the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Spatial and temporal simulation of groundwater recharge and cross-validation with point measurements in volcanic aquifers with variable topography
Abstract. A physically distributed water balance model called WetSpass is applied to estimate the recharge for the semi-humid Lake Tana basin in northwest Ethiopia. Lake Tana basin is one of the growth corridors of the country, where huge waterworks infrastructure is developing. Estimating groundwater recharge at required spatial and temporal scales is a challenge in groundwater management, sustainability and pollution studies. In this study, the WetSpass model is developed at 90 m grid resolution. The spatial recharge map by WetSpass is cross-validated with water table fluctuation (WTF) and chloride mass balance (CMB) methods. The mean annual recharge, surface runoff, and evapotranspiration over the whole basin using WetSpass are estimated at 315 mm, 416 mm, and 770 mm of rainfall, respectively. The mean annual recharge ranges from 0 mm to 1085 mm (0 % to 57 % of the rainfall): 0 mm at water bodies and highest on flat, sandy loam soil and bush land cover. Similarly, a high range of recharge is also noted using WTF and CMB methods showing the strong heterogeneous nature of the hydro(meteoro)logical characteristics of the area. Generally, the recharge is found higher in southern and eastern catchments and lower in the northern catchments, primarily due to higher rainfall amounts in the former parts. A fair general correlation between the recharge by WTF and WetSpass is found. WetSpass is effective in aquifers where diffuse recharging mechanism is the predominant type and recharge is controlled by rainfall. It is less effective in the storage-controlled flat floodplain alluvial and fractured rock aquifer areas. In these areas, the point estimates by WTF and CMB are effective and can be considered as reliable values. The land use change from 1986 to 2014 brought a relatively small hydrological change in recharge although the land use has changed significantly.
- Preprint
(1780 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2021-527', Anonymous Referee #1, 23 Dec 2021
The paper reports regional study based on physically-based water balance WetSpass model, which has been applied to simulate the water balance components, including runoff, evapotranspiration and groundwater recharge, at Lake Tana basin in Ethiopia. Further, the spatial groundwater recharge map by WetSpass was validated through water table fluctuation and chloride mass balance methods. The topic is interesting. However, the paper under its current form suffers from several points concerning its content and structure. I’m not English native, but it looks that the paper needs bit English improvements to reach high written level required by high-tier international journal like HESS. The Introduction section looks too modest. Usually, the Introduction should provide a thorough literature review on the topic, starting from the general to the specific and clearly identifying the gap(s) in current knowledge that the paper addresses. This allows to formulate explicitly research questions, hypothesis, and the contribution of the study. The authors give general info on the recharge at the beginning followed by introducing the characteristics of the Ethiopian volcanic basin. However, it is not clear what is the issue. Novelty and justification seem the main issue of the paper. There is unclear scientific rationale on why the authors are pursuing the proposed study. The literature review seems does not specify clearly research gap(s) of previous studies. The authors stated that ‘‘Different methods are developed for recharge estimation. However, choosing appropriate methods is often challenging’’. This directs the reader to consider that novelty will be ‘‘developing recharge choosing tool’’, but it is not the case. Rather, the paper is oriented towards regional study focusing on groundwater recharge estimating and mapping. It is not clear what is the paper contribution compared to previous recharge studies conducted on the study area (e.g., Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005). Other points when the authors mention that ‘‘the area has a climate with long dry winters and short rainy summer seasons. Hence, the groundwater recharge-discharge processes are expected to vary highly both spatially and temporally’’. I believe that with only two seasons and same type of precipitation (rainfall) at the study area, there are other areas around the world presenting more complicate highly spatiotemporal variation such as the case of the humid northern regions dominated by variety of meteorological inputs or the subhumid regions submitted to many different seasons. My question, where is the situation of the present study compared to other recharge studies undertaken under different climatic conditions. Also, the authors indicated that ‘‘One of the challenges for the point recharge estimation methods is their incapability to estimate it in a spatially distributed way’’. I respectfully disagree with this statement as one of the main advantages stemming from integrating remote sensing with GIS within recharge mapping is in the capability to investigate the recharge at unprecedented levels of spatiotemporally variability. However, I agree that only limited studies have incorporated the spatial distribution of recharge with corresponding rates. Among them the references cited by the authors (e.g., Batelaan and De Smedt 2007). However, there is unclear scientific rationale on why the authors are pursuing the proposed study when they adopt the WetSpass model. The literature review seems does not specify clearly research gap(s) of previous studies related to the WetSpass model and the novelty compared to the previous WetSpass studies. If just using this model, the proposed study should be simple case study that can be fit with other journals like Hydrology Journal: Regional Studies. This seems confirmed with the main objectives cited at the end of the Introduction section. When the authors mention ‘‘studying the recharge mechanisms’’, what is the contribution relative to the existing work of Yenehun (2020) cited in lines 119-120 of the paper. For the methodology (Section 3), I raised some comments which are integrated in the attached document. My main concern is the structure of this section which seems confusing between methods and results. The authors introduce some results (e.g., developed maps) that should be placed in the Results section. Also, many methods are suffering from less details and explanations, and often the authors introduce some information without previous definition. This approach complicates the understanding of the methodology, while it looks appropriate to introduce a flowchart for the model development and explain how each equation, input or parameter has been adopted, calculated, or assumed. In many places of the text, the authors based on their own knowledge, but less detail is introduced. This cannot allow reader to judge this knowledge and understand the background of this knowledge. The authors provided criticism about interpolation in the Introduction section ''it is unwise to extrapolate or regionalized the result by the conventional point recharge estimation techniques''. However, many concepts (e.g., maps) of the proposed paper are based on the interpolation approach. In separate locations, the authors mention some ways that have been adopted for the calibration phase, it will be better to provide a sole section introducing all the calibration process rather to be distributed over several places in the text. I formulated many specific comments in the attached document, but I feel the paper suffers from a lack of novelty and structure. Regarding the raised comments, I believe the paper is not suitable for publication and I suggest directing it to other focusing journals.
-
AC1: 'Reply on RC1', Alemu Yenehun Beyene, 25 Jan 2022
Response to the general comments forwarded by the referee:
English language improvements and corrections have been made including suggestions made by you in the revised version.
In the general comment as well as in the specific comments (in the attached manuscript of the referee), clarity about the objectives has been raised as the main limitation of the manuscript. This could be due to problems of us in explaining the main objectives as explicitly as possible which we thought we have done in the revised version that we will attach.
