| 31 Aug 2026
Attributing summer streamflow decline to climate vs. agriculture
Abstract. Climate warming is reducing the mountain snowpacks that sustain summer streamflow, yet in agricultural and urban catchments this climatic decline is often superimposed on the effects of local water use, and separating the two is difficult. We investigate how much of the continuing dry season-flow decline in a snow-reliant, anthropogenically influenced intermontane catchment with a distinct summer dry season is driven by climate change rather than human water use, and whether the roughly even contributions reported in another study for the last quarter of the 20th century still hold under U.S. Southwest megadrought conditions of the first quarter of the 21st century. Comparing three multi-decadal timeframes across nine catchments in Northern California, we find no significant difference in annual precipitation but a widespread rise in vapor pressure deficit and declining snowpack at lower elevations. Summer flows declined significantly in the catchment affected by human use of groundwater and surface water compared with the near-natural reference catchment. However, unchanged annual total flows indicate seasonal redistribution rather than an overall loss of water. Using a paired-catchment attribution and the mid-20th century record as pre-development and pre-climate change reference, we decompose the late 20th century and early 21st century decline in the managed catchment into climatic and anthropogenic components. The total summer-flow loss nearly doubled between the two climate change periods, from about 21 to 40 Mm³: During late 20th century, climate change and anthropogenic water use contributed almost equally. But under 21st century megadrought conditions the climatic loss (26 Mm³) had grown to roughly twice the anthropogenic impact (14 Mm³), accounting for about two-thirds of the decline. The even contribution no longer holds: climate change now governs the decline. The results suggest that curtailing withdrawals might not sufficiently sustain ecologically relevant late-summer flow and fall pulse flows and additional measures might require managed aquifer recharge, adaptation, and climate mitigation alongside local water management.
GENERAL COMMENTS:
Overall this paper will be a good contribution once the issues that I note below are addressed.
SPECIFIC COMMENTS:
Lines 16-18. "The results suggest that curtailing withdrawals might not sufficiently sustain ecologically relevant late-summer flow and fall pulse flows and additional measures might require managed aquifer recharge, adaptation, and climate mitigation alongside local water management." I don't think you've actually shown anything in this article about the potential effectiveness of curtailment because there no assessment of unimpaired flows and what the total impact of agriculture on flows is (not just the change in agricultural impact since the Historic period). Unless you add some more information and analyses to support this statement, you need to re-phrase it. For example, something like "Given these climate changes, sustaining agriculture and ecologically relevant late summer flow and fall pulse flows might require managed aquifer recharge to increase seasonal storage and emissions reductions to limit further warming".
Line 18. "adaptation" and "local water management" are general terms and it is not clear here what they mean.
Lines 14-15. "But under 21st century megadrought conditions the climatic loss (26Mm3) had grown to roughly twice the anthropogenic impact (14Mm3)" It would be better to use the same terms to describe the anthropogenic and climatic. For example, use "climatic loss" and "anthropogenic loss", or "anthropogenic "climatic impact" and "anthropogenic impact", or "anthropogenic "climatic contribution" and "anthropogenic contribution". The current phrasing of using different terms risks confusing readers (i.e., they might think that "anthropogenic impact" refers to the total effect of agriculture on flow whereas here it really is intended to mean the change relative to the historic period).
Lines 30-34. Consider specifically mentioning mid-winter melt. A good cite for that is Musselman et al. 2021 (https://doi.org/10.1038/s41558-021-01014-9)
Lines 35-37. Consider also specifically mentioning evapotranspiration of vegetation. Also, a great cite (wide geographic area, lots of data sources analyzed) for increased evaporative demand is Albano et al. 2022 (https: //doi.org/10.1175/JHM-D-21-0163.1), consider if that would be a good cite for somewhere in Intro or in Discussion.
