Future Colorado River Basin Drought and Surplus
Round 1
Reviewer 1 Report (Previous Reviewer 3)
The revised manuscript shows some changes that have a bit improved the quality of presentation. The manuscript was and still is well written and structured. Despite this, my perplexity, underlined in the first round of review, from a general point of view, about the approach proposed by the authors still holds. I think that the study and specifically conclusions are rather inconsistent, just describing results of calculations without attempting any discussion about the contradictory differences observed between GCMs runs. In particular, authors show that CanESM2 andCNRM-CM5 models had a wet signal, and HadGEM2 and MI-ROC5 models had a dry signal but no discussions about the reliability of these findings is provided. What kind of conclusion could we draw on the future streamflows of the Colorado river on the basis of such contradictory results ?
GCM's simulations provide a large set of atmospheric variables that would allow in principle an analysis of climate patterns responsible for dry or wet conditions and of the differences among models. To understand why models produce such different results would be very useful to reduce the uncertainties about future projections and to provide reliable scenarios to drive water management actions. Therefore I encourage authors to explore this issue that surely will give worth to the study.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report (Previous Reviewer 2)
Table 1 still does not relate the stations on the map to the station names. The map shows station numbers and the table shows station names, but it is not possible to locate the station name on the map. Table 1 needs a column that shows the station number for each station named.
201-202 “The largest changes in drought magnitude (as indicated by negative values in Table 2) appear in four Upper Basin stations (East River, Yellowstone River, White River, Fish 202 Creek) that show a higher magnitude in all scenarios corresponding to the longer drought periods (Table 3).” This is backwards. Table 2 shows duration and Table 3 shows magnitude.
177 “The basin-wide and station specific results for historical and future… are shown in Tables 2,3 and 4, and Figures 2 and 3.” What is meant by “basin-wide” results? These are only station specific results.
The Tables 2, 3 and 4 and Figures 2 and 3 are misleading in listing Lees Ferry along with the other stations with regard to historical vs future because of the difference in definition of historical for Lees Ferry (115 years for Lees Ferry vs 70 years for the other stations). Also, in the descriptions of the results, Lees Ferry results are compared with the other station results without noting this difference in historical length of data. This is a significant flaw in the study. All the results except for “Further Analysis for Lees Ferry” should be done with Lees Ferry historical data of 1951-2021 for a fair comparison with the other stations. The result with the longer period could also be added, but noting the difference in historical length.
Section 3.2 should then note the extended length of historical in presenting those results.
Thank you for adding the Conclusions section. What is the significance or conclusions for the spatial results? The description of Figures 2 and 3 only address the differences between the climate scenarios, and no comments about the spatial variation. Is there any significance?
Comments for author File: Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report (Previous Reviewer 3)
Authors have significantly improved the quality of manuscript. Thus in my opinion the paper could be published on Hydrology.
Reviewer 2 Report (Previous Reviewer 2)
No further comments.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
General comments: Overall, the paper is well structured and tells a fluid story. Nevertheless, the section on future streamflow data needs to be expanded and clarified because the nature of the data and how they were obtained are not clear. Discussions and conclusions should be more inherent to the results obtained, expanding some (unexpected or contradictory) result, as well as limitations.
Specific Comments:
Line 35. Tier 1 shortage. Please clarify, especially for non-U.S. readers, what is meant by Tier 1.
Line 64. Paragraph 2.1 data. Reading the paragraph, it is not clear the timescale, i.e. whether daily or monthly data were used. Please indicate this explicitly.
Line 78. Figure 1. Please improve this map. For example, the legend is too small, and the contents cannot be read. The authors can insert the north bar and scale, Colorado river network, and in the upper left corner could be inserted a box for this river basin location in the U.S.
Line 83. Paragraph 2.1.2. As above for paragraph 2.1, please specify if daily or monthly data were used. Regarding climate models, nothing is stated about the spatial resolution of these models, or if these are downscaled, please update. Again, regarding climate models and the results presented in Table 2 by each station, it would appear the climate model data were assigned to the corresponding station overlapping the cell of these data, is that correct? For instance, reading Gila river, does the data from climate models include the whole subbasin or only the area (cell) including the measurement station? This is important because in the manuscript is stated “…flows for each station”, which requires considering the inflows/outflows of the entire subbasin considered for correct streamflow evaluation.
