Álamos et al. – The influence of human activities on streamflow reductions during the megadrought in Central Chile

General comments

In this manuscript, the authors investigate the influence of human activities in streamflow reductions before and during the megadrought in central Chile. Their approach is based on the comparison between near-natural streamflow simulations and actual observations. The former is estimated using a linear rainfall-runoff regression model trained on a period of low human-interventions identified according to water uses’ timeseries. The research question is very timely, given the ongoing drought in Chile. At the same time, the manuscript underscores the importance of understanding the role of human activities in either exacerbating or mitigating drought events – a crucial analysis given the increasing human influence in our catchments. I highly encourage the publication of the manuscript after revisions, as there are a few methodological aspects that need to be further specified and a few discussion points that could be further expanded.

On the methodological side, I really appreciate the considerations given to assess possible changes in catchment response driven by the persistent drought conditions, which could have introduced biases to the results. As well as, the thorough analysis for the selection of the low-human intervention period used for the training of the model. However, a few methodological aspects can be further specified. For instance, the authors used a linear regression model to capture the relationship between annual rainfall and runoff. However, relationship between rainfall and runoff are hardly linear and further information on the choice of this model would be beneficial. Further, the implication of the low performance of the winter rainfall-runoff model on the final results need to be discussed.  

Specific comments

Abstract: The abstract is informative, yet quite lengthy and could benefit from concise rephrasing. The initial problem statement (lines 13 -17) and methodology (lines 19-24), spanning several lines, could be condensed to give more emphasis on the results of the study. For instance, the problem statement could be introduced as following: "Since 2010, central Chile has experienced a protracted megadrought. Intensification of the drought has been attributed to the effect of cumulative precipitation deficits linked to catchment memory. Yet, the influence of water extractions on drought intensification is still unclear." Additionally, in line 26, the term 'undisturbed period' is introduced for the first time. For improving clarity, I would suggest to explicitly write that you are referring to a period with low human intervention.

Lines 43-46: this paragraph is a bit disconnected from the first and third paragraphs. I recommend merging it with the first paragraph to enhance overall flow. You could make the text more concise by eliminating the reference to climate change (as it is not the focus of your analysis) and starting to introduce the concept of hydrological drought soon after the first sentence. Or you could connect the first and second paragraphs by introducing a transitional statement like: These alterations in the water cycle may have implications for drought occurrences. If meteorological droughts (..) are mainly controlled by regional climate, hydrological droughts (..) are also influenced by..

Line 53: the word “dominated” is not very appropriate and could be replaced with “driven”. Also, I think you mean: “anthropic water uses”;

Line 57-58 is quite confusing. What do you mean with ‘adaptation and mitigation water management plans’?

Lines 67-69: The sentence is quite disconnected from the previous paragraph: above you mentioned the hydrological drought and water scarcity problem in central Chile, while later you give an example of the possible influence of human activities on the propagation of drought. Or you delete: "for example" or you already introduce in the previous paragraph the potential impact that human activities may have on the intensification and propagation of drought;

Line 67: I would suggest providing the geographical location of the Petorca river basin (as you do for the Aculeo Lake on line 70).

Lines 88-89: It is not very clear: do you mean that annual precipitation concentrates usually during winter (then delete ‘a few’) or that most of the precipitation during the last TOT years occurred only during a few winters or something else?

Lines 96-97: While knowing the actual extent of agricultural land cover is valuable, grasping the influence of these areas at catchment scale becomes challenging without information on the catchment size. I would suggest providing the percentage of the total catchment area covered by agriculture areas. Similarly, it would be good to know the annual water consumption in relation to mean annual flow (particularly in cases of minimal streamflow variation throughout the year).

Figure 1: In Panel A, the proxy station (green diamond) in the Limari catchment (light blue) is not visible. Could this be due to the absence of fill-in gaps in the timeseries extracted from the outlet represented by the red diamond? In Panel C, the colors are quite hard to distinguish. The majority of the area is categorized as 'others,' making it less informative. The agriculture regions are difficult to see, with clearer concentrations observed mainly in Aconcagua and Limari. Moreover, the colors used for urban and mountain landcover are notably similar.

Line 108: Monthly mean streamflow? Same applies, for line110: monthly mean or sum precipitation?

Figure 2: In the text under step 2, the use of the plural in indicating the period of low anthropic activities might be confusing. It may appear as if the model was trained in multiple periods, whereas, in reality, only one continuous period of low water consumption was identified. It would be clearer to use 'in the period' instead. Same applies for line 129, where I would write ‘in a period’.

Line 136: First time you are using the term ‘anthropic variables’, I would continue using the term 'water uses’ for consistency.