Lines 35-39. I am glad to see attention to the increased evaporation here. While the snow-to-rain phase change generally gets a lot of attention in the climate literature and management responses, the changes to evaporation and evapotranspiration are generally underappreciated. [No change necessary, just a comment]
Lines 65-66. "Asarian and Walker (2016) attributed flow declines to human water use rather than precipitation..." That is somewhat of a mischaracterization of the findings of Asarian and Walker (2016). I suggest rephrasing that to something like "Asarian and Walker (2016) found that factors other than precipitation timing and quantity accounted for >75% of September flow declines, but did not quantify the relative contributions of increased water withdrawals and/or increased vegetation/ forest evapotranspiration." [Some key quotes from Asarian and Walker (2016) that support the rephrasing: "A spatial pattern is apparent for unregulated sites in the month of September, which had the most widespread streamflow declines, with factors other than precipitation accounting for over >75% of the streamflow decline at many sites in the Upper Klamath Basin and Upper Rogue Basin as well as the Scott River..." ... "The data and methods we used do not allow for quantification of the relative impacts of the various factors contributing to the declines in precipitation-adjusted streamflow, which include some combination of increased water withdrawals and/or increased vegetation/ f orest evapotranspiration." ... "By carefully examining the trends that have occurred over the study period in watersheds with contrasting conditions and histories, we can develop hypotheses about causal mechanisms that could be tested with additional analyses." ... "Our results appear to support the hypothesis that water withdrawals [at multiple gages, not specifically Scott] are an important factor, but not the only one, contributing to the declining trends in precipitation-adjusted streamflows." ... "In the Scott River, where precipitation-adjusted summer streamflow declined (Figure 8), reductions in base flows since the 1970s have been attributed to increased groundwater pumping and decreased snow accumulation (Van Kirk and Naman, 2008)."]
Line 82: should "its benefits" be "its effects"? Also, rephrase sentence to explain what "its" is?
Line 94. If you keep Trinity Hoopa in your analysis (see general comment above), you need to add it here to the list of sites with large reservoir storage. Trinity Reservoir is the 3rd largest reservoir in California, with volume of 3 cubic kilometers (https://en.wikipedia.org/wiki/List_of_largest_reservoirs_of_California). This reservoir has large effect on summer flows in Trinity River at Hoopa because 450 cfs (approx 10x higher than natural in August/Sept) is released into the Trinity River. There's no major dams in Upper Trinity or South Fork Trinity. There's a low amount of agriculture in the Trinity River watershed, but approx 50% of annual flow is diverted as an interbasin transfer to Central Valley, so it's highly regulated. Hoopa is near the bottom of the watershed, but the dam still has big effects, especially in summer. Flow management has changed a lot over decades also (see dam releases in figure 26 in the Asarian et al. 2023 Trinity report: https://www.researchgate.net/publication/373831644_Synthesizi ng_87_years_of_scientific_inquiry_into_Trinity_River_water_temperatures). In many years since 2003, the Trinity also has high dam releases in August and September to prevent salmon disease outbreaks in the Lower Klamath River (i.e., see section 3.4.6.4 in Asarian et al. 2023).
Line 119. Here you state "summer-flow period" whereas later in the manuscript you use the phrase "summer window".
Line 121. "...end of the low-flow season to year-end" I would recommend clarifying what is meant by year, because there's calendar years, hydrologic year, irrigation seasons, etc. Likely change to "...end of the low-flow season to December 31"? Also, see comments below on line 191 and lines 195-196 regarding standardizing terminology.
Line 124: "The Historic period runs from 1940, or the start of the record where data acquisition began later..." It would be good to cite Table A1 here for gages because it shows start date for gages. Also would be good to add the start date to the snowcourse data in Table B1 (changed to landscaped page if you need more space to fit it).
Figure 1. Minor point on gage naming, just something to consider, not required. I would recommend renaming a few sites to be more intuitive. "Trinity (Hoopa)" could be "Lower Trinity", "Trinity ( Hyampom)" could be "South Fork Trinity", and "Trinity (Center)" could be "Upper Trinity". For most readers, Hoopa, Hyampom, and [Trinity] Center will be very obscure place names. And I've never head people familiar with the Trinity watershed call the upper gage "Trinity (Center)".