The main aims of this paper are:
Each recharge estimation technique has its own limitations, depending on the recharge mechanisms, the aquifer topographical setting, aquifer type (for example fracture vs porous soil aquifers, confined vs unconfined), aquifer geometry, etc. Hence, the effectiveness of recharge estimation techniques is varying, and thus in this study evaluation of the hydrological model (WetSpass) has been made (i.e. validation with measured point values though some variables are estimated in these methods too!!) for the Lake Tana basin having high topographic, variable slope, and comprising different volcanic aquifers. Such aquifer types are found in many parts of Ethiopia and is the major groundwater resources for the global water need. The WetSpass model applied in this study is a physically-based distributed hydrological model which gives spatially distributed spatial water balance terms including recharge. Yes, there are statistical interpolations for some of its input variables but the final spatial recharge and other water balance output maps are different from simply measured point value interpolations, rather they have resulted after optimal global model parameters are set through the calibration process. Hence, one of the objectives of the paper is evaluating the hydrological model, and thus giving suggestions for recharge methods to be better applied for different aquifers lying at different topographical settings: this has been clearly put in the discussion and conclusion section of this manuscript. The paper put concluding remarks, where physical hydrological models, would be effectively applied. Why they are less effective for some aquifers and more for others. This helps to take into account the accuracy of recharge values in the area-specific water management plans and decisions being made at different levels.
The next aim of this manuscript is to produce a spatially distributed groundwater recharge rate map for the Lake Tana basin, which is the source of the transboundary Blue Nile River basin where a high tension of hydropolitics is currently affecting the entire region. The spatially distributed recharge rate map produced in this study for the Lake Tana basin is an important output for further groundwater modeling, and water management issues of the basin.
The other one is identifying the most important hydrometeorological and physical factors and prioritizing them for groundwater recharge distribution (qualitatively) for the study basin, and so for similar study areas, including the effect of land use change on the recharge and other hydrological terms. These have been discussed in the result and discussion section and concluded in the conclusion part.
With due you respect, these objectives, especially the first one are important for the international readers. Many hydrological models for different areas (small catchment-scale to large basins) are developed for estimating groundwater recharge. However, evaluation of the method with point estimations based on direct water level measurement (WTF) and chemical tracer (CMB) gives a good insight for future recharge estimation techniques for similar aquifer types wherever they are located in the world. In this study, identifying topographical and geological characteristics, and thus the recharge mechanisms have found an important factor and starting point in selecting recharge estimation methods. The study also pointed out that the common approach that is being implemented i.e. calculating recharge by a multitude of methods and averaging out the results of the different methods is found unreliable. Rather selecting appropriate one or few technique/(s) and considering that as the optimal result is recommended.
Evaluating the groundwater recharge estimation mechanism for the volcanic aquifers lying at different topographical settings, that represents a vast major part of the world groundwater aquifers, and able to suggest more appropriate method/(s) will benefit different further similar studies and researchers. Hence, this study can be seen as a dual purpose paper: evaluating the physical-based hydrological model (and so other similar models), and giving a spatial recharge rate map for the important Lake Tana basin.
Thank you (the reviewer) for your valuable comments, in the new version, shortcomings in the introduction section have been improved as per the suggestions. The research questions are outlined and the contribution of the research output is mentioned in the new version.
The reason why we talk about “choosing appropriate methods” is to direct our reader that evaluation of a method will be one of the objectives of this manuscript. The reason why different methods were developed for recharge estimation rather than using the simplest and cost-effective method is due to the fact that different aquifers have different recharge mechanisms for which some methods are more effective than others. It is why we did evaluation and validation of the existing method. Furthermore, when we say point recharge estimation, we are referring to recharge estimation techniques that estimate at a given point (e.g. on groundwater well, spring, etc.) such as WTF and CMB methods. Otherwise, yes indeed, we agree with the advance of Remote sensing and GIS, and with the developing capability of the grid-based spatially simulating hydrological models, producing spatially distributed water balance maps is so possible and being widely applied. The applied WetSpass model for this study is one of such kinds, and we only applied an existing model (nothing new is done in this study as far as new methodology is concerned). As we have aforementioned, we pursued this study, for evaluating the physically-based hydrological models for our study basin so that areas having similar hydrogeological characteristics will take into account the suggestions of our evaluation result for future application. Besides, the recharge rate map for the basin, which is an important research product for groundwater exploitation practices and plans, will help groundwater managers and policymakers in this important area which is regarded as one of the growth corridors of the country. Besides, the recharge map is also an important input for further groundwater modeling work.
With due respect, the studies by (Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005) are not all about recharge estimations studies. It is only Demlie et al., 2008, who did a recharge estimation only using the CMB method on a volcanic aquifer for the Akaki catchment located at about 550 km from the Lake Tana basin. The others had made different hydrogeological studies but had put general statements about the recharge variability across the volcanic aquifers. Thus, we can make clear that the presented study in this paper is different from the studies made by these authors.
Some maps in the methodology are shifted to result section as suggested. The shortened explanation of the methodology section is for the reason of shorting the paper. We thought citing the papers that develop the original methodologies and latter modifications will be enough and give detailed background knowledge including the different mathematical equations. However, parameter value modifications that have been done in this paper are discussed and justified in the new version of this paper.
Citation: https://doi.org/10.5194/hess-2021-527-AC1 -
AC2: 'Reply on RC1', Alemu Yenehun Beyene, 22 Feb 2022
- Response to reviewer 1 review comments
Response to the general comments:
English language improvements and corrections have been made including suggestions made by you in the revised version.
In the general comment as well as in the specific comments (in the attached manuscript of the referee), clarity about the objectives has been raised as the main limitation of the manuscript. This could be due to problems of us in explaining the main objectives as explicitly as possible which we thought we have done in the revised version that we will attach.
The main aims of this paper are: - Each recharge estimation technique has its own limitations, depending on the recharge mechanisms, the aquifer topographical setting, aquifer type (for example fracture vs porous soil aquifers, confined vs unconfined), aquifer geometry, etc. Hence, the effectiveness of recharge estimation techniques is varying, and thus in this study evaluation of the hydrological model (WetSpass) has been made (i.e. validation with measured point values though some variables are estimated in these methods too) for the Lake Tana basin having high topographic, variable slope, and comprising different volcanic aquifers. Such aquifer types are found in many parts of Ethiopia and is the major groundwater resources for the global water need. The WetSpass model applied in this study is a physically-based distributed hydrological model which gives spatially distributed water balance terms including groundwater recharge. Yes, there are statistical interpolations for some of its input variables but the final spatial recharge and other water balance output maps are different from simply estimated point value interpolations, rather they have resulted after optimal global model parameters are set through the calibration process (detail explanation about this issue can be found in the response to the line by line comments). Hence, one of the objectives of the paper is to evaluate the hydrological model, and thus give suggestions for recharge methods to be better applied for different aquifers lying at different topographical settings: this has been clearly put in the discussion and conclusion section of this manuscript. The paper put concluding remarks, where physical hydrological models, would be effectively applied. Why they are less effective for some aquifers and more for others. This helps to take into account the accuracy of recharge values in the area-specific water management plans and decisions being made at different levels.
