the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Oct 2024
Groundwater head responses to droughts across Germany
Abstract. Groundwater is a crucial resource for society and the environment, e.g. for drinking water supply and dry-weather stream flows. The recent severe drought in Europe (2018–2020) has demonstrated that these services could be jeopardized by ongoing global warming and the associated increase in the frequency and duration of hydroclimatic extremes such as droughts. To assess the effects of meteorological variability on groundwater heads throughout Germany, we systematically analyzed the response of groundwater heads at 6,626 wells over a period of 30 years. We characterized and clustered groundwater head responses, quantified response time scales, and linked the identified patterns to spatial controls such as land cover and topography using machine learning. We identified eight distinct clusters of groundwater responses with emerging regional patterns. Meteorological variations explained about 50 % of the groundwater head variations, with response time scales ranging from a few months to several years between clusters. The differences in groundwater head responses between the regions could be attributed to regional meteorological variations, while the differences within the regions depended on local landscape controls. Here, the depth to groundwater best explained the time scale of the observed head response, with shorter response times in shallower groundwater. Two of the clusters showed consistent long-term trends that exceeded meteorological controls and could be attributed to anthropogenic impacts. Our study contributes to a better understanding of the regional controls of groundwater head dynamics and to the classification of groundwater vulnerability to hydroclimatic extremes.
- Preprint
(1441 KB) - Metadata XML
-
Supplement
(1154 KB) - BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2024-2761', Anonymous Referee #1, 17 Nov 2024
reply
This study employs a large-sample analysis of wells with high spatial variability in groundwater head responses to meteorological anomalies in Germany. These wells were grouped by similarity in groundwater head into six regional clusters, three meteorologically distinct regions, two response time scales, and two countrywide clusters. The results show that the median response time of the meteorological accumulation times ranged from a few months to several years for systems with longer memory, showing a close link to the frequency, duration, and severity of groundwater droughts. This is an important research that did a good job in classifying and analyzing this large data set. While this work approach clearly originates from the “Large-sample data-driven analyses” as clearly stated at the beginning of the paper (Line 57), it still aims to provide insight to hydrologists and policymakers as summarized at the end of the paper: “Overall, this study increased the understanding of dynamic groundwater responses to droughts and their different regional and local controls and derived vulnerability classes within Germany. The distinct responses to meteorological drivers reveal different implications to be expected under climate change. These insights can inform policymakers, water resource managers, and stakeholders for developing effective strategies for mitigating the impacts of droughts on groundwater systems and ensuring sustainable water management practices.”
This is highly appreciated as these communities can provide insight into the speculated mechanistic reasonings arising from the data in this work, which can be related to hydrological aspects (groundwater response time relation to droughts climate events) and urban anthropogenic drivers (the decline in wells usage in Berlin). However, this is also the weak part of this study. Currently, the methods and results section is written in a way that mainly caters to the hydro-informatics community, while the Implications, Conclusions, and part of the Discussion section aim to include hydrologists and policymakers, in various levels of success. This is not a style difference, but an approach that heavily relies on the data analysis and categorization of it instead of using this analysis to provide and support the general conclusion arising from the data, a conclusion that is definitely there and is relevant to these communities. In the following, I will provide three bullets that exemplify this:
- In the paper, there is a constant reference to the cluster instead of referring to their mechanistic interpretation, which prevents the reader from a clear understanding of how the observation is related to a specific mechanism. For example, lines 564-581 deal with why groundwater heads are increasing in urban areas and how this change is apparent in cluster lt_inc. However, why this cluster is indicative of the water level increase is not stated clearly. While it is clear that this is data-driven research coming from a hydro informatics standpoint, the relevance to the hydrology community resides in drawing the mechanistic aspects between markers, like the clusters, to the processes, like the water level, and the reason for the observed change and correlation.