Lines 138-139. "We ranked all candidate catchments by how closely their Historic summer regime (day of year 150–300) matched that of the Scott River..." It's slightly confusing to use the term "summer " whereas elsewhere it's called "late summer" in Figure 5. See other comments regarding standardizing terminology.
Line 146. Consider citing the updated version of Statistical Methods in Water Resources (Helsel et al. 2020 https://doi.org/10.3133/tm4A3 instead of the older Helsel and Hirsch 2002)
Lines 148-149. "...yields the Scott River flow expected under climate change alone, absent any change in agricultural use and specifically absent the onset of groundwater pumping at the beginning of the Intermediate era." I recommend re-phrasing the second half of this sentence because it could interpreted (by inattentive readers) to imply that groundwater is explicitly included in the analysis. For example, replacing "and specifically absent the onset of groundwater" with "such as the onset of groundwater" or "including the onset of groundwater" would make it more clear you are just comparing between periods.
Table 1. This table was not very intuitive to me and I had to read the descriptions carefully to make sense of it. Also, it doesn’t provide the magnitude of the changes. I think it would be more effective to summarize the data in a different way (i.e., focus on magnitude of differences between periods, not the probability that there’s a differences between periods, see general comment 1). Perhaps a multi-panel of boxplots (similar to Fig 2, but with the time series removed), sites as columns and the metrics (Precip, VPD, etc.) as rows. But that’s not required, just something to consider. I acknowledge that I have a bias towards figures over tables, and that other researchers may not share that perspective.
Table 1 caption. I'm not sure what this means: "April 1 SWE is the most significant snow course per catchment (always its lowest)." Can you clarify? Does it mean that the tests were run for multiple snow courses but that only the snow course with the largest changed is reported? Does "lowest" mean lowest elevation or lowest p-value?
Table 1. Should End and Post-Season Mean should be included here? Some of those results are mentioned in the text but not presented in tables or figures.
Lines 165-169. p<0.05 is an arbitrary cutoff, there isn't really any important difference between p=0.055 and p=0.05. See general comment on this topic.
Lines 171-174. Not required, but I think it would be a great addition to the paper to do some analyses between snow course elevation and SWE declines. For example, see Figure 4 of Van Kirk and Naman 2008, although I think you can even do something simpler (i.e., just a basic scatterplot with X-axis as elevation, Y-axis as % of Historic, with one panel for Intermediate and one panel for Modern). You haven't really clearly shown yet, in an easily understandeable way, that lower elevation snowpack sites have greater declines.
Figure 2. This figure is helpful, shows the actual time series. Since you have the formatting for this plot already, and could presumably easily (or perhaps even have already done so) plot the other gage-associated parameters (precip and the flow metrics), etc.), I think it would be good to add similar for all parameters and put in a Supporting Information. Also the snowpack data too.
Line 190. "No metric differed significantly between the Intermediate and Modern periods". That's surprising, given the data shown in figure 3A and 3B. See comments elsewhere regarding p-value thresholds.
Line 191. "mean discharge after the season" is inconsistent terminology (see comment below regarding line 195-196).
Lines 194-195. This is true (no changes necessary here), but contradicts the statement on lines 96-98 that the Trinity River (as part of the "other catchments" category) is "effectively unimpaired by direct anthropogenic influence other than a changing climate". See general comment above about the Trinity at Hoopa gage.
Lines 195-196. "The Shasta catchment showed a significant decline only in post-season discharge." At first I was confused about what "post-season discharge" was, but now I see that it is defined using a different phrase on line 120-121. This concept is also mentioned with different phrasing in line 191. It would be better to use consistent terminology.
Line 227. Should "...consistent with projections of shifts in precipitation timing rather than amount" instead be "consistent with projections of shifts in precipitation phase and timing rather than amount? You haven't done any analyses here of the timing of when precipitation occurs.