- The next aim of this manuscript is to produce a spatially distributed groundwater recharge rate map for the Lake Tana basin, which is the source of the transboundary Blue Nile River basin where a high tension of hydropolitics is currently affecting the entire region. The spatially distributed recharge rate map produced in this study for the Lake Tana basin is an important output for further groundwater modeling, and water management issues of the basin.
- The other one is identifying the most important hydrometeorological and physical factors and prioritizing them for groundwater recharge distribution (qualitatively) for the study basin, and so for similar study areas, including the effect of land use change on the recharge and other hydrological terms. These have been discussed in the result and discussion section and concluded in the conclusion part.
With due respect, these objectives, especially the first one are important for international readers. Many hydrological models for different areas (small catchment-scale to large basins) are developed for estimating groundwater recharge. However, evaluation of the method with point estimations based on direct water level measurement (WTF) and chemical tracer (CMB) gives a good insight for future recharge estimation techniques for similar aquifer types wherever they are located in the world. In this study, identifying topographical and geological characteristics, and thus the recharge mechanisms have found an important factor and starting point in selecting recharge estimation methods. The study also pointed out that the common approach that is being implemented i.e. calculating recharge by a multitude of methods and averaging out the results of the different methods is found unreliable. Rather selecting an appropriate one or few technique/(s) and considering that as the optimal result is recommended.
Evaluating the groundwater recharge estimation mechanism for the volcanic aquifers lying at different topographical settings, that represents a vast major part of the world groundwater aquifers, and able to suggest more appropriate method/(s) will benefit different further similar studies and researchers. Hence, this study can be seen as a dual purpose paper: evaluating the physical-based hydrological model (and so other similar models), and giving a spatial recharge rate map for the important Lake Tana basin.
Thank you (the reviewer) for your valuable comments, in the new version, shortcomings in the introduction section have been improved as per the suggestions. The research questions are outlined and the contribution of the research output is mentioned in the new version.The reason why we talk about “choosing appropriate methods” is to direct our reader that evaluation of a method will be one of the objectives of this manuscript. The reason why different methods were developed for recharge estimation rather than using the simplest and cost-effective method is due to the fact that different aquifers have different recharge mechanisms for which some methods are more effective than others. It is why we did an evaluation and validation of the existing method. Furthermore, when we say point recharge estimation, we are referring to recharge estimation techniques that estimate at a given point (e.g. on groundwater well, spring, etc.) such as WTF and CMB methods. Otherwise, yes indeed, we agree with the advance of Remote sensing and GIS, and with the developing capability of the grid-based spatially simulating hydrological models, producing spatially distributed water balance maps is so possible and being widely applied. The applied WetSpass model for this study is one of such kind, and we only applied an existing model (nothing new is done in this study as far as new methodology is concerned). As we have aforementioned, we pursued this study, for evaluating the physically-based hydrological models for our study basin so that areas having similar hydrogeological characteristics will take into account the suggestions of our evaluation result for future application. Besides, the recharge rate map for the basin, which is an important research product for groundwater exploitation practices and plans, will help groundwater managers and policymakers in this important area which is regarded as one of the growth corridors of the country. Besides, the recharge map is also an important input for further groundwater modeling work.
With due respect, the studies by (Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005) are not all about recharge estimations studies. It is only Demlie et al., 2008, who did a recharge estimation only using the CMB method on a volcanic aquifer for the Akaki catchment located at about 550 km from the Lake Tana basin. The others had made different hydrogeological studies but had put general statements about the recharge variability across the volcanic aquifers. Thus, we can make clear that the presented study in this paper is different from the studies made by these authors.
Some maps in the methodology are shifted to the result section as suggested. The shortened explanation of the methodology section is for the reason of shorting the paper. We thought citing the papers that develop the original methodologies and later modifications will be enough and give detailed background knowledge including the different mathematical equations. However, parameter value modifications that have been done in this paper are discussed and justified in the new version of this paper (some of the details can be found in the following response to your line-by-line comments).
Response to detail line by line comments
Comment1: Language editing such as deletion, addition, and rewriting are suggested.
Reply: kindly accepted, and language edition has been done throughout the body of the manuscript, according to the suggestions.
Comment 2: The different correction suggestions for the location map (Fig. 1)
Reply: accepted, and appropriate corrections have been done in the new revised manuscript.
Lines 128-126: “That means years of seasonal or monthly time series data are averaged into single seasons or months.” How this limitation is considered in this study?
Reply: The objective of this study (as far as WetSpass is concerned) is to produce long-term average spatial recharge and other water balance component maps showing the spatial variation of the components. Furthermore, it is to see the seasonal variations of these water balance terms. It is not to see the fine-time scale temporal variability of them. It would indeed be nice if we, for example, were able to see the change of recharge over the years, but the WetSpass model cannot do that.
Line 139: how this parameter is calculated
Reply: thank you for the comment. The parameter is the amount of rainfall that reaches the surface (total RF minus the one intercepted by the plant canopy). The interception is calculated by the model using the land use at each modeling pixel.
In the WetSpass model, depending on the type of vegetation, the interception fraction represents a constant percentage of the annual precipitation value. Thus, the fraction decreases with an increase in the annual total rainfall amount (since the vegetation cover is assumed to be constant throughout the simulation period). The detail about it is found in Batelaan and De Smedt (2007, 2001).
Line 142: how EP is calculated? What do you mean by open water?
Reply: Open water used here to mean a water body (could be lake water, dam reservoir, etc.).
The description how evapotranspiration is calculated in WetSpass is mentioned as follows.
For the calculation of seasonal evapotranspiration, a reference value of transpiration is obtained from open-water evaporation and a vegetation coefficient:Trv= c Eo
Trv = the reference transpiration of a vegetated surface [LT-1];
Eo = potential evaporation of open water [LT-1] and
c = vegetation coefficient [–].
This vegetation coefficient can be calculated as the ratio of reference vegetation transpiration as given by the Penman-Monteith equation to the potential open-water evaporation, as given by the Penman equation
For vegetated groundwater discharge areas, the actual transpiration (Tv) is equal to the reference transpiration as there is no soil or water availability limitation:Tv= Trv if (Gd - ht)≤ Rd
Gd, is groundwater depth [L];
ht is the tension saturated height [L] and
Rd is the rooting depth [L].
For vegetated areas where the groundwater level is below the root zone the actual transpiration is given by:
Tv= f(θTrv if (Gd- ht ) < Rdf(θ) is a function of the water content
the detail about it is found in the original model developing literature by Batelaan and De Smedt (2007, 2001).