Since there are different techniques for streamflow forecasting (e.g., time series, artificial neural network, AI models), the authors should better explain how the data were processed/modelled (procedures, accuracy, etc.) and how data is assigned to each of the 16 stations/subbasins.
Line 192. Figures 2 and 3. As for figure 1, the legend is too small. In addition, the model’s RCP boxes should be better placed.
Line 284. Paragraph 4. I suggest expanding the discussions (and presenting the paragraph as Discussion and conclusion), by including critical elements and limitations of the study, for example, those related to the nature and resolution of climate models, or how streamflow forecasting is obtained (i.e., different models), etc.
Reviewer 2 Report
See pdf version also attached
This paper compares droughts and surpluses in the Colorado River Basin for historical data with the projected droughts and surpluses resulting from climate change model scenarios. The comparison is in terms of number, length and severity of the drought and surpluses. Four climate models from CMIP5 with varying wet/dry leanings are used, each for both RCP4.5 and RCP8.5. A strength of this paper is that looking at both surpluses and droughts is unusual in the CRB literature and the results are interesting and compared with other studies.
The references are appropriate and cover recent and important findings from the past ~20 years.
Major weaknesses that must be addressed:
1. A research gap is not identified; this should be added in the initial section.
2. All stations except Lees Ferry are irrelevant to this analysis. The headwater flows all end up at LF and represent a tiny portion of the total; and the Arizona stations do not contribute to the allocated portion of the CRB, so would not be of interest to the managers of the CRB in looking at future droughts and surpluses. (But if you choose to leave them in, either explain their importance or emphasize that these other stations do not contribute to the research question; also identify the names with the stations on the map and explain absence of Gila R data)
3. Confirm the historical period used – is it 70 years as claimed in the data section? Or entire historical record as shown in Fig 5? Explain reason for how it is defined. Mention the paleo record and note in introduction and conclusions that the study could be enhance by using the paleo.
4. Add a methodology section that clearly explains what was done and how. This goes after data section and before results. Exactly how were the statistics for drought and surplus computed? What avg flow values were used? How was the F-test used and why? What are the ensembles – how calculated? How were the box plots generated? Exactly how were the statistics for drought and surplus computed? Please provide the details.
5. Is there a rational for the definition of drought and surplus that is based on 2 years? What was the justification in the earlier study that was not in CRB? Does the fact that CRB UB has 4 avg flow years of storage affect this rational?
6. Strengthen the justification for the CC models used – previously justified for California. What evidence is there that they are appropriate for CRB? Also, these models are known to be wet, dry, etc. So, in results, point out where the models behave as expected, and where they have unexpected results that need to be explained. (Currently, some of the obvious/expected behavior of these models is noted as a result of analysis, i.e., a “discovery”.)
7. Clarify and correct the language throughout and make all statements in the text consistent with the figures. Many specific examples are listed below.
8. The discussion/conclusion section needs strengthening. Summarize the novelty and findings of this study. The findings are summarized well in the abstract, but should be also in the Conclusions. Good to mention studies that agree, but remove information about studies that are not relevant to this one.
Comments about specific text or figures
Note: quotes from the paper are in black. Review comments in red. Some of these are addressed also in the items above.
9. 60-62 “observations and conclusions are drawn on changes in extremes (both droughts and surpluses) as these contribute (or deplete) a large portion of the water supply for the basin.
Only 1 – Lees Ferry - contributes /depletes a large portion of the water supply for the basin. (This is based on area, since flows are not provided.) The Salt/Gila stations supply Arizona only and not as part of the Colorado River allocation.
Hist Streamflow
10. Line 69-70 Unimpaired streamflow stations are needed to conduct this analysis and remove any anthropogenetic effects
But except for LF, they do not reflect the actual runoff in the basin. Also, they do not “remove” but rather avoid anthropogenetic effects. Also, should be “anthropogenic” rather than “anthropogenetic”
Most of the stations in UB are at the divide – so they don’t reflect a large area. Only LF reflects a large area. In LB, the Gila and Salt do not provide water to the LB water users, so are not relevant.
Years 1951-2021 is only half of the historical record. It does not cover the early 20th century pluvial period. How/why is the historical period so short?