Lines 230-232. See comments above on lines 171-174. Table B1 doesn't clearly demonstrate that snowpack declines are greater at lower elevations. Can you make that more obvious and clear?
Figure 5. Is "late-summer" the best label for this? Days 150-300 are May 30 to Oct 27 (in a non-leap year). I don't have a great suggested alternative, perhaps "summer and early fall" or "low-flow " or "summer"? See other comments regarding terminology
Line 269. Consider citing Asarian et al. 2023 (https://doi.org/10.1029/2022WR032915) here showing that lower flow increases water temperatures in Scott River.
Lines 278-279. "Our results suggest a structural limit to how much this can achieve for late-summer flow." Actually, I don't think you have shown this in the paper. You've shown that climate has been the biggest driver of recent flow declines, but you haven't done any comparisons here of the total impact of agriculture on flows (i.e., you haven't also included the effects that were already present in Historic period). Somewhere in the paper, likely Discussion (but perhaps in Intro or Methods), I think you should present some information comparing measured flows to unimpaired flows. Without this context, readers can't really understand what the management options are. There is information on this in the scenari os in Siskiyou County Flood Control and Water District Groundwater Sustainability Agency (2021). See related comments elsewhere.
Lines 282-283. "Measures that add water rather than withhold it". I think this phrase should be revised to be more clear, i.e., what is the activity or geography where the water is being withheld or added? Should it be "Measures that add water rather than limiting water extraction"? or "Measures that add water rather than limiting pumping and diversions"?
Lines 300-302. "Even minor influences on the Scott not shared with the Salmon—changes in forest cover and evapotranspiration, wildfire, diversions, or shifts in the gauge rating—is folded into the agricultural estimate and may inflate it." From an a-priori perspective, it's far from clear that these changes would necessarily inflate the agriculture estimate, it seems equally likely that it could reduce it, right (perhaps I'm misunderstanding)? That said, the key real-world landscape changes (beyond agriculture) that have occurred differently in Scott and Salmon since 1977 is wildfire. The vast majority of Salmon River watershed has burned at least once since 1977, whereas only a small percentage of the Scott has burned. Unfortunately, I don't have a citation to suggest for already-computed watershed statistics. However, even a brief review of fire history maps shows this, so I think you could probably cite the Calfire historic fire perimeter datasets (i.e., see map viewer at htt ps://experience.arcgis.com/experience/b72eede32897423683a94c61bf9d3027). There is abundance evidence in the literature that wildfires increase streamflow, for example see Boisramé et al. 2019 (https://doi.org/10.1029/2018WR024098) and Williams et al. 2022 (https://doi.org/10.1073/pnas.2114069119). My summary recommendation on this topic is that I think you need to either change "inflate" to "inflate or reduce", or provide some evidence (e.g., cite fire history) about why you think it’s more likely to inflate rather than reduce.
Lines 304-305. "The reference catchment itself is imperfect: its vapor pressure deficit did not rise significantly, hinting that it may warm slightly less than the Scott." I think this illustrates the problem with relying on arbitrary cutoffs for "significance". Looking at the time series in Figure 2, I don't really see any obvious difference between Salmon (panel E) and others (i.e., Trinity Hoopa panel H) where p-values are <0.05.
Lines 327-336 (and especially lines 327-33). I think is the right way to frame the issue of the causes and solutions for flow declines. It's not that curtailing agriculture can't necessarily get flows back to what they were in the Historic period (this paper hasn't presented any analyses of whether or not that's possible), it's that that would require cutting in the Historic use. No changes necessary here, I'm just noting this to contrast it with how this is discussed elsewhere in the paper.
Line 333-334. "...and ultimately with mitigation of the warming that increasingly governs the decline." Would it be better to state this in a less jargony way, e.g., "...and ultimately reducing the emissions that cause the warming and increasingly govern the decline"?
Line 439-440. Is there a specific reason to cite the 2021 public draft instead of the final version? If not, it would be better to cite the final version (2022?).