the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Mapping groundwater dependent ecosystems using a high-resolution global groundwater model
Abstract. Global population growth, economic growth, and climate change have led to a decline in groundwater resources, which are essential for sustaining groundwater dependent ecosystems (GDEs). To understand their spatial and temporal dependency on groundwater, we developed a framework for mapping GDEs at a large scale, using results from a high-resolution global groundwater model. To evaluate the proposed framework, we focus on the Australian continent because of the abundance of groundwater depth observations and the presence of a GDE atlas. We first classify GDEs into three categories: aquatic (rivers and lakes), wetlands (inland wetlands), and terrestrial (phreatophyte) GDEs. We then define a set of rules for identifying these different ecosystems, which are based, among others, on groundwater levels, and groundwater discharge. We run the groundwater model in both steady state and transient mode (period of 1979–2019) and apply the set of rules to map the different types of GDEs using model outputs. For steady-state, GDEs are mapped based on presence or absence, and results are evaluated against the Australian GDE atlas using a hit rate, false alarm, and critical success index. Results show a hit rate above 80 % for each of the three GDE types. From transient runs, we analyse the changes in groundwater dependency between two time periods, 1979–1999 and 1999–2019 and observe a decline in the average number of months that GDEs depend on groundwater resources, pointing at an increasing threat to these ecosystems. The proposed framework and methodology provide a basis for analysing how global impacts of climate change and water use may affect GDEs extent and health.
- Preprint
(1153 KB) - Metadata XML
-
Supplement
(729 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-112', Tom Gleeson, 08 May 2024
Overall, I think this is an interesting, important and worthwhile manuscript. I appreciate the purpose (improved temporal-resolution modeling of different GDE types) as well as the method (coupled hydrologic model) and geographic focus (Australia where there is good data), and the results seem reasonable. I have a few critiques of the methods that I think would improve the manuscript.
I hate suggesting to include a few contributions that I have been a part of but i can't see anyway around this. This is a recent overview of groundwater and ecosystems that could provide more background on terminology and processes: Gleeson, T., Huggins, X., Vázquez Suñé, E.. Arrojo-Agudo, P., Connor, R. (2022) Groundwater and Ecosystems. Chapter 6 of Groundwater: Making the invisible visible, UNESCO World Water Development Report.
I would be good to at least mention that you are not covering subsurface ecosystems.We took a similar approach to mapping terrestrial and aquatic ecosystems in this: Huggins, X., Gleeson, T., Serrano, D., Zipper, S., Jehn, F., Rohde, M.M., Abell, R., Vigerstol, K., Hartmann, A. (2023) Overlooked risks and opportunities in groundwatersheds of the world’s protected areas. Nature Sustainability.
We used an inference based terrestrial and aquatic inference-based approach to map terrestrial GDEs, lentic aquatic GDEs and lotic aquatic GDEs.Based on these, my most significant critique is dividing GDEs in lentic (non-flowing; lakes/wetlands) vs. lotic (flowing; rivers, streams) rather than aquatic vs wetland. Ecologists often differentiate this way since they function very differently.
This critique leads to my next concern: defining aquatic ecosystem dependency as the ratio to groundwater discharge to streamflow. This might be appropriate for rivers or streams but is inappropriate for lakes which may have very little flow but still may groundwater dependent.
Two more methodological concerns:
5 m is deep water table as minimum for wetlands seems too deep and needs more justification to me...
I was also surprised that the difference between known and likely GDEs was not distinguished. I would test the importance of this assumption.Citation: https://doi.org/10.5194/hess-2024-112-RC1 -
AC1: 'Reply on RC1', Nicole Gyakowah Otoo, 08 Nov 2024
Overall, I think this is an interesting, important and worthwhile manuscript. I appreciate the purpose (improved temporal-resolution modeling of different GDE types) as well as the method (coupled hydrologic model) and geographic focus (Australia where there is good data), and the results seem reasonable. I have a few critiques of the methods that I think would improve the manuscript.
We thank Reviewer 1 for his kind remarks about our paper.