Line 147: give detail explanation about the parameter modifications
Reply: Detailed explanation is added in the revised one. We mean some adjustment for the parameters’ values on the default land use parameter table has been made, for example, the root depth of forest land is changed. In the model parameter table, each land use type has given impervious, bare land, vegetation, and open water percentages. In the default parameter of the WetSpass model, the vegetation area coverage for the Bush and Shrubland is 100%. This is based on the observation of the land use class type in the temperate zone (the Netherlands and Belgium). However, the Bush and Shrubland land use type in the Ethiopian (tropics) context is different: the vegetation is sometimes sparsely distributed, and is with some bare land component, during our field verification on the land use type, we tried to guess (using parcels land on which we do some measurement), and came up with about 10% is bare land and the rest 90% is consisting of vegetation. Similarly, adjustments have been made by Gebreyohannes et al. (2013) during their application of the model for Geba catchment, in northern Ethiopia. Similarly, the sub-afro-alpine vegetation land use type found in our area (consisting of about 0.3% of the total area coverage) is not present in the default land use classes of the WetSpass model. However, we made an equivalent with wet meadow land use type, and following a similar procedure, we modified the land use percentage to 80% vegetation and 20% bare land (it was 100% vegetation for wet meadow land cover type in the model parameter table).
Line 165: Paragraph above does not include land use as a parameter for calibration??
Reply: Yes, indeed land use is not used for calibration. It is to mean the land use which is used for model calibration is the one made using 2014 satellite images. The model parameters were optimized using this land use and then after the developed model was rerun using other land uses (1986 and 2000), just to see the change of the water balance components as a result of the change in land use over these years. Some paraphrasing has been done to make clear the issue.
Line 166: why these years (186, 2000, 2014) are chosen?
Reply: Because we have existing land use maps made using the satellite images taken in those years.
Line 170: It is unclear how the different model parameters serve the model?
Reply: as it is mentioned earlier in our response to your general comments, putting every detail of the model here would make the paper too long as the objective of this manuscript is neither modifying the model nor testing it. It is rather for applying the model and comparing (validating) with other methods, see how effective are such models (physically-based hydrological models) in estimating groundwater recharge (given the different assumptions these models have though they give fine spatially distributed maps), and identifying the controlling physical and meteo(hydro)logical factors, etc. Hence, detailed equations are better to be cited and be read from the published papers (containing model development and latter modifications). However, thank you to you, we found it is short and we added some more elaborations in the revised version.
Line 179: why??
Reply: Because the source for all water balance components including recharge is rainfall. For the months with no rainfall at all (most of the winter months) is less important. However, it does not mean that knowing at what month of summer (June, July, August, September ) is the most important recharge is happening, and at which month is the less, and why that is happening is insignificant for the hydrological knowledge of the area. It is so important and other works dedicated to it should focus on and study it.
Line 183: Unclear
Reply: it is to the rainfall amount, corrected accordingly.
Line 184: provide the number 8??
Reply: 8 of the stations are established by the project that funds this study. It is BDU-IUC project (funded by VLIR-UOS, by the Flemish Government of Belgium).
Line 187: In the introduction section you criticize the interpolation techniques.
Reply: Yes, in general, interpolation has drawbacks as it is estimations at several places given real measurement at some spatial points of an area. However, it depends on the property of the parameter (the matter) for which that interpolation is being executed for. For example for our specific case, interpolation for RF and other meteorological parameters has been done as the WetSpass model needs spatial maps as an input. However, interpolation of some parameters e.g. groundwater recharge (which is mentioned in the introduction section) is highly dependent on the ground hydrogeological condition of the area and is highly variable irrespective of spatial proximity for which interpolation mostly depends. Hence, instead of interpolating estimated results at measuring points of an area (e.g. by the WTF method), it is highly better to produce it using some method which can take into account all those hydrogeological factors (which use them as input). Hence, playing with spatially distributed input factors in the mathematical calculations (though some are prepared by interpolation) is not the same as final point value interpolation. It is true that we need spatially distributed groundwater values (maps) for example for groundwater modeling, and further management works, and hence, preparing it by interpolation of point values is unwise and will have significant errors and uncertainties.
Line 192: some detail on this method is required, at least the equation
Reply: Penman-Monteith method is a well-known method in hydrology. We think citing it would be enough for the reader.
Line 199: How these data on the groundwater level are collected??
Reply: the detail about it is discussed in WTF section, but it is true that we have to put some information on how we collect them. We have given some details in the revised one.
Line 201: is it possible to show them in your map?
Reply: Yes, they are shown on Fig. 6.
Line 204: separate section for calibration
Reply: thank you, yes, we agree, and hence, we put the calibration details under the heading model calibration in the revised manuscript.
Line 206: is there no climate change over the years 1986-2014, and 2012 to 2013? Selection should be justified here.
Reply: yes, there might be climate change over 1986 to 2014. However, the objective here is to see how the land use change affects the major water balance components using an already optimized WetSpass model. yes, in reality, the values of the actual water balance terms for 186, 2000, and 2014 might be more or less than what is simulated in this paper due to climate change or due to variabilities in meteorological parameters, however, it is common to keep other factors constant to evaluate the effect one influencing factor. Thus, the effect of land use on the hydrological variables was evaluated keeping other variables constant. The reason why these specific years were used for the evaluation is already mentioned above (it is because of the availability of land uses, mapped based on the satellite images taken in those years).
Line 222: the limitation factor of what?
Reply: corrected, it is to mean one of the limitation of this study.
Line 231-232: should be explained
Reply: It is to mean the same codes that the WetSpass model uses for each land use class has been given for each class of land use map used in the model as an input variable. Otherwise, the model would not read it. Corrected accordingly in the revised version.
Line 247: is there fieldwork that has been done? should be detail in this case
Reply: There was no fieldwork for this part, however, the authors use their general area knowledge.
Line 278: what about Johnson (1967)? Provide some detail.
Reply: It is a literature compilation of different possible specific yield values for different geological materials (aquifers in this case). We added some explanation about it in the revised one.
Line 279: where is the formula?
Reply: the formula, how it is developed, and detail about it is found in the paper cited. It is an empirical formula. The equation is included in the revised version.
Line 281: is it possible to provide it a supplementary material?
Reply: Unfortunately, the pumping test analysis paper is written and being submitted to a journal as a continuous part of this study. Hence, providing here as a supplementary material may not be possible. Sorry, for the inconvenience in the follow-up of the studies. the pumping test paper has to be published before this manuscript submission.
Line 293: period of what?
Reply: the period that the total RF is collected. Corrected accordingly.
Line 296: how many groundwater samples are considered, sampling map is required.
Reply: 138 groundwater samples distributed all over the basin have been used. Ok, the groundwater sample map is included in the revised version.
