the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Widespread increase of root zone storage capacity in the United States
Abstract. The root zone is the upper part of the unsaturated zone, where water and nutrients are accessible to plants, controlling hydrological responses, vegetation dynamics, biogeochemical processes, and land-atmospheric interaction. The root zone storage capacity (Sumax) represents the maximum subsurface moisture volume that can be accessed by the vegetation’s roots, controlling the partitioning of precipitation into storage, runoff and percolation. Previous work has illustrated that Sumax varies spatially, largely responding to climatic conditions. It can be therefore expected that Sumax varies temporally as well in response to climate change. However, this hypothesis has not been tested. In this study, we utilized a conceptual hydrological model and a dynamic parameter identification analysis method, to quantify the temporal trends of Sumax for 497 catchments in the USA. We found that 423 catchments (85 %) showed increasing Sumax, which averagely increased from 178 to 235 mm between 1980 and 2014. The increasing trend was also validated by multi-sources data and independent methods. Our results suggest that ecosystems dynamically adapt their root zone in response to climate change, which significantly affects hydrological processes and water resources availability. Moreover, the increase of Sumax significantly correlates to hydroclimatic indicators and vegetation dynamics. These results highlight the importance of considering the co-evolution of climate, ecosystems, and hydrology.
- Preprint
(1774 KB) - Metadata XML
-
Supplement
(191 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-550', Anonymous Referee #1, 14 Apr 2024
Review Liang et al. HESSD
This paper by Liang et al. investigates if the root zone storage capacity varies temporally in response to changing climatic conditions. To do so, the authors analyze 497 US catchments from the CAMELS database. The authors estimate temporally varying root zone storage capacities via two methods; the first one is based on a bucket-type hydrological model that contains a parameter S_umax, and the second one is based on the so-called mass curve technique (MCT) that uses a water balance approach to calculate the maximum water storage deficit. The authors find that root zone storage capacity increases in many places with time and attribute this increase to adaptation to changing climatic conditions, such as an increase in precipitation and potential evaporation. They conclude that vegetation dynamically adapts the size of its root zone to climatic conditions, a process that many hydrological models omit.
Summary and general comments
Overall, the paper is very well structured. The language is good but sometimes a bit imprecise, so the paper would benefit from a thorough reread. The topic is of interest to the HESS readership and has the potential to become a very interesting contribution, but it requires major revisions before it can be published.
As a side note, I would always encourage authors to include tables and figures in the text. This makes it easier to review.
Major comments
While the two methods use different datasets and approaches, I wonder how independent they really are. If the forcing data roughly agree (especially on longer than daily timescales and trends), then they contain similar information. The core of the FLEX model is more complex than a simple water balance, but follows a similar idea. The phrasing “multi-sources data and independent methods” made me expect a bit more than “just” another water balance approach, perhaps some independent observations/estimates of vegetation dynamics or rooting depths. I would thus suggest to rephrase “validated by multi-sources data and independent methods” in the abstract and the corresponding parts in the manuscript and state right away that these two methods were used. And I would encourage at least a discussion on other kinds of more independent data that could be used to evaluate the findings.
In the manuscript, it sometimes sounds like it is obvious that this parameter (and the MCT estimate) actually relates to root zone storage capacity, but somehow I'm still a little unsure (or perhaps unconvinced) about the physical meaning of the parameter. So, I think a bit more discussion on the physical meaning of these estimates would be good. For instance, in l.69 it says “such problems can be attenuated through specific modelling choices…” but later the problem of parameter identification and the difficulty of relating parameters to measurable properties (i.e. the root zone) isn’t discussed much.
Related to that, I was wondering if the authors ever tried to use the MCT on model outputs. It would be interesting to see if one could infer the then known parameter S_umax from an idealized modelling situation using the MCT.
I am a bit surprised by the results. If I am not mistaken, S_umax increases with aridity (PET/P) in Gao et al. (2014), but the temporal trend here shows the opposite. Why is that? Why would an increase in P lead to larger S_umax? Wouldn’t more precipitation lead to a smaller root zone? Does a concurrent increase in P and PET (an intensification if you will), require more storage to maximize ET? Or is the increase in P caused by more intense precipitation events that require more buffering potential in the root zone? Related to the last question, I was wondering why storm characteristics (e.g. inter-storm duration, event intensity) were not analyzed. Inter-storm duration was a strongly correlated climate attribute in Gao et al. (2014), so it would be interesting to see how that correlates with the observed changes, also given that it might change in the future. Overall, I think the discussion about spatial and temporal relationships between root zone storage capacity and climate attributes requires more depth and also a clearer description of the hypothesized processes. I think currently not all statements are totally coherent and the some of the statements are a bit vague or only loosely relate to the findings of the paper.