I hate suggesting to include a few contributions that I have been a part of but i can't see anyway around this. This is a recent overview of groundwater and ecosystems that could provide more background on terminology and processes: Gleeson, T., Huggins, X., Vázquez Suñé, E.. Arrojo-Agudo, P., Connor, R. (2022) Groundwater and Ecosystems. Chapter 6 of Groundwater: Making the invisible visible, UNESCO World Water Development Report.
Thanks for this useful reference. We have duly referenced it in the manuscript.
It would be good to at least mention that you are not covering subsurface ecosystems.
We took a similar approach to mapping terrestrial and aquatic ecosystems in this: Huggins, X., Gleeson, T., Serrano, D., Zipper, S., Jehn, F., Rohde, M.M., Abell, R., Vigerstol, K., Hartmann, A. (2023) Overlooked risks and opportunities in groundwatersheds of the world’s protected areas. Nature Sustainability.
We mention subsurface GDEs in our introduction and will add a justification for not including them in our revised manuscript. We have also included the suggested reference.
We used an inference based terrestrial and aquatic inference-based approach to map terrestrial GDEs, lentic aquatic GDEs and lotic aquatic GDEs.Based on these, my most significant critique is dividing GDEs in lentic (non-flowing; lakes/wetlands) vs. lotic (flowing; rivers, streams) rather than aquatic vs wetland. Ecologists often differentiate this way since they function very differently.
We acknowledge that this would be a useful further subdivision of aquatic ecosystems, especially from an ecological perspective. There are of course several ways of dividing ecosystems into different classes. The division that we chose puts the groundwater-dependent part first, following Foster et al. (2006) and also fits the Australian GDE atlas (Doody et al. 2017). In this sense, both lakes and streams belong to aquatic ecosystems. Wetlands ecosystems are mapped based on permanent soil saturation (‘Wetlands’) and shallow groundwater depth (groundwater dependent), whereas ‘Terrestrial groundwater ecosystems’ are mapped based on root interactions with groundwater depth.
References
Foster S, Koundouri P, Tuinhof A, Kemper K, Nanni M and Garduño N 2006 Groundwater Dependent Ecosystems the Challenge of Balanced Assessment and Adequate Conservation Briefing Note Series 15 (Washington DC: GW·MATE/Worldbank)
Doody, T. M., Barron, O. V., Dowsley, K., Emelyanova, I., Fawcett, J., Overton, I. C., ... & Warren, G. (2017). Continental mapping of groundwater dependent ecosystems: A methodological framework to integrate diverse data and expert opinion. Journal of Hydrology: Regional Studies, 10, 61-81.This critique leads to my next concern: defining aquatic ecosystem dependency as the ratio to groundwater discharge to streamflow. This might be appropriate for rivers or streams but is inappropriate for lakes which may have very little flow but still may groundwater dependent
In our modelling approach, lakes do have flows and, due to their position (mostly in convergence zones and alluvial plains) and size, are subject to significant groundwater inflow. Hence, our classification approach also works for lakes. Nevertheless, we have excluded them from our analyses for two main reasons. First, for determining groundwater dependence, lake volume is likely more important than flow and absolute lake volumes are not explicitly included in our model. In addition, we cannot account for other complexities such as groundwater influence on lake stratification. Also in the revised manuscript, we will make more clear that we focus on lotic aquatic systems and exclude lenthic aquatic systems this case being lakes.
Two more methodological concerns:
5 m is deep water table as minimum for wetlands seems too deep and needs more justification to me..We believe that a groundwater table threshold < 5 meters effectively captures groundwater dependent wetlands. While groundwater levels closer to the surface (0.5 to 3 meters) support core wetland functions (Eamus et al., 2006; Winter, 1999), wetlands in arid and semi-arid regions can still exhibit groundwater dependence with water tables up to 5 meters, particularly in peripheral or drought-adapted areas (Stromberg et al., 2010). The 5 m threshold thus accommodates both core and peripheral groundwater-supported zones across varied climates and is consistent with topographic variations and uncertainties in surface elevations within a 10x10 km grid cell. The 50% soil saturation was added to discern dry areas with shallow groundwater levels from actual wetlands that are typically situated in areas with topographic convergence. We also performed sensitivity analysis varying groundwater depth and saturated area fraction with a total of 100 combinations, which shows this threshold produces the highest critical success index when validating against the Australian GDE atlas. We will add this description of the rationale behind our threshold to the supplementary of the revised manuscript.