Line 297: there is some confusion related to runoff characteristics and coefficients. Could you provide us more details.
Reply: As it is shown in equation 5, the CMB method takes into account the runoff amount because what is needed in the equation is the amount of the infiltrated water i.e. the rainfall minus the surface runoff amount. As it is difficult to get runoff amount at each groundwater sampling point, we subdivided the area based on studied runoff characteristics. Dessie et al. (2015), studied the runoff characteristics of the Lake Tana basin using a conceptual based hydrological model and calculated average runoff coefficients ( = surface runoff amount divided by rainfall amount) for each area. In their study, they classified the basin into southern, northern, eastern, and western catchments, based on the runoff coefficient values and characteristics. We used those coefficients and calculated the runoff amount needed in the CMB equation. Accordingly, the recharge amount for each sub-area is shown in table 4.
It was possible to get runoff values at each groundwater sampling point using our distributing runoff map (from the WetSpass model output) and perhaps calculation of recharge at each sampling point would seem possible. The reason that we did not do it is because our objective i.e. we need to compare the different methods' results independently. Besides, as we have taken average chloride values of rainfall samples for each sub-area, the chloride values at each groundwater sampling point would be less certain (our rainfall samples were not sampled at each groundwater sampling point).
Line 298: I do not have any idea about these four catchments. Should be detail.
Reply: Thank you, some descriptions about these catchments grouped based on runoff characteristics are added in the revised version of the manuscript. Also few descriptions about how runoff is calculated have been also added.
Line 330: maybe in different periods as you have only 60 wells.
Reply: About 138 groundwater samples for chemical analysis (chloride concentration) were considered in this study (not 60). Equation 5, shows chloride concentration in the rainfall is needed in the CMB method. However, the problem here is the chemistry of the rainfall is not constant temporally and spatially. The rainfall may have different sources: seawater or maybe also fresh water for which chloride concentration reaching at a specific area in different seasons or months or periods would be variable. In this study, we tried to consider such possible causes of uncertainty to a certain extent, for the spatial variation (as stated earlier, we grouped the catchments), and for the possible temporal variation, we tried to sample both the groundwater and the rainfall in the similar period so the chloride concentration in the groundwater is mostly from the sampled rainfall that infiltrated recently. At least being able to sample the groundwater which percolates from the rainfall that we sampled too will give a reasonable recharge rate. But if we, for example, sample the groundwater in June and the rainfall in August, perhaps there will be a high error in the recharge value estimated.
Line 341: it is better to start by Evap, runoff and after R. Including the precipitation is valuable.
Reply: the suggestion is accepted, and corrected accordingly
Line 346: why so include in the result and discussion part.
Reply: thank you for the comment, moved to the methodology section.
Line 352: is there analysis (e.g. statistical analysis) to prove this.
Reply: Unfortunately, there is no such a solid analysis for all the factors rather we used our physical observation of the maps. But we extract the recharge map by the different class combinations of soil, land use, slope, and rainfall amounts, and tried to see the attribute table of extracted raster maps of class combinations vs recharge rate. In doing so, we tried to judge which factor is the most important and gave rank in controlling the recharge amount. It is impossible to explain it in quantification.
Line 355: some confusing with this value of 0 mm.
Reply: at the water body recharge is assumed 0 (there is no infiltration beneath the floor of the water body).
Line 360: it looks no appropriate to consider the 0 mm at water bodies for determing an average.
Reply: I think we have to consider rather. Because we are reporting the average value of the basin including the lake body. It is one of the land use classes where the other water balance components are influenced on. Since we are comparing it with RF that rains over all parts of the basin, the one (even 0 value) should be also included in the average.
Line 386: As winter is the dry season, we expect to have evp more important than summer, but this is not the case of the present study. Why?
Reply: yes, winter is the dry season, and hence there is no enough rainfall as the water source for evapotranspiration, the less value in winter is reasonable. The source water for evapotranspiration is basically coming from rainfall plus the evaporation on the lake. The lake evaporation is happening throughout the year irrespective of being summer or winter (it only depends on the amount of potential evaporation). Even compared to the low amount of rainfall, the 44% of winter evapotranspiration is high. This is due to the evaporation taking place from the lake water surface during the long dry winter.
Line 395: validation section
Reply: Ok, we will shift it to the validation section.
Line 397: It is better to compare with other methods.
Reply: Yes, it is also compared with other literature values made by other methods (shown in table 2, and has been discussed in the text preceding the table).
Line 402: coarser time scale??
Reply: It means not for fine time scale like for daily or hourly but it is per seasonal. The term is common in GIS, RS, and other spatio-temporal modeling works.
Line 428: where is R2 in the figure?
Reply: It will be added in the revised one.
Line 443: 1:1 line??
Reply: It is called an 'identity line' or 'line of equality'. It is standardizing the axis to compare measured data with predicted data, or two different models. The starting and ending point of both axes should be the same (through the axis origin in our case).
Line 465: about name the caption preceding or following the description?
Reply: thank you, we corrected it according to the suggestions.
Line 473: rephrasing suggestion on the subtopic??
Reply: accepted, and corrected as per suggested.
Line 500: you used WTF for validation in the previous section. So, it is better to introduce this result before validation.
Reply: accepted, rearrangement is made.
Line 511: what do you mean by degree of weathering?
Reply: It means the rate of weathering. A term used to mention the different grades of rock weathering which in turn refers to the different physical strengths of lithological materials from rock to soil. Hydraulic properties like porosity, hydraulic property, transmissivity, specific yield, etc. depend on weathering effect.
Line 542: It is better to include the chemical analyses of Cl both for rainfall and the selected groundwater samples. The effective precipitation values used for calculation are also required.
Reply: kindly accepted, a map consisting of the groundwater sampling location (for 138 GW samples), with a label of its chloride values is presented in the revised version. Also, the Cl amount of the RF is presented in table 4.
Line 552: I do not understand what represents this value as you have several periods and catchments.
Reply: accepted, and corrected accordingly. These values are the mean annual values for the whole Lake Tana basin. They are the average values for the whole basin estimated by the three methods.
Line 616: This R2 is good but as can be seen in figure, there is some lag between simulated and observed values.
Reply: yes, there are overestimations of river discharge by the WetSpass model (except for three of the major rivers: Gumara, Gilgel Abay, and to a certain extent, Gelda rivers). The possible reasons have been discussed in Lines 426-441 of the old version. The possible reason would be the assumption of the method that we followed to calculate the total river discharge: as it is already stated, we assumed the total discharge is a summation of surface runoff (SR) plus groundwater recharge (GWR) that the model gives both spatial maps and as a single total average value for the whole basin. However, theoretically, river discharge can be lower than this summation when there is significant deep groundwater flow that flows through the subsurface (without emerging to the surface) i.e. part of the water goes without being measured by the river gauging station. Hence, there is a possibility that the measured river discharge value is less than the sum of SR and GWR. In other words, it is with a special assumption that baseflow is equal to groundwater recharge (when all the recharge water emerges to the surface and flows through the river flow measuring station). The real geological conditions for the different catchments are already mentioned with reference to previous studies in the discussion section.
