The authors provided an extensive response-to-review document, which is really appreciated, as well as a thoroughly revised manuscript. I'm very impressed with the response to the review provided by the authors, because it provided many insights into the new percolation scheme in CROCUS. And to many issues raised by the reviewers, the responses were very informative and clear. However, I'm a little bit disappointed that some of the interesting points in this document did not end up in the revised manuscript, although I think that by doing so, the significance of the manuscript would be enhanced and it would result into a much more interesting manuscript. Please find detailed comments below. I consider point 1, 2 and 5 are of particular importance to be addressed adequately, after which I can recommend publication.
Some remarks about the response-to-review document:
1) Particularly, I think that the test of switching on/off the compaction routine is very insightful, and the comparison of Fig. A and B in the response document should end up in the final manuscript in my opinion. First, Fig. A suggests that the numerical implementation of Richards equation is behaving numerically stable, and second, the comparison of Fig. A and B shows clearly where the next developments should be. This is not only for the CROCUS team, but also for the snow community as a whole: how to accurately describe the effect of liquid water flow on compaction and wet snow metamorphism, as well as how to take care of feedback mechanisms between both. This discussion would just increase the group of researchers for which the manuscript is interesting.
2) Related to the previous issue: It's very clear now that the unrealistic alternating pattern arises from the interaction of the water percolation routine and the snow settling. As far as I now understand it, the problem is that higher values of LWC are typically associated with stronger settling rates. This leads to more snow compaction, higher densities, which in turn results into denser packing of the snow crystals, higher capillary suction, and thus, larger values of LWC, which in turn increases the snow settling, etc. Am I correct here? If the authors agree with this reasoning, I think it is a good idea to explicitly explain it in the manuscript (it is now rather implicit). Actually, an obvious, and easy solution to this problem could be to limit the LWC used in the parameterizations for snow compaction and snow metamorphism to the 5% from the bucket scheme, with the motivation that this is done because the parameterizations for settling and metamorphism are typically developed and tested using the bucket scheme, and future studies may address the behaviour of snow with very high LWC. Did the authors tried this? I strongly recommend to try this out and report the results in the manuscript. If it doesn't solve the problem, it shows that the problems are of more substantial nature.
3) I asked to mention the CPU time needed somewhere in the manuscript. I do understand the point of the authors that many factors influence the CPU time, as for example the file output resolution. However, I still think it is very important for such model/numerics description papers to give readers an idea of the computational burden of the proposed model/numerics improvements. So maybe the authors can just provide the relative extra CPU time to use the new Richards equation scheme over the old bucket scheme, or at least state that the CPU time needed for the Richards equation scheme is of the same order of magnitude as the bucket scheme, as it seems to be. Note that it is important to consider that a switch to 1-dimensional Richards equation is generally accompanied by a very acceptable increase in computational time, in contrast to the three-dimensional snow models recently proposed (for example by Hirashima et al. 2014, Leroux and Pomeroy (2017)), which are accompanied by such a large increase of computational time, that a useful application on seasonal time scales or large spatial scales on real natural snowpacks is not (yet) feasible.
4) I made the remark that a plot of snowpack runoff is very useful. The in response provided plot of the soil moisture in the upper soil layer is also interesting, but not so informative when it comes to snowpack runoff. The reason I asked about runoff is that in snow modelling, many researchers are interested in the hydrological aspects, mostly snowpack runoff. It should just be easy to plot it, as it results directly from the free-drainage boundary condition (just translate gradient in pressure head to the flux). Two important aspects that I hope show up is a shift later in the day of the arrival of the meltwater at the bottom of the snowpack, and a recession curve at night. This aspect is not a show-stopper for me, but it would just enhance the impact of the manuscript.