References
Eamus, D., Froend, R., Loomes, R., Hose, G., & Murray, B. (2006). A functional methodology for determining the groundwater regime needed to maintain the health of groundwater-dependent vegetation. Australian Journal of Botany, 54(2), 97-114.
Stromberg, J., Lite, S., & Dixon, M. (2010). Effects of stream flow patterns on riparian vegetation of a semiarid river: implications for a changing climate. River research and applications, 26(6), 712-729.
Winter, T. C. (1999). Relation of streams, lakes, and wetlands to groundwater flow systems. Hydrogeology Journal, 7, 28-45.
I was also surprised that the difference between known and likely GDEs was not distinguished. I would test the importance of this assumption
Thank you for the suggestion. Attached is the outcome of comparing our mapped GDEs to known GDEs as reported by the Australian Atlas. We will also include this in our revised manuscript.
GDE Type Hit rate (Known and Likely GDEs) Hit rate (Known GDEs) Aquatic 87 % 79 %
Wetland 95 % 90 % Terrestrial 92 % 89 % Citation: https://doi.org/10.5194/hess-2024-112-AC1
-
AC1: 'Reply on RC1', Nicole Gyakowah Otoo, 08 Nov 2024
-
RC2: 'Comment on hess-2024-112', Anonymous Referee #2, 09 Oct 2024
This scientific work is very interesting and well-structured. It focuses on the study of groundwater-dependent ecosystems (GDEs) in Australia under both static and transient conditions, with a methodological approach based on coupling two models (PCR-GLOBWB 2 and MODFLOW). The objective is to identify these GDEs, comparing them with the official atlases of this nation, and also study their temporal evolution at a fine resolution. It uses a large dataset of observed data to calibrate the simulated levels. My only remaining concern is the high error between the observed and simulated level data. In Lines 225-226 the authors stated: “Our simulated heads are deeper than the observed, however with ~ 70 % having a bias ranging from 0 to 5 m…The relative variance shows an underestimation of groundwater level variation ~ 80% with a relative variance < 0. 6.” In my opinion, the error is not small if it can be as large as a 5 m difference. For a coastal aquifer it is a large and significant error that can delineate a different flow system. In the transient condition where you want to get an idea of how the whole system evolves over time, such a large error can delineate totally different scenarios. One could better argue this result in the discussion section, although it has already been mentioned as a limitation. Another question is about how the authors calibrate and validate the first model (PCR-GLOBWB 2), because the results of this model represent the input data in MODFLOW, therefore the overall results rely on the accuracy of the output data in the first stage.
In the following, a few suggestions:
- Line 41: insert the two citations into the same parenthesis
- Lines 104-105: avoid repetition of the word “only”
- Lines 130-131: is the second model used MODFLOF 2005? If yes, please specify it in this sentence
- Lines 205-206: repetition of saturated word
- Line 236: delete the double point at the end of the sentence
- Numbering of figures should agree with their appearance in the text. This also applies to supplementary material. For example, in the results, figure S3 is commented on first before S1 or S2
- It could be better if the authors add in the maps some toponyms mentioned in the text like Tasmania, Great Artesian basin, Murray Darlin Basin, New South Wales for a better understanding of the readers
- Line 171: “If there were multiple wells within a 1km cell, we take the average of these.” Considering the same month/year?
Citation: https://doi.org/10.5194/hess-2024-112-RC2 -
AC2: 'Reply on RC2', Nicole Gyakowah Otoo, 08 Nov 2024
This scientific work is very interesting and well-structured. It focuses on the study of groundwater-dependent ecosystems (GDEs) in Australia under both static and transient conditions, with a methodological approach based on coupling two models (PCR-GLOBWB 2 and MODFLOW). The objective is to identify these GDEs, comparing them with the official atlases of this nation, and also study their temporal evolution at a fine resolution. It uses a large dataset of observed data to calibrate the simulated levels.
