The paper presents an evaluation of the spatial pattern of distributed hydrologic model simulation (ET in this paper) against independent ET map derived from remote sensing data. The method to quantify the spatial pattern matching uses EOF. The paper updated ET parameterization of their hydrologic model based on remote sensed vegetation data to overcome issues of previous parametrization i.e., spatial discontinuity of parameter fields that are derived from per-region calibration. The paper show that updated parameterization improves ET spatial pattern over use of the previous model parameters.
The paper points out an important aspect of hydrologic model evaluation – spatial distribution of model simulations. Overall introduction nicely lays out this problem. I agree that few studies have looked at this aspect of evaluation. The update of the ET parameters for the DK model using remote sensing data is somewhat specific to their model, but I think the paper provides a good case to use remote sensing data in a conjunction with transfer functions to compute spatially distributed parameter values. I would recommend minor revision before being accepted for HESS publication. Most of my comments could be addressed by further revising texts.
Section1. One minor comment is on P3, L10-11. Is this really the most important goal of this paper? I understood that TSEB ET map is one method to generate ET maps that can be used for evaluation but potentially other ET products could have been used for this study. A primary goal of this paper is to evaluate spatial pattern of distributed ET from hydrologic model and improve the ET parameterization. If so, revise the statement of this paper goal.
Section 2.1- TSEB setup, especially theory part is a little confusing to me. It was perfectly fine till L28, and then rest of the descriptions give me hard time understand how latent heat is actually computed. There is one unknown Tc (canopy temperature). So I assume the code iterates model simulation by changing Tc, but why does this need Priestley-Taylor? Need some improved descriptions for this model iteration and how final ET is estimated. A few more minor comments include - 1) what is the assumption for ground heat transfer? And 2) it could be helpful to include a table showing all the input data (or maybe Figure 3 list complete parameter/variables?).
Section 2.1.2 describe great deal of LAI, Green Fraction (GF). I wonder why authors did not use MODIS LAI GF product, which are widely used for the other studies.
Section 2.2. P9 L12-16 discuss crop coefficient (Kc). I understood that the model use some lookup table to get reference ET for a given land cover and use this Kc to scale to get ET. I would suggest describing reference ET. If readers (including myself) are not familiar with Mike-SHE structure, just need to guess it.
Section 2.3. I feel that Joint EOFs need descriptions on how two matrices (observation and simulation) are combined. Joining matrices in row (time) or column (space)? It would be helpful to describe how to interpret EOFs with combined matrix (like description in EOF section in Koch et al. 2015). Equation 18 is not shown in the results. I am not sure if this equation is needed.
Section 3.2. ET maps (Figures 5-9) look nice and support the descriptions in the result section. Somehow I expected that EOF analysis look at two comparison -1) original model’s ET and TSEB ET, 2) modified model’s ET and TSEB ET. What is justification of comparing two models ET instead of model versus independent dataset (i.e. TSEB ET).
Section 3.3. First of all, does “WBE” mean runoff volume bias (against observed discharge), or literally mean long-term water balance closure error (P-RO-ET should be close to zero for long term mean)? I understood the former. If not, I wonder what is going on the model water balance (water balance error is large).
Second, this paper modifies the parameterization related to ET, therefore I would expect runoff volume is affected (that is why I thought WBE was runoff volume bias). Although NSE is the first metric to look at, maybe look at NSE decomposed metrics (bias, variability, and correlation) – see KGE paper (http://dx.doi.org/10.1016/j.jhydrol.2009.08.003). I would expect that bias is mostly affected, which contributed to NSE deterioration. This way, you could discuss more why (or how) NSE is deteriorated.
Line by line textual comments
P5, L12. Spell out VZA (I think it appears at the first time).
P6, L11. 46 daily ET MAP?
P7, L14. This is the first time that study period is mentioned. Maybe specify beginning of Section 2.
P7, L7 and L10. Here notation is confusing because Rn is used for both instantaneous and daily. Also actual -> instantaneous?
P7, L28. What is spam here?
P10, L14. AET. Is this actual ET? This paper does not mention “potential” ET, so just use ET?
P11, L15-18. This paragraph seems to be out of place, and not necessary.
P13, L1-4. I think that discussion on this scatter plots fit better before discussing clay, LST and LAI maps (Figure 9).
P13, L31. RS->remote sensing?