Generally, I think the discussion would benefit from some major edits. Some of the paragraphs do not relate much to the findings (e.g., l.396-405 or l.451-459 do not add much for me) and could be shortened or removed. Instead, I would find it interesting to see more discussion on the physical mechanisms that drive a change in root zone storage capacity (see above). Why would an increase in P lead to an increase in S_umax? Other questions that would be interesting are: How could we actually set the S_umax parameter dynamically if we do not have calibration data? What other independent data sources could be used to evaluate the findings of the study?
Minor comments
Abstract and also l.39: “The root zone is the upper part of the unsaturated zone…” – What if roots are below the water table?
Abstract: “controlling the partitioning into storage, runoff and percolation” and *transpiration*? This seems to be the most important flux that depends on the root zone.
Abstract: In the abstract, it sounds like the main focus is on the model-derived S_umax, but in the main text it sounds like this approach and the MCT approach are both equally important for the analysis. I would suggest reporting the numbers in l.330-341 also in the abstract and to add to the abstract that the MCT was used as a second method, rather than just writing “validated by multi-sources data and independent methods”.
l.39: “in which vegetation buffers water” – I am not sure if I’d say that vegetation buffers water, because it’s stored in the soil/weathered rock after all
l.41: “the real shape” – what do the authors mean here? The shape of the bucket? Or just the value of S_umax?
l.57: It is perhaps costly but one could think of measuring rooting depths etc. over time (using many similar plots etc.).
l.76: Hauser et al. (2022) estimated global changes in rooting depths. The study might be relevant here.
l.94: Perhaps also add Newman et al. (2015), which was a precursor to CAMELS US.
l.101: Was PET calculated by the authors or taken from CAMELS? I also assume that NDVI is not from CAMELS? Please clarify.
Figures 1,2: Finding colors for so many variables is difficult, but there should be better color palettes than this one with all the greens and reds.
Figure 2: I was not entirely clear to me over which time windows the variables were averaged. And what are the units? mm/yr?
Table 2 and 3 and the model description could be moved to an appendix. Instead, one summary paragraph could be added to the methods that explains the most important aspects of the models, incl. relevant parameters. For instance, I do not see why the routing parameters are explained, as they shouldn’t affect the partitioning in the root zone.
l.179: Was the model performance assessed over 5 years and was the window then also moved by 5 years or only by 1 year? That was not completely clear to me.
l.198: How exactly did the authors define net precipitation? Rain (i.e. liquid precipitation) minus interception?
l.215: Please check the sentence “continues in the end of the end”.
Methods: Was the MCT curve applied to the same catchments by extracting the fluxes from ERA5 via catchment shapes? This was not clear to me.
l.219-231: The selection of the 10yr, 20yr, 40yr, etc. based on vegetation type seems a bit arbitrary. It would be good to add some references and possibly also an uncertainty analysis.
l.227: How exactly did the authors calculate S_r for different return periods? Maybe add a few more explaining sentences. Are the return periods chosen (e.g. in Figure 3) just based on the vegetation type? Or are the return periods in Figure 3 chosen based on the results, so that they fit best? If this is too much repetition from Gao et al. (2014) it could be added to an appendix.
l.237: If the evaporation coefficient is calculated using Q/P, doesn’t it contain exactly the same information as the runoff coefficient? The correlations etc. seem to be exactly the same, so I don’t see why this is added. And why 7 data points per regression (which really is not a lot to calculate a trend).
l.239: Please define aridity and seasonality index.
l.248: I’d suggest deleting “unique”.
l.250: Do the authors mean “cluster averages” when they say “spatially mean”?
Figure 3 and l.269: This figure does not really show the spatial distribution, which to me sounds like a map, but rather the different clusters. I word suggest changing the phrasing here.
Clusters: some clusters are really small (n=6) while others are large (n=188). For example, the seven smallest clusters have fewer catchment than the largest. Does that affect the results and if yes, how?
l.329: The authors suddenly talk about a widespread increase, but so far no temporal changes have been presented.
Are the dynamics of S_umax only modelled with FLEX or also with the MCT? How was S_r calculated for 40yr return periods using a 5 year window?