We thank the reviewer for the kind words.
My only remaining concern is the high error between the observed and simulated level data. In Lines 225-226 the authors stated: “Our simulated heads are deeper than the observed, however with ~ 70 % having a bias ranging from 0 to 5 m…The relative variance shows an underestimation of groundwater level variation ~ 80% with a relative variance < 0. 6.” In my opinion, the error is not small if it can be as large as a 5 m difference. For a coastal aquifer it is a large and significant error that can delineate a different flow system. In the transient condition where you want to get an idea of how the whole system evolves over time, such a large error can delineate totally different scenarios. One could better argue this result in the discussion section, although it has already been mentioned as a limitation.
This is indeed a valid point. However, we are confident that the results of the groundwater model are useful for GDE-mapping, given that we found a much lower bias (we refer to the SI, Figure S3) and better evaluation statistics for shallower groundwater levels. To clarify, we will also separately report the validation statistics for shallow groundwater in the revised manuscript as well.
Another question is about how the authors calibrate and validate the first model (PCR-GLOBWB 2), because the results of this model represent the input data in MODFLOW, therefore the overall results rely on the accuracy of the output data in the first stage.
Thanks for bringing this up. In fact, we did not perform a calibration exercise in this paper, and still, we obtain acceptable results. The reason that we do not calibrate the model to our study area is that we want to test our approach for its global accuracy, including for data scarce regions. We will focus more on calibrating the models in a follow-up study with a global extent, using groundwater level data that are now becoming more globally available.
In the following, a few suggestions:
- Line 41: insert the two citations into the same parenthesis
- Lines 104-105: avoid repetition of the word “only”
- Lines 130-131: is the second model used MODFLOW 2005? If yes, please specify it in this sentence
- Lines 205-206: repetition of saturated word
- Line 236: delete the double point at the end of the sentence
- Numbering of figures should agree with their appearance in the text. This also applies to supplementary material. For example, in the results, figure S3 is commented on first before S1 or S2
- It could be better if the authors add in the maps some toponyms mentioned in the text like Tasmania, Great Artesian basin, Murray Darlin Basin, New South Wales for a better understanding of the readers
- Line 171: “If there were multiple wells within a 1km cell, we take the average of these.” Considering the same month/year?
Thanks for these suggestions. We will implement these suggested changes in a revised version of the paper.
Citation: https://doi.org/10.5194/hess-2024-112-AC2
-
RC3: 'Comment on hess-2024-112', Anonymous Referee #3, 17 Oct 2024
1.general comments
This paper investigates a method for mapping the potential continental-scale distribution of GDEs using a global groundwater model, which aligns with the scope of HESS. The framework and steps for mapping GDEs are clearly delineated. Although there are some discrepancies between the results and the actual atlas, the findings are still acceptable for large-scale research, and the results are adequate to support the interpretations and conclusions.
Unfortunately, the data and parameters necessary for coupling large-scale hydrological models with groundwater models—such as precipitation, evaporation, groundwater extraction, infiltration zoning, aquifer characteristics, and groundwater levels—are not yet clearly presented. Additionally, the impact of climate change and human activities on groundwater recharge has not been thoroughly analyzed. There has been insufficient discussion comparing the application of modeling methods with other remote sensing techniques.
2.specific comments
(1)Abstract
The manuscript states that at the end of the Abstract section "The proposed framework and methodology provide a basis for analyzing how global impacts of climate change and water use may affect GDEs extent and health", However, there appears to be no data or analysis of outcomes regarding climate change and water use to support this claim.
(2) Introduction
Since remote sensing is used as one of the effective methods in both regional and continental scale for identifying the potential GDEs distribution, why does the author abandon the approach and instead directly employ the surface water-groundwater model method for simulation and mapping of GDEs? Why not combine remote sensing method with surface-groundwater modeling method together?