5) Fig. A in the response-to-review document is very important and I strongly recommend to take it into the manuscript. However, Fig. A shows one confusing thing: after passing of the melt water front, a significant amount of liquid water (typically 2-3%) should be held in the capillaries, although the plot seems to suggest that the water content after passing of the melt water front falls back to almost 0 in some layers. One suspicion I have is that Eq. 7 is not implemented correctly. As far as I understand the code, Eq. 7 from the manuscript translates into the following source code line:
ZTHETA_R= MIN(ZTHETA*.75, 0.02)
However, this is not consistent with Eq. 7. It should translate into:
IF(ZTHETA.LT.0.02)THEN
ZTHETA_R=0.75*ZTHETA
ELSE
ZHETA_R=0.02
ENDIF
The latter approach is in my opinion the better one. Otherwise, a condition can occur that even when theta>0.02, theta_r is reduced below 0.02, after which theta gets smaller, after which theta_r is reduced, and this continues all the way to 0, although we know that a bulk liquid water content for a wet snowpack is typically more than 2%. Note that for numerical stability, it may be better to write something like IF(ZTHETA.LT.(0.02 + theta_min)), such that ZTHETA is always significantly larger than ZTHETA_R, which is required in the van Genuchten model.
6) The discussion about the time steps inside the Richards equation solver in combination with the mass balance check, as provided in the response-to-review document, should be present in the manuscript I think, in a much more condensed form of course. It is important to note in the manuscript that the mass balance error is acceptable. Otherwise, I can imagine that other readers will also feel that the allowed mass balance error is large compared to the minimum allowed time step.
Some remarks about the manuscript:
7) Section 6.2.2: I think the authors are too negative about the free-flow bottom boundary (p13,l10-11). First, in most snow models, the outflow from the snowpack is not at all constrained by the underlying soil. The bucket approach in CROCUS is probably also not taking into account the conditions of the underlying soil. One can always argue that the soil module of ISBA should take care of the incoming water flux from the snowpack, i.e., decide if it infiltrates into the soil, or creates overland flow. Note that in reality, a frozen, saturated, or extremely dry soil can have such a reduced infiltration capacity, that meltwater from the snowpack creates a significant amount of lateral overland flow, and thereby constitutes a significant flood risk. I agree with the authors that the approach of SNOWPACK to solve the snow-soil continuum at once has the advantage that these processes can be adequately captured, and with the SNOWPACK model, we are indeed able to reproduce melt pond formation from snowpack runoff in case the soil has limited infiltration capacity. On the other hand, one looses the sophisticated coupling some hydrological models have from the unsaturated zone to the aquifer and streamflow (I'm not sure how this is with ISBA). I think in section 6.2.2., the authors may want to discuss some of these aspects there.
8) Prewetting amount: If I understand correctly, Fig 8 A and D should be identical with Fig. 10 A and B, as both have the same prewetting amount of 10^-5? Yet, there is a clear difference, but it is not clear where this originates from. Is it a typo in the caption of Fig 8, that it should actually be 10^-6? Or what else changed between both simulations? Another issue here: sections 5.3 and 6.4 now fail to explain why the prewetting is so important, but I think the reason is that when it is set too high, hydraulic conductivity becomes already so significant, that water percolates, even when we should still consider the snowpack "dry". When more water is added after the prewetting and this is refrozen every time step, heat is advected. If authors agree with this explanation, they may consider adding it to the manuscript. Note that in SNOWPACK, we don't refreeze prewetting water every time step. We have hysteresis, i.e., the threshold for executing phase changes is a factor 10 larger than the prewetting amount. I think a similar approach in CROCUS would help to reduce the warming effect. But I consider this something for future work. In my opinion, the discussion of the sensitivity of the pre-wetting amount is not so interesting, and as I suggest to add figures describing the influence of switching on/off the compaction routines, this section and Fig. 10 could be removed to save space, if the authors wish to do so.
Technical corrections (line numbers refer to revised manuscript):
(Note that I think that some of the technical corrections should have been identified before submission by a proper proofreading by author and co-authors.)
- p1,l11: "thought" -> "through"
- p1,l12: add gravity: "capillary suction, gravity and hydraulic conductivity"
- p1,l15: "coved" -> "covered" (?)
