is a variational approach. It aims to simultaneously correct several model parameters, through the minimisation of the observed surface temperature simulation over a time window (1 year here). The study is very comprehensive and is presented in a logical and progressive way, starting with a general case study (detailed and presented in the article for a particular site), set up from several angles of analysis: sensitivity analysis, 2 optimisation methods tested, assimilation of various characteristic features of the LST, then extension to 34 European study sites). The article then moves progressively towards an application more suited to the large scale, seeking to identify the most promising assimilation path, and attempting to assess the impact of the average parameter set obtained for each of the PFTs.

General comments

Overall, the article is clear and well-written and worthy of publication in the journal HESS. It is very well documented and very explicit in both the presentation and justification of the proposed assimilation method. The results are presented clearly, progressively and comprehensively, particularly in terms of analysing the impact of assimilation on energy flows. However, it is regrettable that surface process analysis has been carried out in relation to assimilation, or at least justified in the text if this is not possible. This concerns at least the impact of optimisation on the phenology of the vegetation simulated by the model, and even the availability of soil water, which mainly controls surface fluxes. This is all the more regrettable for LAI, which remains a global products which can easily assimilated and a key variable both in the partitioning of energy between soil and vegetation, and in the partitioning of radiative and convective fluxes. Such an analysis would also allow us to better qualify the differences obtained for energy flux performances.

Another regrettable point is that the article deals results in a ‘results and discussion’ section, which remains very (too) focused on the quantitative analysis of flux performance, and not necessarily from a broader angle of approach, as would be expected in a discussion. Given the great interest of the paper and the advances it makes from the point of view of exploiting LST data, it would surely have been wiser to bring all the different discussion points scattered throughout the current section, in a dedicated section also built around perspectives. As exemaple, discussion point could be introduced on the recommendations arising from the study, on the characteristics and potential of future satellite missions (e.g. the TRSIHNA mission mentioned in the article); the consequences of optimisation precisely on vegetation, carbon fluxes or soil moisture; possible expectations for future assimilation on a regional scale; the benefit or possible complementarity of LST assimilation with other global products provided, in particular the LAI, the benefit of which had previously been shown in ORCHIDEE (e.g. Demarty et al. , GRL, 2018).

I therefore believe that the article is acceptable after minor corrections, in particular after reinforcing these elements with both discussions (ideally in a separate section) and process analysis (impact of parameterisation on the simulated processes; possible addition of an analysis or figure on the simulated LAI, in the appendix).

Detailed/minor comments

Line 123: why do you choose to assess net radiation (NR), which merges solar and thermal effects (possible compensation between them)? Is it not possible to distinguish the analyses of SWout and LWnet, which would allow a better understanding of the potential of LST?

Line 128: The choice of a particular year (2018) is not without consequence in the optimisation of the parameters. Even though 2018 seems to be a commun year for all the sites and could maximise the number of sensitive parameters, particularly because of the stress periods, it does not preclude optimisation being carried out in different years for each site (similarly, optimisation does not always focus on the same parameters for each site). Has the effect of choosing 2018 on optimisation been analysed? This is not currently discussed in the article.

Section 2.1.2: Do LAI observations exist? Can you precise in the section please.

Line 163 – Equation 1: Explain me why a minus sign appears in the first factor dedicated to observations and a plus sign in the parameter factor ? This is not very intuitive. Indicate in text that the ‘T’ in Eq.1 represents the transpose of the matrix.

Line 205-213: What exactly is meant by ‘trajectories’ (parameter p) ? This term seem to differ from the number of levels considered for each parameter (line 209), but I is clear what is it ? If the number of levels differs for each parameter, is the formulation (n+1)p no longer valid? More details need to be provided in the text in order to be able to well understand what has been done.

Line 219 - Table 1 - Appendix B

• Table 1 is both general to the 34 sites and specific to the ES-Ab site, which is difficult to understand when reading the section (although this point is better understood later). I suggest adding a few clarifications to the text and the legend of Table 1. It should also be made clear that the asterisk is associated with a scaling factor for the parameter concerned by the nomenclature. Finally, it would appear that there are 12 parameters listed in bold and not 11 as specified in the text and figure in Appendix B (θcrit parameter in bold?). Adding a column with the a priori values of the ES-Ab parameters (not its scaling factor) also seems useful.
• Appendix B is interesting and relevant. However, it remains very concise and lacks commentary. I didn't quite understand how the 11 parameters out of all those tested were selected. If a threshold on the sensitivity score is set for this purpose, then the value chosen should be indicated and whether it is common to all 34 sites.

Line 228: The introduction to the TRISHNA mission appears a little abruptly in the text. Please add 1-2 sentences of explanation and a reference.

Line 259: I think equation 2 is wrong. There must be a parenthesis missing in the wording.

Line 372 + Figure 3 and its link with Table 1: Note that for parameters modulated by scaling factors, Figure 3 shows ranges of variation in parameter values (and not their scaling factors). The names of the parameters in the titles of the sub-figures should therefore be changed (by removing the asterisk) or the sub-figures changed accordingly to scale factors. Wouldn't Figure 3 be more explicit about the performance of the optimisation if you fixed the ordinate scale to the ranges of variability given to each of the parameters? 

Section 3.1.3 + Appendix C: The analysis is very interesting, but remains very general for a low weight in the article. The question is whether to keep this section in the article or to move it completely to the appendix. If you want to keep it in the main part of the article, we are currently awaiting answers in the text on the distinction between the effects of ‘data availability’ and their ‘associated uncertainties’ (these two factors are analysed together). These elements would be useful in formulating recommendations to the remote sensing community, such as the TRISHNA mission (highlighted in the text). A point of detail about Figure C1, for a possible addition of the boxplots obtained in the previous section. This would greatly improve the readability of the document, as the reader currently has to navigate throughout the article to visually compare the 2 figures.

Lines 415-425 + Figure 6: In arid environments, the heat conduction flux in the soil (G) and the water infiltration capacity in the soil are processes that are becoming predominant. The question of their representation in ORCHIDEE may arise, in particular to explain the significant differences in the H flux. The results may suggest that the LST is improved for the wrong reasons in the minimisation. If the elements of discussion remain included in this section, more elements of analysis need to be introduced on the description and representation of the dominant processes in the model.

Sections 3.2.1 and 3.2.2: The analyses are interesting but currently remain very focused on the general performance of the assimilation procedure. Yet one of the interests of the analysis over the multiannual period lies precisely in the consequences of the parameterisation obtained after assimilation (even though no interannual drift is highlighted by PFT). The final sections sound too much like a fairly basic qualitative presentation of the scores obtained, whereas analyses could have been made, for example, of the impact of optimisation in 2018 on the total period. These elements would make it possible to discuss the limits of the contribution of the LST to represent certain key processes, such as LAI or soil moisture. It would also make it possible to discuss the complementarity of the data sources that can be assimilated (for example by means of a joint assimilation of several types of data, which is technically feasible with the proposed procedure). It is therefore at this stage of reading that the feeling of a slight lack of height and/or perspectives in a dedicated ‘discussion’ section becomes more firmly rooted.