- The following statement in line 593: “This study indicated that there is a large spatial variability in groundwater response time scales to meteorological forcing, even within the same region. This implies different vulnerability to the different types of driving meteorological drought events, i.e. meteorological extremes with respect to different time scales represented by different accumulation times.” It is indeed important and should be of interest to both the hydrological community and policymakers, yet the source of variability is not clearly presented in the paper in a way that is accessible to these communities. At the moment, it requires meticulous effort to follow the cluster names, indexes, acronyms, etc. The conclusions and implications are indeed important to these communities, making it very suitable to HESS, yet an effort should be made to “talk” to these communities. To do so, add to the names, indexes, and acronyms the hydrological aspects that they represent within each relevant part. This way, readers from this community can easily trace the relevant aspects of their community and appreciate the result’s relevance without being bogged down by the details.
- I find section 4.3 to be extremely important. This section draws its importance from the conclusions driven by the data analysis, yet understanding how these conclusions relate to the data in this work is really hard to deduce. For example, in line 627: “These locations with short-term memories could potentially be more strongly affected by increasing hydroclimatic seasonality. As these systems are often located in proximity to streams, this could moreover result in more losing or intermittent streams, as mentioned above.” Where is the role of losing streams in the context of hydroclimatic seasonality mentioned? The closest relation to the data I found was in line 344: “The distance to streams of fourth order (sd_order_4) was ranked as the sixth most important and significant feature (whole range of importances >1; see Fig. S6 panel a and b), followed by the second order stream distance (sd_order_2) in the resptSPEI and accSPEI RF models.”, where the SPEI is for the meteorological anomalies. The need to trace each acronym and parameter is cumbersome, and it worsens when the terms vary through the manuscript (hydroclimatic seasonality or meteorological anomalies?). I’m sure this is the standard presentation for hydroinformatic, but this is not true for other fields. I suggest that the terms will be uniform throughout the manuscript. In addition to parameters and model, a two-word description can be explicitly added where needed, like the above sentence. There are more than 30 terms in this study, and tracking the meaning of all of them is next to impossible in a first, second, or even third read. As such, this should be clearer and more approachable for the reader.
That being said, this is still excellent work that is relevant and highly suitable to HESS. The fact that HESS aims to reach a broad audience from various communities only makes a stronger case for this work to be published there, but it should be altered to be more approachable to these communities. The following has some specific comments:
- In line 85, it is stated that:” The groundwater head data used in this study are monthly mean groundwater head time series (from originally daily to monthly observations), aggregated.” It is unclear why monthly if you have daily data. If not all the data is daily and the aggregation allows to have a data set with the same time scale it should be stated here. If it is done to make the analysis easier, please comment on what we lose in terms of temporal resolution.
- Line 86: A reference and explanation for “CORRECTIV” is needed. I understand this is a German network, and it is referred to in the acknowledgments, but as the data originates from them, a more detailed explanation of the data and how it is collected is needed to appreciate the data quality.
- Line 90: Can't the meteorological data be used to "fill" the gaps instead of linearly interpolating it?
- Line 96: What are BGR and SGD? Are they references? Acronyms?
- Lines 102, 103: Km2 -> Km^2
- Line 104: It’s hard to understand which data or wells are important and which are irrelevant from the text. Can it be clarified?
- Line 147: Can you provide an equation for these SPI and SPEI? Generally speaking, each indicator comes from an equation representing a statistical analysis that has transformed into the jargon of a specific community. In an effort to be approachable to more communities, jargon should be reduced or better defined.
- Line 165: Is it possible to elaborate on the method in the supplementary?
- Lines 195-201: What are the advantages of ML that can't be achieved with statistical analysis and Bayesian statistics?
- Line 254: The last two paragraphs should be presented or refer to a table in the paper. At the moment, the details are hard to follow and rank.
- Table 3: What is the acceptable R^2? Is there a meaning to R^2<0.5? I know that in the hydro-informatic communities, this presentation is “standard,” but for a hydrologist, an R^2<0.7 is already questionable.
- Line 489: “Note the positive (although not strong) correlation between the mean gwdepth and topographic variables: for the elevation r=0.20, for the slope r=0.43 or the twir=0.38 (Spearman correlation).” Note where? is it in a figure?
- Line 602, 479, 477 : Fig. appears twice.
- Line 314, 315: Add a space between 4 to panel.
Citation: https://doi.org/10.5194/egusphere-2024-2761-RC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 249 | 68 | 10 | 327 | 36 | 4 | 5 |
- HTML: 249
- PDF: 68
- XML: 10
- Total: 327
- Supplement: 36
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1