Eq. 2: Is the rainfall-runoff model applied on an annual or monthly scale? It is good to specify in the text what 't' is. And to change the time scale of your time series do you use the average or the sum? I suggest providing this information in the text.

Eq. 2 and Eq. 3: You are considering a linear relationship between precipitation and runoff, however annual runoff data is generally skewed and not linearly correlated to precipitation. Several researchers, for example, used a linear rainfall-runoff relationship only after verifying the assumption of a normal distribution of rainfall values and subsequently transforming the runoff data with a Box-Cox transformation to ensure they follow a normal distribution (Saft et al., 2015: The influence of multiyear drought on the annual rainfall-runoff relationship: An Australian perspective, Avanzi et al., 2020: Climate elasticity of evapotranspiration shifts the water balance of Mediterranean climates during multi-year droughts). The decision to employ a linear regression model for capturing the rainfall-runoff relationship should be then justified in the text or explicitly mentioned as a limitation in the analysis within the discussion section. In line 167 -170, you mentioned that multiple regression equations have been tested, however you do not specify in the text which other regression equations have been used. Further clarification on this can provide valuable information.

Line 192: The choice of the threshold for detecting hydrological drought should be justified, since influencing the periods in which near-natural and observed streamflow are compared.

Figure 4: The terms streamflow and runoff in the caption and figure have been used interchangeably, although different. Is the blue line indicating runoff or annual streamflow? The same occurs in Table 1 (line 234).

Table 2: I found it strange that the mean bias error for Equi and Aconcagua is 0% even though they have a low r-squared (under 70%). Could it be that this is because the model both over- and under-estimate during the training period and so the average bias converges to 0? And how does this low model performance for some catchments in winter affect the results in Figure 7? Could it be that the difference you capture between the near-natural streamflow and the observed streamflow is the error in the model estimation of the near-natural flow instead of the additional water stress resulted from water abstractions? Because with a low r-squared and high mean bias in the regression model for the winter season, it might be that you are capturing the bias in the near-natural streamflow more than the difference between the near-natural streamflow and the observed streamflow.

Further, in the Study Area section (2.1), you also mention that the winter season is the most important for water replenishment since the high concentration in precipitation in that season (Garreaud et al., 2017).

Additionally, you use annual time series on water uses of industry, energy, mining, livestock, and drinking water sectors. But information from the literature on how these uses change over a year (e.g., intensified in summer or winter) can better help interpret the results and might give relevance to a particular season over another.

Some reflections on the above points in the discussion section might be helpful.

Figure 6: I would move Figure 6 in section 3.2 since you already refer to it in that section. You can also think to merge this figure with table 2 (indeed in line 267-268 you mention the r-squared values, which however are not shown in the figure).

Figure 7: Precipitation and near-natural streamflow have very similar colors, I would differentiate the colors a bit more.

Section 4.1: Would it be possible to compare the results on the impacts of human activities on water availability together with the differences in trends of water users (Figure 5)? Water uses for agriculture and mining differ between catchments and over time with some peaks, lows, and some constant values. Perhaps reflecting on this could provide valuable information for local authorities trying to understand which water sector has the greatest impacts on drought intensification.

Lines 336-337: maybe good to briefly specify somewhere in the text that groundwater does not contribute to streamflow in the catchment under analysis (if this is the case).

Lines 357-372: There is a bit of disconnection between the previous paragraph and this one. Although, you start the paragraph by referring to the intensification of streamflow deficits, you then discuss about drought impacts (357-363) to then start to talk about human activities (Line 363). I would actually remove the part from ‘’Although..”(line 357) to “US$56 million during 2010-2020” (line 363) or move it below.

Grammar check: Line 56: ‘’to assess catchments’ vulnerability’’; Line 60: ‘’we focus on central Chile’’;  Line 71: ‘by using’ instead of ‘and used’ (to reduce the use of ‘end’); Line 96: land cover (without s); Line 176: ‘compared’; Line 199: ‘defined’ (consistent with the past tense form used for the method section).

Review of “The influence of human activities on streamflow reductions during the megadrought in Central Chile” by Alamos et al.