Hence, it does not mean that the WetSpass model overestimates the river discharge but it maybe because of the assumption of the baseflow amount to be equal to groundwater recharge in the total river discharge calculation of this paper. As the WetSpass model does not give total river flow amount as an output, we have followed this procedure to equate it so that we can compare it with the measured river discharge amount measured at the river outlets.Citation: https://doi.org/10.5194/hess-2021-527-AC2 - Response to reviewer 1 review comments
-
AC1: 'Reply on RC1', Alemu Yenehun Beyene, 25 Jan 2022
-
RC2: 'Comment on hess-2021-527', Anonymous Referee #2, 26 Jan 2022
General
In this study Alemu Yenehun et al. estimate groundwater recharge and study the spatial and temporal recharge patterns for the Lake Tana region in Ethiopia. Three (well-) established recharge estimation methods are applied, a model approach using the physically based WetSpass model, the groundwater table fluctuation method, and the chloride mass balance method. I applaud the Authors for conducting such an extensive study in a research area where data acquisition I assume can often be challenging. I think the main contributions from this paper are 1) an improved understanding of the groundwater recharge fluxes in the case study area, and 2) the comparison of different recharge estimation methods. These contributions are valuable, as groundwater is an important source of drinking water in the region and a better understanding of the resource in this region would serve many. I found the manuscript generally well written (minor textual improvements are required), but some restructuring and changes to the figures might be required to improve the readability of the manuscript. The main concern I have for this manuscript are related to the methodology and the description thereof, discussed in detail below and in separate line comments. Additionally, while I think the study is worth publishing, I am unsure whether the contribution fits the scope of HESS (“substantial new concepts, ideas, methods, or data”) in its current form. In my opinion, it would fit much better in other journals where it can be submitted as a case study (e.g., Hydrogeology journal), and reviewed as such. If the Authors wish to publish in HESS, I think a more elaborate analysis and discussion of the uncertainties in recharge estimation could be done to better fit within the scope of the journal.
Methods description
From the description of the methodology and data following in this manuscript, it is hard to reproduce the results and gain a full understanding of the modeling procedure. This could partly be solved by sharing the scripts, input, and output data used in this study. However, often a more detailed description of the modeling procedure is also required. The WetSpass model requires substantial data input, which is not always clearly described in the manuscript. From the description of the data, it seems that different time periods were used to generate the input data for WetSpass. Perhaps this could be clarified using a table that summarizes the different data sources and time series characteristics (e.g., measurement interval, period, operator). Given the high temporal variability of the different hydrometeorological variables, I assume the same time periods for all variables are used, but I could not verify this from the current manuscript. A few times it is mentioned “expert judgement” is used, but it remains unclear what values were applied and why. The calibration process is only briefly described, and the calibrated parameters are not reported. I general, I think it would be good to rewrite the methods section with reproducibility in mind.
Uncertainties in recharge estimates
The Authors mention in the introduction that it is important to take uncertainty in recharge estimation into account (line 47), and thus I was expecting a more elaborate analysis or discussion of the uncertainty in recharge estimation methods applied in this manuscript. As the WetSpass model is manually calibrated, no parameter uncertainties are available. It therefore remains unknown how uncertain these recharge estimates are. This could be addressed by a sensitivity analysis or a more elaborate uncertainty analysis. A discussion of the limitations of the different methods and the uncertainties of the recharge estimates at the end of section 4 would be a welcome addition to the current manuscript.
Title
The title suggests a more general study on volcanic aquifers, while a case study is presented. I think it would be better if the title reflects the fact that it is a case study. Additionally, the title suggests that there are “point measurements” of groundwater recharge. In my view it be better to refer to these as ‘point estimates’, as they are empirical estimates from a recharge estimation method and not real measurements.
Line comments
L31: add a comma after floodplain
L72: regionalize
L77: were all these studies in a specific study area? If so, good to mention that.
L123: Perhaps refer to the GitHub repo: https://github.com/WetSpass
L123: No reference to Wang et al (1996) required?
L132/135/139: Sentences introducing these equations would be nice.
L133: Should there not be a change in storage term?
L142: Which evaporation equation was used?
L144: I think it would be good to state which physically based equations (e.g., Darcy) were used and elaborate a bit on the model (e.g., finite differences).
L147-156: This description is rather vague, what was done exactly? How were the values changed?
L157-162: This could be elaborated and made more specific. How many parameters were calibrated, how many data points were used for calibration, was there a formal objective function used or only visual goodness of fit?
L164-164: “to validate the recharge estimates from WetSpass.”
L165: I was a bit confused here, perhaps change to “During model calibration”.
L174: remove “our” or mention the operating organization
L177: remove “relatively”
L178-179: Add a reference for this statement
L183: Can mostly be specified?
L187-188: I think at least a reference would be appropriate here.
L191: The naming “so-called first-class stations” seems inappropriate here, I suggest rephrasing, also in the Figures.
L198-201: How many monitoring wells were used? What period were the groundwater tables observed? Does the period overlap with the discharge data and the meteorological data? Is this the same data as that mentioned in L306?
L205-206: One value per year, does that mean the model is calibrated to just a very few data points? How many rivers were in the dataset?
L231-232: Unclear what “appropriate” is here.
L279: How many pumping and slug tests were done? What values for Sy were found, are these reported or available somewhere?
L282: Perhaps a separate subheading for the CMB.
L306: So the recharge estimates from the WTF method are only representative for that period. Is this considered when comparing to the other methods?
L325: Perhaps this could be combined with Fig. 2. with the hydrogeological setting?
L336: Where is 4.1? Also, I suggest starting with 4.3 (verification), before discussing the water balance components.
L342: At this point I do not know for what period WetSpass computes the recharge. Perhaops write: “The annual recharge over the period 20XX-20XX..”
L348: minor typo.
L351: Change “Next to” to “Apart from”.
L360: minor typo
L366: values
L379: Exclude the lake in figure 7B. Perhaps add subplot with the precipitation. The legend in Fig7a has a mistake in the values. Why was a continuous coloring scheme not used (e.g., from yellow to blue), this would make the figures much easier to interpret. In the figure caption, specify “long term” by mentioning the exact years.
L407-447: I do not think this is a proper “Model verification”, as the same data is used for calibration! We cannot verify a model using the same data that was used for calibration. Perhaps this section can be renamed to “Calibration results”.
