General comments
Although I was not part of the first review round, I generally agree with the comments from the previous reviewers. That is, the paper describes a simple approach to correct for biases in CLM simulations of evaporation over CONUS by rescaling the model ET to GLEAM. Although the approach is not really innovative and does not result in improved physical understanding, it is effective in reducing biases and improving state and flux estimates, and therefore, can be of interest to the community. I also believe the authors have substantially improved the manuscript after revision, particularly by including additional independent evaporation datasets (other than GLEAM) for validation, also at the daily time scale. Below, I still formulated some minor comments, which I believe could help further improving the manuscript. For future reference, please include line numbers in the manuscript to help tracking comments.
Specific comments
1. Title: given that GLEAM evaporation is more of a model product (forced by remote sensing observations), than an actual remote sensing product, I would recommend to change the title accordingly.
2. Introduction/methods: I think it is important to add a caveat somewhere in the introduction or method section, in that the approach assumes full trust in the GLEAM evaporation, and no trust at all in the CLM evaporation. From a data merging/assimilation perspective, it could be more desirable if information from both CLM and GLEAM would be exploited, based on their relative uncertainties/performances. In its current form, the approach can degrade the performance of ET estimates in places/periods where GLEAM evaporation may perform less well. For instance, GLEAM has a very simple snow module, so it might have issues over snow-covered areas, like the Sierra Nevada (an area that actually stands out in Figure 2).
3. Page 4, 3rd paragraph: The authors mention data assimilation as another approach for reducing biases. I would rephrase this part, as the general purpose of data assimilation is not to remove biases, but to correct for random errors (e.g. in the forcings). Biases between model forecasts and observations should be corrected prior to assimilation.
4. Page 6, 1st paragraph: The authors state that CLM can realistically capture the spatial pattern of ET over CONUS. This seems to be contradictory to what is shown later in the results section, where large biases occur over space (e.g. Figure 2) and time. Maybe this statement needs to be nuanced, by mentioning the scale to which it applies, or mentioning the reference dataset on which it is based.
5. Page 6, 3rd paragraph: I was wondering if, by applying constant monthly scaling factors, there may be jumps in simulated states/fluxes from one month to the other. Alternatively, the authors could consider interpolating scaling factors to get a smooth transition. Please comment on whether such jumps are observed in the simulations (e.g. of daily time scale runoff, soil moisture, etc.)
6. Page 7, 2nd paragraph: Please describe which adjustments are made in CLM if soil moisture cannot support the evaporative demand. This is a quite important aspect in view of the approach.
7. Equations 1-5: Are absolute values of S and R considered for calculating statistics like bias? If not, and when averaging over grid cells, simulations which have negative bias in one place and positive bias in another place would show up as bias-free, which is incorrect.
8. Results: I believe this section would improve if the authors would at least try to formulate some hypotheses on why biases in CLM simulations of ET occur over some areas and time periods. This could particularly help future studies to improve physical model aspects. Based on your extensive validation analysis of CLM ET, what can be learned? Are ET biases likely originating from biases in soil moisture, biases in radiation/temperature, model physics or parameters (and if so, which parameters)? Is this different for different regions, e.g. water- vs energy-limited regions, etc.?
9. Page 13, 2nd paragraph, line 12: I believe it is important to stress that improvements in model performance are relative to GLEAM. GLEAM has its own shortcomings, so improving towards GLEAM not necessarily means improving estimates overall.
10. Page 20, 2nd paragraph: I appreciated the statement that the approach does not replace model improvements through better parameterization of physical processes. This frames the value of your approach, and helps readers to identify shortcomings and advantages of the approach.
Technical corrections
I would strongly encourage the authors to perform an in-depth check of the grammar throughout the manuscript. There are still many mistakes, e.g., on using plural nouns and tenses. The following list is not comprehensive.
Page 6, 3rd paragraph, 1st line: capable instead of cable, patterns instead of pattern
Page 7, 1st paragraph, 5th line: components instead of component
Page 7, 2nd paragraph, 4th line: start instead of starts
Page 8, last paragraph: Do you mean time steps instead of time series?
Page 9, 1st paragraph, line 11: show instead of shows
Page 9, 1st paragraph, line 12: spatial resolution of the GLEAM dataset
Page 9, 1st paragraph, line 13: GLEAM instead of GELAM
Page 10, 2nd paragraph, 1st line: 16 sites / used to validate the model / those sites span four …
Page 11, 2nd paragraph, line 9: agrees
Page 11, 2nd paragraph, line 13: number of stations
Page 11, 2nd paragraph, line 17: linear instead of liner
Page 13, 2nd paragraph, line 13: statistics are superior
Page 15, line 2: CONUS instead o CNOUS
Page 15, line 3: SON is the reason … please rephrase
Figures: Please make sure you refer to panels (a, b, c, …) in all figure captions (for instance figure 4) |