(3) Data and methodology
This section outlines three steps of the GDEs mapping framework with illustrations and figures. However, the third step in Fig.1 presents the content of static GDEs mapping and analysis, which appears to lack the transient GDEs mapping analysis.
Section 2.1 Defining GDE classes (step 1):The author defines that the saturated area fraction greater than 50% and a shallow groundwater table (less than 5m) classified as a wetland GDEs. What is the scientific basis for that?
Section 2.2 Model set-up, sensitivity analysis and output evaluation (step 2):This section does not explore how the conceptual models of hydrology and groundwater are constructed, particularly how the boundary conditions of continental-scale groundwater models are established, how the permeability coefficients of phreatic aquifers, confined aquifers and riverbeds are acquired, and how the boundaries and hydraulic connections between the adjacent basins or hydrogeological units are determined. Is it necessary to mesh-refinement so that the conductance data from riverbeds be utilized in groundwater models? In determining net recharge, how can data be obtained on evaporation for aquatic area, wetland and terrestrial area, as well as groundwater infiltration due to precipitation? For the Pcr-globwb-2 model to simulate the saturated area fraction, which soil parameters or parameters from the unsaturated zone and saturated zone need to be input?
Section 2.3 GDE mapping (step 3):
When selecting transient GDE mapping, why are the two time periods 1979-1999 and 2000-2020 chosen? Are they related to climate change (such as changes in precipitation) and shifts in human activity (such as groundwater extraction)?
(4) Results
Groundwater depth is a crucial parameter for determining the GDEs, particularly the terrestrial GDEs, for example depth of 5m just mentioned in the paper, as it most directly reflects the distribution of GDEs. Why not select the typical years from the period of 1979-1999 and 2000-2020 to create a contour map of groundwater depth and compare it with the atlas that has already been produced?
This paper examines the contribution of groundwater to the stream in the Murray-Darling basin. However, how can you explain the decline in the dependency ratio when both groundwater levels and stream flow are decreasing? Were the monitoring sites for groundwater levels and stream flow selected from the upper, middle, or lower reaches of the basin? We are unsure whether the simulation accuracy at the watershed-scale will be higher than that at the continental-scale. Why not to map the distribution of GDEs at the basin-scale for typical years and compare the results with the actual atlas and then get a hit rate?
(5)Discussions and conclusions
The method discussed in this paper still exhibits a notable gap in evaluation accuracy when compared to the actual GDEs distribution derived from the Australian atlas. Can it be utilized for assessments at other regional, continental, and even global scales? What are the advantages and disadvantages of this method in relation to other scholars' combined remote sensing hydrogeological survey techniques?
Has it been compared and analyzed against the relevant results of the following article? ——Rohde, M.M., Albano, C.M., Huggins, X. et al. Groundwater-dependent ecosystem map exposes global dryland protection needs. Nature 632, 101–107 (2024). https://doi.org/10.1038/s41586-024-07702-8
3.technical corrections
Line 19: “using a hit rate, false alarm, and critical success index,” Perhaps the term “missing rate” was lost. It would be changed to “using a critical success index derived from hit rate, false alarm, and missing rate”
Line 99, Fig.1: The abstract outlines a step for evaluation of transient mapping; however, Figure 1 does not provide an analysis and its arrow indication is unclear. The name of fig.1 is “Groundwater dependent ecosystems (GDE) modelling framework or mapping framework?
Line 110-111: Please confirm it is that the maximum rooting depth is greater than the depth to groundwater table.
Line 229: groundwater level variation ∽80% with a relative variance < 0. 6). _ missing a bracket.
Figure 3: Lack of scale bar
Line 261: false alarm ratio, or false alarm rate?
Line 262-263: and green represents hit rate.
Line 285 Figure 5: The figure is unclear and lacks a scale bar.
Line 294: depends on groundwater level and streamflow.