- p1,l22: "The parameterization ... crust layers." This is an assumption (although a well justified one) and not really supported by data in the manuscript, so I think this sentence is misplaced in the abstract.
- p1,l30: either remove comma or write: "time consuming, and LWC"
- p1, l30-31: "change over timescales that are"
- p2, l2: remove comma: "rescue workers have reported"
- p2, l26: "due to suction and sloping terrain and water pooling"
- p2, l29: "and heterogeneous" -> "as well as heterogeneous"
- p3, l22: "however" -> "although"
- p5, l2: "found the speed"
- p6, l12: note that Eq. 7 is not a continuous function, there is a break at theta=0.02. Either make the function continuous, or say "a piecewise function".
- p6, l17: I suggest to write: "which corresponds to a minimum pressure head, for which it holds for every dry snow layer that the liquid water content is smaller than a prescribed minimum value \theta_min."
- p6: Eq. 8 and 9 uses different symbols to indicate multiplication. The "x" is not adequate for scalar multiplication.
- p6, l28: add comma "is a complex system, it is"
- p7, l2: Eq 8 should refer to Eq. 10? (Occurs twice)
- p7, l7: add comma "computations, the following"
- p7, l8: I suggest "There must be a substantial snowpack: if there are less than 3 layers"
- p7, l9: The sentence is incomplete.
- p7, l26: "borders"
- p8, l25: I would move the sentence "The lower boundary is the soil-snow interface." Before starting the paragraph: "There are two options". Otherwise it is not clear that the bottom boundary condition refers to the snowpack, and not the soil.
- p9, l2: Eq. 10 should point to Eq. 11?
- p9, l24: For clarity, I suggest: "The peak shortwave radiation"
- p10, l20-21: This is not so clear. I suggest: "causes the surface layers to get wet, while deeper layers remain below freezing".
- p10, l24: "shows the formation"
- p11, l4: I would write: "pore space that is filled by water (i.e., the saturation) ...", as "saturation" is the term used in the figure.
- p11, l21-22: This sentence is not grammatically correct. Should be something like ".. is drastically different .., although the timing ..."
- p12, l4-5: "Routines such as for compaction and grain metamorphism were ..."
- p12, l14-20: I guess the authors do not plan to print this in bold.
- p12, l18: "modles"
- p12, l25: I would start this sentence with: "When using the upper soil layer as lower boundary, the hydraulic conductivity" to make clear that this is in contrast to the free-flow boundary.
- p13, l8: "there is there is"
- p15, l6: add comma "snow layer, snow"
- p15, l8: I think a nicer formulation would be: "are the only crystal type that needs to be described by a water retention curve."
- p15, l13: note that a recent study shows the effect of hysteresis in snow, a study that I recommend citing here: Leroux and Pomeroy (2017).
- p16, l5: "The parameterizations"
- p15, l7 and p16, l9. Note that the abbreviation MF is not introduced, but as it is only used twice, I would recommend to write "melt forms" in both cases and not use an abbreviation.
Fig 9 and 10: Make the captions more in line with Fig. 8, for example: "Crocus output for Neverland forcing using different time steps ..."
Fig 10: Subfigure B has different label, and wrongly formatted. I prefer the B style, that the labels show everyhwere "theta_min = XXXX"
Fig 10: The caption only indicates what A and B refer to, but not C and D, and according to the figure itself B is also with pre-wetting 10^-5 and not 10^-7 as is written in the caption.
References:
Hirashima, H., Yamaguchi, S., and Katsushima, T.: A multidimensional water transport model to reproduce preferential flow in the snowpack, Cold Reg. Sci. Technol., 108, 80-90, doi:10.1016/j.coldregions.2014.09.004, 2014.
Nicolas R. Leroux, John W. Pomeroy, Modelling capillary hysteresis effects on preferential flow through melting and cold layered snowpacks, Advances in Water Resources, Volume 107, 2017, Pages 250-264, ISSN 0309-1708, http://dx.doi.org/10.1016/j.advwatres.2017.06.024. |