11. Line 79 “Table 1 and Figure 1 summarize all the stations used in this study. Water year streamflow data were shown 79 as the million acre-ft…” But there is no streamflow data in Table 1 or Fig 1. Where were they shown?
The stations should be identified so later when discussed the reader can locate them on the map. Please add this, perhaps by giving a letter reference and a table that relates the letters to the station names.
The Reclamation Natural Flow database is maintained for 29 gauges in the basin. (Some of these 16 USGS stations could be used instead of the tiny headwater stations.)
What about the paleo record? Some mention, how important it is, and why not considered here, since it is so widely known and used in the CRB.
2.1.1 Future Projections
12. 106-107 “For this study there is no assumption about the model bias and direct reference is made to the models that best represent climate of the region.”
Where is the reference made? The above paragraph implies that the models were selected based on southwest/California scales. To what extent are the models appropriate for the stations in the CRB?
13. 107-108 Ensembles are calculated based on the average of the four models for each RCP. Describe the ensembles and how they are calcualated and how they are used to produce the results.
Will the calculations of ensembles be explained?
2.2 Definition of drought and surplus
14. drought is 2 or more years of below average and ends only when 2 or more years of surplus. How is average computed? Are same averages used for historical and for future years? Do future hydrologies show a trend?
Why is this 2-year criterion a good definition for the CRB? (or why not?) how do others define it? what is normal variability? CRB has 4 years of storage in system, i.e., is designed for variability.
2.3 testing of differences
15. Add a methodology section here
Please explain what differences are being tested. What is meant by “sample populations”? What is the F-test? Why compare variance of historical data to 8 future projections for each station, rather than simply the number of droughts or surpluses? What is exact methodology?
3. Results
It would be convenient to know where on the map the various stations are located (see comments above).
16. Line 131 “for the evaluation of drought (surplus) durations and magnitudes… this should not show (surplus since the paragraph is about droughts only.
17. 3.1.3 Changes in Water Year Flow Please explain the changes being discussed:
18. 169-170 “how the water year flows change” change from what to what?
19. 170-172 “Table 4 presents the historical water year average flows for each station and the changes that would occur under each of the scenarios.”
The numbers in the Historical column of Table 3 do not seem to be water year average flows. Lees Ferry is 52.3 MAF, which is about four times the annual average. Also, why are “historical average flows” shown with a negative number? Lines 172-173 refer to the numbers in the scenarios columns to be “the changes that would occur under each of the scenarios” What is changing – is it the average flow? How could one of scenarios for drought be -123.4 MAF different than the historical flow that is listed as -52.3 MAF? What are the historical average flows different for the Drought and Surplus cases?
172-178 comparison of result of the various models/RCPs does not seem to provide information about future flows (droughts or surpluses), but rather provides information about these climate models. Similarly, for the results lines 153-161. These results seem fairly consistent with the characterization of the 4 models in lines 102-105.
20. 184-186 “For instance, both RCPs have similar changes in the ensemble water year mean - six (6) stations are lower (drier) and 11 are higher (wetter).”
What is the “ensemble water year mean”? Although there is a reference to ensembles being calculated in line 107, no explanation of methodology regarding ensembles has been provided. Here there is no explanation of the meaning of “ensemble water year mean”.
21. 186-187 “Of note, is the approximate 3% (from 14.7 MAF to 15.1 MAF) increase in water year flow at Lees Ferry”
There is not a reference to Table 4 here, so the reader is still looking at Table 3. Please clarify.
Caption for Table 4, is first mention of what “ensemble mean” might be referring to. Please add Methodology section and explain in advance.
22. 187-189 “It is important to note that this is an ensemble from two models that produce higher averages (on the order of 17 MAF) and two models that produce lower average (on the order of 12-13 189 MAF).”
Without a Methodology section, it is not possible for the reader to understand this sentence. Please add this section to the paper and clearly explain your analysis and the rational for the calculations in advance of the results section.
The very close RCP 4.5 and 8.5 results are unexpected, particularly the lower avg flows for two of the RECP 8.5 scenarios when compared to their 4.5 counterparts.