-
AC3: 'Reply on RC3', Nicole Gyakowah Otoo, 08 Nov 2024
This paper investigates a method for mapping the potential continental-scale distribution of GDEs using a global groundwater model, which aligns with the scope of HESS. The framework and steps for mapping GDEs are clearly delineated. Although there are some discrepancies between the results and the actual atlas, the findings are still acceptable for large-scale research, and the results are adequate to support the interpretations and conclusions.
We thank the reviewer for these encouraging remarks.
Unfortunately, the data and parameters necessary for coupling large-scale hydrological models with groundwater models—such as precipitation, evaporation, groundwater extraction, infiltration zoning, aquifer characteristics, and groundwater levels—are not yet clearly presented.
We did not present this information because it is available in several of our previous publications that we refer to (Sutanudjaja et al. 2018; De Graaf et al. (2017), Verkaik et al. (2023). However, we agree that it would be informative for the reader to provide more context. Therefore, we will:
- Add a table to the SI with an overview of the datasets and their source. These will include meteorological forcing, land use, soil, soil physical parameters, the surface water network, groundwater extraction estimates, aquifer characteristics, and groundwater level data.
- To illustrate spatial variability in forcing and parameter fields of the groundwater model, we will add to the SI: maps of average rainfall, average potential evapotranspiration, average groundwater recharge, groundwater extraction for the last year, and conductivity and thickness of confining layer and the aquifer. We will add a reference to the groundwater model data repository (Verkaik et al., 2023) where these time series of the forcing (monthly time step) and the parameter fields can be obtained.
Additionally, the impact of climate change and human activities on groundwater recharge has not been thoroughly analyzed.
We concur that climate change and human activities have impact on the GDE occurrence that we observe. Therefore, we: a) analyse the impacts of groundwater table trends and wet/dry periods on aquatic ecosystems; b) we quantify the differences in groundwater recharge between two periods (1979-1999 and 1999-2019) and compare these to the identified patterns of changes in GDEs (S5 and 6). We include additional information to make this clearer in the revised paper. We did not analyse the impacts of climate change and human impacts separately using a full factorial analysis, because we consider this out of the scope of this paper. The scope of this paper is to test a mapping method and its sensitivity to changes in groundwater depth over time. We believe that this has been sufficiently covered by our analyses. In subsequent work we will focus more on how past and future changes in GDE extent relate to climate signals and changes in human water use.
There has been insufficient discussion comparing the application of modeling methods with other remote sensing techniques.
We stress that other techniques are mentioned in the introduction section, among which remote sensing. We focus on modelling, since we aim to project future changes in GDE extent in a follow-up study, which is not possible with remote sensing.
2.specific comments
(1)Abstract
The manuscript states that at the end of the Abstract section "The proposed framework and methodology provide a basis for analyzing how global impacts of climate change and water use may affect GDEs extent and health", However, there appears to be no data or analysis of outcomes regarding climate change and water use to support this claim.
We note that we do provide a tentative assessment of impacts of both climate change (reflected in groundwater recharge change) and groundwater pumping, as explained in our reply to one of the main comments above. Nevertheless, we will formulate the claim more carefully as:
“The proposed framework and methodology provide a first step towards a method for analysing how impacts of climate change and human water use may affect the extent and condition of GDEs globally."
(2) Introduction
Since remote sensing is used as one of the effective methods in both regional and continental scale for identifying the potential GDEs distribution, why does the author abandon the approach and instead directly employ the surface water-groundwater model method for simulation and mapping of GDEs? Why not combine remote sensing method with surface-groundwater modeling method together?
The main reason for using a modelling approach, as stated in the Introduction, is that we intended to develop a mapping method that can be used in a next step (follow up work) to project GDE change in the future. This cannot be done using remote sensing alone. We could however use remote sensing data for evaluation and calibration, and we do this in a sense, since the Australian GDE Atlas that we evaluate our approach on is partly based on remote sensing (in addition to field surveying and expert knowledge) (Doody et al. 2017).
(3) Data and methodology
This section outlines three steps of the GDEs mapping framework with illustrations and figures. However, the third step in Fig.1 presents the content of static GDEs mapping and analysis, which appears to lack the transient GDEs mapping analysis.