L427-429: A high R2 was obtained after the calibration, which may be interpreted as that the model can explain a large part of the observed discharge variation. However, Fig. 8 also clear shows a large systematic error between modelled and observed discharge, which I think could be more clearly stated. Contrary to what is stated in the text, the simulated discharge is always higher compared to the observations. This would be a good point to get back to in the discussion.
L448: Here the section “Model verification” starts in my opinion.
L448-463: Is there a reason the estimates from the CMB method were not used here?
L465: Figure 9 could be condensed/smaller.
L500-549: This section could be placed earlier, as the results described in this section were previously used to compare to WetSpass. Add “method” after WTF throughout the section.
L506: “has been taken”
L526: Change “to catch up” to “to capture”.
L557: Perhaps the higher range of values could be reported here.
L560-591: Just a suggestion. Perhaps the Authors can come with a nice plot that visualizes the different recharge estimates from all the other studies, and those from this study.
L595: minor typo.
L612: Perhaps I misinterpret, but I added the percentages of runoff (29%), recharge (22%), and evapotranspiration (53), and these do not add up to 100% (=104%). Why is this, are there model errors or changes in storage?
L640: Some recommendation/implications for future studies and work could be added at the end.
Citation: https://doi.org/10.5194/hess-2021-527-RC2 -
AC3: 'Reply on RC2', Alemu Yenehun Beyene, 22 Feb 2022
Response to Referee 2
General
In this study Alemu Yenehun et al. estimate groundwater recharge and study the spatial and temporal recharge patterns for the Lake Tana region in Ethiopia. Three (well-) established recharge estimation methods are applied, a model approach using the physically-based WetSpass model, the groundwater table fluctuation method, and the chloride mass balance method. I applaud the Authors for conducting such an extensive study in a research area where data acquisition I assume can often be challenging. I think the main contributions from this paper are 1) an improved understanding of the groundwater recharge fluxes in the case study area, and 2) the comparison of different recharge estimation methods. These contributions are valuable, as groundwater is an important source of drinking water in the region and a better understanding of the resource in this region would serve many. I found the manuscript generally well written (minor textual improvements are required), but some restructuring and changes to the figures might be required to improve the readability of the manuscript. The main concern I have for this manuscript are related to the methodology and the description thereof, discussed in detail below and in separate line comments. Additionally, while I think the study is worth publishing, I am unsure whether the contribution fits the scope of HESS (“substantial new concepts, ideas, methods, or data”) in its current form. In my opinion, it would fit much better in other journals where it can be submitted as a case study (e.g., Hydrogeology journal), and reviewed as such. If the Authors wish to publish in HESS, I think a more elaborate analysis and discussion of the uncertainties in recharge estimation could be done to better fit within the scope of the journal.
Reply: Thank you so much. We have tried to do an uncertainty analysis of the model parameters and have added it in the revised version.
Methods description
From the description of the methodology and data following in this manuscript, it is hard to reproduce the results and gain a full understanding of the modeling procedure. This could partly be solved by sharing the scripts, input, and output data used in this study. However, often a more detailed description of the modeling procedure is also required. The WetSpass model requires substantial data input, which is not always clearly described in the manuscript. From the description of the data, it seems that different time periods were used to generate the input data for WetSpass. Perhaps this could be clarified using a table that summarizes the different data sources and time series characteristics (e.g., measurement interval, period, operator). Given the high temporal variability of the different hydrometeorological variables, I assume the same time periods for all variables are used, but I could not verify this from the current manuscript. A few times it is mentioned “expert judgement” is used, but it remains unclear what values were applied and why. The calibration process is only briefly described, and the calibrated parameters are not reported. I general, I think it would be good to rewrite the methods section with reproducibility in mind.
Reply: Yes, indeed, a table containing the different data sources and time period of the WetSpass model input variables will be added in the revised final manuscript. It is true that for some of the model input variables, the data for different time periods have been considered. As the model is using long-term averages of the different variables as an input (in our case: long-term average summer and winter), hence, observing long-term time-series temporal variations of the water balance components with the WetSpass model is impossible. The model is limited as it is simulating only long-term monthly averages as a fine time scale (time series results are not possible). In other words, the model is basically developed to simulate long-term average monthly/seasonal conditions, and most importantly the spatial variations. The model is basically developed for spatial simulation of groundwater recharge (which is later coupled with groundwater modeling-MODFLOW) whose short -time (day to day or hour to hour variations) is less important compared to the other water balance terms.
Except for the groundwater level map (which is based on recently collected data), meteorological variables (rainfall, average temp., wind speed, all meteorological parameters used to calculate PET using the Penman-Monteith method) were more or less in similar time periods. Yes, it is true that the meteorological variables have high temporal variations (most importantly the seasonal or monthly, or daily variations); year-to-year variation is relatively small. In other words, there is a small variation for a given season or month for different years. Given the objective of this paper i.e. the spatial variation of recharge and to a certain extent other water balance terms, these would not be a major problem.
The other non-meteorological (physical) parameter that changes with time is the land use. For the long-term average data (2005 to 2018), we used the land use map using the 2014 satellite image and ground truth data, for the modeling; land use of 1980 and 2000 for the evaluation of the effect of land use change on the water balance terms. The use of one-time data for the land use and long-term average for the other parameters might have slightly departed from reality and may have caused some uncertainties. However, the model is not robust enough to use different land uses for calibration. Furthermore, the change in land use is not so fast, and hence, the land use of 2014 might have represented more or less the possible land use status of the area during 2005-2018.
Explicit statements for what do we mean by “expert judgment” will be added in the revised version. We mean some adjustment for the parameters’ values on the default land use parameter table has been made, for example, the root depth of forest land is changed. In the model parameter table, each land use type has given impervious, bare land, vegetation, and open water percentages. In the default parameter of the WetSpass model the vegetation area coverage for the bush and shrubland is 100%. This is based on the observation of the land use class type in the temperate zone (the Netherlands and Belgium). However, the Bush and Shrubland use in the Ethiopian (tropics) context is different: the vegetation is sometimes sparsely distributed, and is with some bare land component, during our field verification on the land use type, we tried to guess (using parcels of land on which we do some measurement), and came up with about 10% is bare land and the rest 90% is consisting of vegetation. Similarly, adjustments have been made by Gebreyohannes et al. (2013) during their application of the model for Geba catchment, in northern Ethiopia. Similarly, the sub-afro-alpine vegetation land use type found in our area (consisting of about 0.3% of the total area coverage) is not present in the default land use classes of the WetSpass model. However, we made an equivalent with wet meadow land use type, and following a similar procedure we modified the land use percentage to 80% vegetation and 20% bare land (it was 100% vegetation for wet meadow land cover type in the model parameter table.