23. 192-193 “Figures 2 and 3 present the spatial changes in the basin and the magnitude of the water year flows. There is a tendency for the headwater basins to show increases in water year flows in most of the model scenarios”
24. 199-200 …tendency for wetter conditions in CanESM2 and CNRM-CM5 models and drier in the Had-GEM2 and MI-ROC5 models (consistent for both RCP 4.5 and 8.5). These seem consistent with the characterizations of the CC models in lines 102-105.Why is this information useful for the basin?
25. Spatial differences: the headwater gauges are not independent since the headwaters end up at Lees Ferry. The Gila and Salt gauges are in Arizona rivers that do not contribute to the allocation of the CRB. There has not been a case made for the significance of any of the stations other than Lees Ferry in this analysis.
26. Figure 4. How are the box plots calculated?
27. 253-254 “It is noteworthy that the changes in surplus and deficit are more extreme under 8.5 scenarios” This sentence is not useful or appropriate in a research paper. Why is it notable? What insight does this provide?
28. Figure 5 Time series of historical and future projections at Lees Ferry for all scenarios. This plot is inconsistent with the historical period of 1951 to 2021 (lines 74-75). This calls into question the numbers presented thus far in the study – are they based on the entire period of historic record shown in Fig 5, or on just the 70 years stated in the data section?
29. Figure 6. Same question – what is the historical period represented by this data?
30. 272-273 “Under future conditions, maximum drought and surplus severity increases to levels of -6.4 to 14.0 MAF/year.” The plot seems to show values of severity from about -8 to over 9 MAF/year.
Discussion
31. Should have “Conclusions”, not just “Discussion” This section should state the major findings of the study and why they are unique and important. There is no mention of the novelty of this study. Please summarize in this section the novelty of the study and the main findings.
32. That the results conform to a California study is not useful unless there is information about the California study that demonstrates that it would or should be similar.
33. 289-291 “Other studies focused on ensemble projections have shown the potential for declines at Lees Ferry anywhere from 13.8 MAF (Woodhouse et al., 2021) and -10% for the water year [11]. This sentence does not make sense. Please clarify. How does this compare with this study?
34. 291-296 “It is noteworthy that some of these studies [11] have shown the change in seasonal runoff with higher streamflow (10 – 20%) occurring during the December – March period and less during the critical April – July period. It should be noted that these projections use a large number of models (32) from CMIP5 and the study presented here focuses on four models that are applicable to the western United States.
This appears to be completely irrelevant to this study and should be deleted unless the connection and relevance can be established. There is nothing in this study that addresses intra-annual variability for which this would be relevant. Noting the number of models used for this comparison is further of no interest.
35. 297-302 “Various studies have confirmed the wet signal identified in the research presented here. Hoerling et al. [8] noted an increase in the median precipitation from CMIP5 projections for both RCP 4.5 and 8.5 precipitation over Colorado by the middle of the 21st century. Ayers et al. [19] showed higher precipitation projections in CMIP5 (compared to 300 CMIP3) along with streamflow for the Upper Colorado River basin. Lukas et al. [20] also noted this potential for higher streamflow along with an earlier snowmelt”
This is a good outcome and comparison. Suggest to first summarize, noting the “wet signal” identified in this research, then mention these other similar findings.
36. 306-307 “These may not necessarily be extended dry or wet periods that were seen in the historical (or paleo) record.”
It seems that this study shows that there are likely to be extreme or extended dry or wet periods in the future, similar to the historic. If so, the last sentence is true and important!
Comments for author File: Comments.pdf
Reviewer 3 Report
Although the manuscript is well written and structured, I suggest rejecting it , because authors face a complex issue as change in streamflow trend in a river under global warming scenarios is so elementar and simplified, that it is very difficult to find any worth in their proposal.
Authors have taken some data from available datasets, filled some tables, drawn some figures and found some contradictory results for the different scenarios examined that they have in any way not attempted to explain or to motivate.
In my opinion approaches as those proposed by the authors should be discouraged, because they are fundamentally wrong. Data from CGMs should be carefully taken since they are affected by a number of uncertainties as structural, parametric or due to non linear dependence on initial conditions. GCMs use coarse spatial resolution and need to be downscaling. Precipitations, and as a consequence streamflows, are generally biased, and the bias correction methods are not exempt from criticism.