For transient, we split our simulation time into two periods of 20 years each (1-1-1979 to 31-12-1999 and 1-1-2000 to 01-01-2019) and we analysed the changes in the average number of months that GDEs depend on groundwater between period 2 (1-1-2000 to 01-01-2019) and period 1 (1-1-1979 to 31-12-1999). Here we kindly refer to Figures 5, 6 and 7.
Section 2.1 Defining GDE classes (step 1):The author defines that the saturated area fraction greater than 50% and a shallow groundwater table (less than 5m) classified as a wetland GDEs. What is the scientific basis for that?
Thanks for this comment. As explained also in our reply to one of the comments of Referee 1, a groundwater table threshold < 5 meters effectively captures groundwater dependent wetlands as levels closer to the surface (0.5 to 3 meters) support core wetland functions (Eamus et al., 2006; Winter, 1999). In arid and semi-arid regions, wetlands can still exhibit groundwater dependence with water tables up to 5 meters, particularly in peripheral or drought-adapted areas (Stromberg et al., 2010). This threshold accommodates both core and peripheral groundwater-supported zones across varied climates and is consistent with topographic variations and uncertainties in surface elevations within a 10x10 km grid cell. While the 50% soil saturation was added to discern dry areas with shallow groundwater levels from actual wetlands that are typically situated in areas with topographic convergence. We also performed sensitivity analysis varying groundwater depth and saturated area fraction with a total of 100 combinations which shows this threshold produces the highest critical success index when validating against the Australian GDE atlas.
References
Eamus, D., Froend, R., Loomes, R., Hose, G., & Murray, B. (2006). A functional methodology for determining the groundwater regime needed to maintain the health of groundwater-dependent vegetation. Australian Journal of Botany, 54(2), 97-114.
Stromberg, J., Lite, S., & Dixon, M. (2010). Effects of stream flow patterns on riparian vegetation of a semiarid river: implications for a changing climate. River research and applications, 26(6), 712-729.
Winter, T. C. (1999). Relation of streams, lakes, and wetlands to groundwater flow systems. Hydrogeology Journal, 7, 28-45.
Section 2.2 Model set-up, sensitivity analysis and output evaluation (step 2):This section does not explore how the conceptual models of hydrology and groundwater are constructed, particularly how the boundary conditions of continental-scale groundwater models are established, how the permeability coefficients of phreatic aquifers, confined aquifers and riverbeds are acquired, and how the boundaries and hydraulic connections between the adjacent basins or hydrogeological units are determined. Is it necessary to mesh-refinement so that the conductance data from riverbeds be utilized in groundwater models? In determining net recharge, how can data be obtained on evaporation for aquatic area, wetland and terrestrial area, as well as groundwater infiltration due to precipitation? For the Pcr-globwb-2 model to simulate the saturated area fraction, which soil parameters or parameters from the unsaturated zone and saturated zone need to be input?
Here we kindly refer to our answer to the general question above about the input data and parameterization.
Section 2.3 GDE mapping (step 3):
When selecting transient GDE mapping, why are the two time periods 1979-1999 and 2000-2020 chosen? Are they related to climate change (such as changes in precipitation) and shifts in human activity (such as groundwater extraction)?
We have 40 years of simulation. We split this period in half to estimate if changes could be picked up by the mapping method and if these could be explained from groundwater level changes subject to recharge and pumping. The change in recharge between the two periods is not necessarily related to climate change, but the average climate could be different between the periods due to the occurrence of large climate modus (e.g. ENSO). Furthermore, pumping is larger in the second half due to socioeconomic development.
(4) Results
Groundwater depth is a crucial parameter for determining the GDEs, particularly the terrestrial GDEs, for example depth of 5m just mentioned in the paper, as it most directly reflects the distribution of GDEs. Why not select the typical years from the period of 1979-1999 and 2000-2020 to create a contour map of groundwater depth and compare it with the atlas that has already been produced?