Figure 4: I’d suggest to just show the important parameters here. Or rather move this figure to the appendix (like the model description).
l.368: drought index = aridity? Also, “humidity” is used in l.361. If this refers to the aridity index, I would suggest to always use aridity to have consistent terminology.
l.385: “splitter” seems like an odd wording here.
l.396-405: I am not quite sure what the take home message of this paragraph is.
l.424: “Both studies demonstrated the increase of S_umax in response to an increase of the aridity index. But Figure 7 and the associated results section states the opposite, doesn’t it?
l.434: What do the authors mean by human activities? There shouldn’t be much irrigation in the CAMELS catchments.
l.461: “According to the Budyko framework … is highly influenced by S_umax”. That is not true, the Budyko framework (at least Budyko’s original ideas) states that it is only aridity that controls the partitioning into streamflow and evaporation, and there is no storage parameter involved.
l.470: How would the authors suggest to use dynamic parameters if we do not have calibration data?
Discussion: it might be interesting to connect the results a bit more to literature that does not come from the hydrology community and provides a somewhat complementary view on the topic, e.g. Tumber‐Dávila et al. (2022).
On the term runoff: I would suggest using streamflow if you talk about what is measured at a gauge, because it is less ambiguous.
References
Hauser, E., Sullivan, P. L., Flores, A. N., Hirmas, D., & Billings, S. A. (2022). Global‐Scale Shifts in Rooting Depths Due To Anthropocene Land Cover Changes Pose Unexamined Consequences for Critical Zone Functioning. Earth's Future, 10(11), e2022EF002897.
Gao, H., Hrachowitz, M., Schymanski, S. J., Fenicia, F., S_riwongsitanon, N., & Savenije, H. H. G. (2014). Climate controls how ecosystems size the root zone storage capacity at catchment scale. Geophysical Research Letters, 41(22), 7916-7923.
Tumber‐Dávila, S. J., Schenk, H. J., Du, E., & Jackson, R. B. (2022). Plant sizes and shapes above and belowground and their interactions with climate. New Phytologist, 235(3), 1032-1056.
Citation: https://doi.org/10.5194/egusphere-2024-550-RC1 - AC1: 'Reply on RC1', Hongkai Gao, 24 May 2024
-
RC2: 'Comment on egusphere-2024-550', Anonymous Referee #2, 02 Jul 2024
This manuscript presents an interesting study on root zone storage capacities. However, I have some rather fundamental concerns, namely, using ERA5 data and neglecting parameter uncertainties. Both are straightforward to address, but this would require significant work for new model runs. Below, I discuss these two concerns and list a few minor concerns:
- Use of ERA5 data: there might be good reasons to use ERA5 data. These data are, for instance, a suitable option for global studies. However, one has to be aware of the substantial uncertainties of these data and for the US, the use of ‘real’ (station-based) observations is preferable, especiallyy for precipitation data (for a detailed analysis see https://egusphere.copernicus.org/preprints/2024/egusphere-2024-864/). From the text, I am not fully sure whether ERA5 data is used for both approaches or only the second one (MCT). Still, both cases would be problematic and I would recommend using station-based data, especially for precipitation, in both cases.
- While details on how the model has been calibrated are missing, my understanding from reading the manuscript is, that there was one single parameter set for each catchment. This would not be state-of-the-art but one should consider an ensemble of good/acceptable parameterisations. Please explain how parameter uncertainty was considered. If it was neglected, as I am afraid of, I would recommend to use some approach that allows for considering multiple parameter sets.
Minor comments:
L122: please provide details on how the model was calibrated
L153: this sounds like the soil routine in the HBV model. If yes, a reference to Bergström et al. would be suitable.
L169ff: with the focus on soil water storage, I wonder whether it is appropriate to use KGE as the objective function, which is known to focus on performances during high-flow events.
L182: Sorry, but I am not sure I fully understand how parameter sets are combined within one cluster. Despite the grouping by Jehn, parameter sets must vary largely within each cluster. If anything, aggregating simulated time series might be more appropriate than averaging parameter values.
Equations: please avoid multi-letter variable names, these are mathematically incorrect (SM means S times M)
Figures: I find the figures hard to read. Partly this might be due to the small size. In Figure 5 the different axes use a different scale, which is unfortunate.
Citation: https://doi.org/10.5194/egusphere-2024-550-RC2 - AC2: 'Reply on RC2', Hongkai Gao, 10 Jul 2024
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
570 | 145 | 29 | 744 | 44 | 21 | 19 |
- HTML: 570
- PDF: 145
- XML: 29
- Total: 744
- Supplement: 44
- BibTeX: 21
- EndNote: 19
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1