General comments

Alamos et al. study the impact of human water use on streamflow (Q) in four river basins in Central Chile, by comparing observed and naturalised Q for a period of high human influences encompassing the ongoing megadrought. They assess the impact of human water use on annual/seasonal Q and on streamflow droughts, by showing Q reductions and an intensification of streamflow droughts due to human water use, but a lower relative contribution of human influences on the reduction of seasonal/annual Q during the megadrought than before its onset. The topic is of high relevance and this work can provide a valuable contribution to the still limited literature on the matter, with specific regard to multi-year droughts. For these reasons, I see the paper as suitable for potential publication in HESS and the SI, it is submitted to. I appreciated the detailed description of methodological choices, the integration of water use data in the identification of high/low human influence periods, and the in-depth discussion of the results in the Chilean context. Yet, I believe that some methodological choices, the presentation of results and their implications beyond this exemplary case study for the broad scientific community should better explained prior to publication, as I detail in the following comments.

Specific comments:

• Section 3.1. Why did the authors not exploit water use data only to identify periods with low/high human influences? Provided good quality (some more details on them would be helpful anyway, given the common challenges around collecting this kind of data), I think these could give direct information on change points in the different basins due to human intervention.
• Section 3.2. I appreciated that the authors investigated also potential climate-driven non-stationarities on Q during the multi-year drought, given their evidence in Chilean catchments (Alvarez-Garreton et al., 2021) and their relevance for hydrological modelling (e.g., Saft et al., 2016). I assume runoff ratios in Table 3 were computed on naturalized Q, given the Section in which they appear. If so, how can we infer that human activities are responsible for the intensification of the Q reductions (L254-256)? Furthermore, are the upstream-downstream catchments fully comparable? Could you please provide more information on their catchment attributes? In some cases (e.g., Choapa), I wonder if additional factors - other than human influences only - may play a role in the different behaviour of the catchments, given the significant difference in catchment area. Why not computing shifts in rainfall-runoff relationship (e.g., Saft et al., 2015, Alvarez-Garreton et al., 2021) on naturalized Q for downstream basins? This could provide a clue of potential climate-driven non-stationarities in the basins during the megadrought. Furthermore, if such non-stationarities happen, how do you ensure that the linear model for the naturalization can properly reproduce Q? I recommend adding at least some discussion on this latter point. Concerning the choice of the naturalization approach, I also wonder why upstream-downstream comparison (e.g., Van Loon et al., 2022 and references therein) was not chosen, provided comparability of the upstream/downstream catchments.  
• Section 3.3. I would suggest to focus on a 10-year period, instead of 1988-2009, to ensure the same record length for the computation of anomalies before and during the megadrought.
• Section 3.4. The fact that most streamflow drought events after 2010 in Figure 8 appear mostly as human-induced may sound a bit contrasting the result of decreasing human contribution to Q reduction during the megadrought (Figure 7). I suggest showing in Table 4 the relative difference in drought characteristics due to human influences (e.g., Van Loon et al., 2022), before and during the megadrought to provide additional insights on this. An additional event-scale analysis on the role of human activities in aggravating/inducing streamflow droughts could also provide novel understandings on the role of human influences during prolonged meteorological droughts (L370-372).
• Why was the drought analysis performed at seasonal (6-month) scale and not monthly, as frequently done? A monthly scale is usually relevant for water management and monthly data are available here (L108). Furthermore, why not using near-natural simulated streamflow for the whole period for the identification of the drought threshold (L201)?
• While I appreciated the schemes to illustrate the methods, I believe figures could be more informative in general, and Figure 2 itself as well, since it does not fully convey the two main analyses on the effect of human activities on seasonal/annual Q and on streamflow droughts. To this end, I suggest for instance highlighting the megadrought period (here and elsewhere, to help in the readability of the time series as well). Moreover, Figure 5 is hard to read, due to the different ranges of variability for the time series. Maybe a log-scale or two different y scales could help? A common x-scale for the different subplots in Figure 6 would also facilitate comparison among them. I would also enjoy the same units in the figures and text (e.g., percentages in Figure 7, since anomalies are then presented in percentage at L279fff) to make figures easier to grasp, as well as the same y scales in case of multiple subplots. Finally, please make sure figures and tables are mentioned in the text in the same order as they appear, which is not the case for Figure 8 and Table 4 at the moment.

Technical corrections:

• L81, what “for the same period” refers to is not completely clear, since two periods are mentioned in the sentence before.
• L90, how are potential changes in glacier cover over time taken into account in the analysis? If they are not, a brief discussion on this point would be helpful.
• L111, where do basin polygons come from?
• L166, how is the uncertainty of the regression model quantified? Is it then used to derive the human influence later other than being visualized in Figure 6?
• L169-170, could you please add these results in an appendix? This could also add more confidence to the model.
• Figure 3, I would plot a variable drought threshold to make the figure more informative, since this kind of threshold is then actually used.
• L192, a sensitivity analysis on the influence of this threshold on the conclusions would be valuable, similarly to what done for the statistical tests to identify periods with high/low human influences.
• L271, how were anomalies computed? Please clarify in the Methods.
• L271, “observed and near-natural simulated streamflow” -> “observed and near-natural simulated SUMMER streamflow”?
• L273, “Appendix A (Fig. A1)” -> redundant cross-reference to the figure
• I suggest a careful check of the language throughout the manuscript to avoid inconsistencies in verb tenses and forms (e.g., both past and present in the same sentence, L287-288), and terminology (e.g., STARS/Rodionov test, anthropogenic/anthropic, streamflow/flow/runoff used rather interchangeably which generates some confusion), as well as potential typos, also in the legends of the figures (e.g., “this suggest”, L332, “treshold” in Figure 3).
• I think some re-organization among the sections would help the reading flow (e.g., I suggest moving L210-211 to the Methods and L212-214 to the Discussion, and so on).
• L393, please check the first link, since I got a “404 Error” when trying to access the website.

References:

Alvarez-Garreton, C., Boisier, J. P., Garreaud, R., Seibert, J., and Vis, M.: Progressive water deficits during multiyear droughts in basins with long hydrological memory in Chile, Hydrol. Earth Syst. Sci., 25, 429–446, https://doi.org/10.5194/hess-25-429-2021, 2021

Saft, M., A. W. Western, L. Zhang, M. C. Peel, and N. J. Potter (2015), The influence of multiyear drought on the annual rainfall-runoff relationship: An Australian perspective, Water Resour.Res.,51, 2444–2463, doi:10.1002/2014WR015348

Saft, M., M. C. Peel, A. W. Western, J.-M. Perraud, and L. Zhang (2016), Bias in streamflow projections due to climate-induced shifts in catchment response, Geophys. Res. Lett., 43, 1574–1581, doi:10.1002/2015GL067326

Van Loon, A. F., Rangecroft, S., Coxon, G., Werner, M., Wanders, N., Di Baldassarre, G., Tijdeman, E., Bosman, M., Gleeson, T., Nauditt, A., Aghakouchak, A., Breña-Naranjo, J. A., Cenobio-Cruz, O., Costa, A. C., Fendekova, M., Jewitt, G., Kingston, D. G., Loft, J., Mager, S. M., Mallakpour, I., Masih, I., Maureira-Cortés, H., Toth, E., Van Oel, P., Van Ogtrop, F., Verbist, K., Vidal, J. P., Wen, L., Yu, M., Yuan, X., Zhang, M., and Van Lanen, H. A. J.: Streamflow droughts aggravated by human activities despite management, Environ. Res. Lett., 17, 044059, https://doi.org/10.1088/1748-9326/ac5def, 2022.

General summary

The paper quantifies the impact of water extractions in 4 adjacent catchments in central Chile. The paper adopts the methodology proposed by van Loon et al. 2016 and Rangecroft et al. 2019. Key findings are that water extractions significantly exacerbate droughts in all 4 catchments and the size of the effect is likely due to the level of agricultural development in each catchment. Overall, the paper reads quite well and is suitable for publication in the special issue.

General comments

• Please provide a more thorough definition of hydrological drought in the introduction as per the methods adopted.
• It would be good to define rainfall/climate elasticity and streamflow a little better and perhaps include a reference or two. It will put more focus on why it is important to try and include close to natural periods/catchments to compare to. Water extractions are bound to affect hydrological drought but by how much will be influenced by how elastic or inelastic the relationship between rainfall and streamflow is.
• In your discussion, you make the point that extraction is inelastic, and you allude to why this can be the case in the conclusion i.e. farmers being issued with additional water rights. I think that it would be important to discuss this in a bit more depth, particularly around how a potential switch is being made toward groundwater. Importantly, discuss how the current policies might tip the system into the next drought category – agricultural drought.
• Please further clarify some of the consequences of maladaptation. Perhaps you can add thoughts to “there is still a gap in understanding how human activities contribute to basin's vulnerability to drought.”

Specific comments

• Line 52: Do you mean “no” significant rainfall deficits?
• Line 168: perhaps use , instead of – i.e. forms and variables, including evapotranspiration and temperature, were tested. This is also slightly vague as none of the equations list T and ET in their current form. Are you suggesting that all models ended up being only dependent on P?
• Lines 221 -228: Totally understand that much of the information that is provided is based on local experience and while some events are very defined such as the construction of a dam, it would be good to have references to point to changing land use in the region if it exists.
• Lines 293-296: Agricultural water demand will increase in drought – perhaps you can tease out this paradox a little more.
• Line 332: perhaps make clear that total consumption is the sum of surface and groundwater. This makes the link with lines 335-337 clearer.