We will add a more extensive justifation of the threshold of 5 m (see also our replies to earlier comments on this point). We show the average groundwater depth map and the differences in groundwater depth between the first and the second period in the SI, as well as comparison with observations.
This paper examines the contribution of groundwater to the stream in the Murray-Darling basin. However, how can you explain the decline in the dependency ratio when both groundwater levels and stream flow are decreasing?
If groundwater discharge is decreasing more rapidly than streamflow, the ratio declines. What happens to the ratio is thus very much dependent on how drought or desiccation impacts the groundwater system and surface runoff. This can be different from one place to the next and is captured by our modelling system.
Were the monitoring sites for groundwater levels and stream flow selected from the upper, middle, or lower reaches of the basin?
We use all the monitoring sites with sufficient data for the evaluation for the groundwater levels simulations. In the example we use examples from the lower reaches in the MD basin. We will show the location of the monitoring stations on a map in the revised paper.
We are unsure whether the simulation accuracy at the watershed-scale will be higher than that at the continental-scale. Why not to map the distribution of GDEs at the basin-scale for typical years and compare the results with the actual atlas and then get a hit rate?
We are not sure whether we understand this remark. The Murray-Darlin results are extracted from the continental scale simulation and are thus not necessarily more accurate.
(5)Discussions and conclusions
The method discussed in this paper still exhibits a notable gap in evaluation accuracy when compared to the actual GDEs distribution derived from the Australian atlas. Can it be utilized for assessments at other regional, continental, and even global scales? What are the advantages and disadvantages of this method in relation to other scholars' combined remote sensing hydrogeological survey techniques?
Has it been compared and analyzed against the relevant results of the following article? ——Rohde, M.M., Albano, C.M., Huggins, X. et al. Groundwater-dependent ecosystem map exposes global dryland protection needs. Nature 632, 101–107 (2024). https://doi.org/10.1038/s41586-024-07702-8
We respectfully beg to accept that we disagree about the accuracy of our method, which we consider quite high when looking at the high hit rates, low miss rates and high critical success ratios. We really appreciate the paper by Rohde et al. (2024), which we also noted ourselves. This paper was not published at the time we submitted our paper and we will now refer to it in the revised version. When comparing their approach to ours, we note the following:
- Their evaluation results in Australia are similar to our evaluation results
- This is again a method based on stationary mapping and not suitable to project future changes of GDE-occurrence;
We will add a short discussion on comparison of our results with their findings.
3.technical corrections
Line 19: “using a hit rate, false alarm, and critical success index,” Perhaps the term “missing rate” was lost. It would be changed to “using a critical success index derived from hit rate, false alarm, and missing rate”
Line 99, Fig.1: The abstract outlines a step for evaluation of transient mapping; however, Figure 1 does not provide an analysis and its arrow indication is unclear. The name of fig.1 is “Groundwater dependent ecosystems (GDE) modelling framework or mapping framework?
Line 110-111: Please confirm it is that the maximum rooting depth is greater than the depth to groundwater table.
Line 229: groundwater level variation ∽80% with a relative variance < 0. 6). _ missing a bracket.
Figure 3: Lack of scale bar
Line 261: false alarm ratio, or false alarm rate?
Line 262-263: and green represents hit rate.
Line 285 Figure 5: The figure is unclear and lacks a scale bar.
Line 294: depends on groundwater level and streamflow.
Thanks for these suggestions. We will implement these suggested changes in a revised version of the paper.
Citation: https://doi.org/10.5194/hess-2024-112-AC3 -
AC4: 'Reply on RC3', Nicole Gyakowah Otoo, 08 Nov 2024
Publisher’s note: this comment is a copy of AC3 and its content was therefore removed on 8 November 2024.
Citation: https://doi.org/10.5194/hess-2024-112-AC4
-
AC3: 'Reply on RC3', Nicole Gyakowah Otoo, 08 Nov 2024
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
384 | 129 | 22 | 535 | 43 | 18 | 15 |
- HTML: 384
- PDF: 129
- XML: 22
- Total: 535
- Supplement: 43
- BibTeX: 18
- EndNote: 15
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1