<< insert table1: the model land use parameter table here, attached in the supplement file>>
The steps followed in the calibration process, and the table for the calibrated parameters will be added in the revised version of the manuscript.
The modeling section of the methodology is more elaborately written in the revised version. However, to reduce the pages of the manuscript, all the detailed equations, and the empirical formulas developed and followed in the WetSpass model are better not to be included, we would rather cite the papers that developed the model first and improved further.
Uncertainties in recharge estimates
The authors mention in the introduction that it is important to take uncertainty in recharge estimation into account (line 47), and thus I was expecting a more elaborate analysis or discussion of the uncertainty in recharge estimation methods applied in this manuscript. As the WetSpass model is manually calibrated, no parameter uncertainties are available. It therefore remains unknown how uncertain these recharge estimates are. This could be addressed by a sensitivity analysis or a more elaborate uncertainty analysis. A discussion of the limitations of the different methods and the uncertainties of the recharge estimates at the end of section 4 would be a welcome addition to the current manuscript.
Reply: The WetSpass model has different parameters and needs input variables to give a reasonable output for the understanding of water balance components of a given river basin/watershed. Yes, to evaluate the efficiency of the WetSpass model applied in this study, the sensitivity analyses of all parameters and input variables would be a good addition of the paper. A simple sensitivity analysis is carried out and added in this revised version of the manuscript. Since the objective of doing this sensitivity analysis is to know the possible uncertainties that might be incorporated into the estimated water balance values, we did a sensitivity analysis for the calibration of the global parameters of the WetSpass model. These are alfa coefficients (it is a parameter adjusting soil moisture value), interception coefficients (adjusting interception by the plant canopy), LP coefficient (parameter adjusting evapotranspiration), rainfall intensity, and x coefficient (runoff delay factor). The other physical factors (called local parameters) such as the contribution of the slope (40%), landuse (30%) and soil (30%) in determining the balance components share, are originally developed from literature (experimental research). Thus, we kept them as they are (we have taken the default values of the model), and hence, we have not done a sensitivity analysis though their local variation might have caused some uncertainties in the estimated recharge and other water balance terms. Furthermore, we do not perform sensitivity analysis of the meteorological variables used in the model perhaps they might cause some uncertainties during measurement, interpolation, etc. They were fixed variables in our calibration similar to the physical parameters mentioned above.
Among the global model parameters calibrated in the model, the most sensitive global model parameter is RF intensity. On average, for every 1 mm/hr increase of the intensity, there is a mean annual recharge increment of 38.2 mm. Fortunately, we measured the rainfall intensity at some of the meteorological stations that the BDU-IUC project (the project that sponsors this study) established (using the automatic rain gauges), and hence, we used the range of these data in our calibration.
<<insert Figure 1 attached in the supplementary file here>>
Figure 1. Mean annual groundwater recharge (mm) versus average rainfall intensity (mm/hr.) for the different possible intensity values. The model was running for these RF intensity values by keeping the other model parameters constant.
In general, except for the RF intensity (which highly depends on the rainfall characteristics of an area), the other global model parameters have given a possible range of values in the WetSpass model (LP coefficient and x coefficient: 0.1 to 1; and alfa coefficient and a interception 1 to 10). Hence, the sensitivity testing for values within these ranges has been performed.
For every variation in LP coefficient, there is a 4 mm increase or decrease of the annual recharge.
<<insert Figure 2 attached in the supplementary file here>>
Figure 2. Mean annual groundwater recharge (mm) versus LP coefficient for the different possible LP coefficient values.
On average, for every 1 value variation for alfa coefficient there is only about 2.3 mm variation in annual recharge rate.
<<insert Figure 3 attached in the supplementary file here>>
Figure 3. Mean annual groundwater recharge (mm) versus Alfa coefficient for the different possible Alfa coefficient values.
The variation in a interception is small and only for the initial values. For the relatively higher values, the recharge is almost not changing, and hence, the recharge can be considered as insensitive for the variation of a interception. This has been also shown with the low R2 value.
<<insert Figure 4 attached in the supplementary file here>>
Figure 4. Mean annual groundwater recharge (mm) versus a interception coefficient for the different possible a interception coefficient values.
The groundwater recharge is not changing for the possible changing intervals of the x-coefficient, and hence, the x-coefficient is the most insensitive parameter among all the global model parameters for the Lake Tana basin.
<<insert Figure 5 attached in the supplementary file here>>
Figure 5. Mean annual groundwater recharge (mm) versus x coefficient for the different possible x coefficient values.
The limitation and the sensitivity analysis which would bring uncertainties to the reported values have been added in the revised version.
Title
The title suggests a more general study on volcanic aquifers, while a case study is presented. I think it would be better if the title reflects the fact that it is a case study. Additionally, the title suggests that there are “point measurements” of groundwater recharge. In my view it be better to refer to these as ‘point estimates’, as they are empirical estimates from a recharge estimation method and not real measurements.
Reply: Thank you so much. Yes, indeed, they are point estimates. For example, the variables in the calculation of the WTF method are the change in the water level and the specific yield (Sy). The water level is based on fine-time scale measurement, however, there is estimation for Sy though it is also analyzed from slug and pumping tests (real measurements), estimation from Johnson (1967) literature compilation has been also used. Given the high aquifer variability (compared to the representativity of each test well), it is true the WTF values are also estimations. Thus, the topic is corrected accordingly.
Yes, indeed the methods are tested in the case study area, however, we do believe the evaluation part that recommends the effective application of the methods for aquifers lying at different topographical and hydrogeological conditions would make the study beyond being case study, and will attract global readers.
Line comments
L31: add a comma after floodplain
Reply: thank you, added.
L72: regionalize
Reply: accepted, and corrected accordingly.
L77: were all these studies in a specific study area? If so, good to mention that.
Reply: mentioned, and thus corrected accordingly.
L123: Perhaps refer to the GitHub repo: https://github.com/WetSpass
Reply: thank you so much. Cited accordingly.
L123: No reference to Wang et al (1996) required?
Reply: Wang et al (1996) is about the WetSpa model. Maybe it is a good idea to add it as WetSpa is the starting point and the basis for the development of WetSpass. Hence, it is added in the new revised version.
L132/135/139: Sentences introducing these equations would be nice.
Reply: thank you, added.!
L133: Should there not be a change in storage term?
Reply: Yes, in the WetSpass model, the storage term is neglected. There are two ways of incorporating change in storage to the model on a seasonal basis. The first instance is that the plant available soil moisture reservoir in the summer is assumed to be filled up while it can be depleted in the model in the winter (for areas like Lake Tana basin); the second case is that in the model a different groundwater depth can be used for winter and summer. In our methodology, we followed the second option, we had prepared the groundwater table map for the summer and winter seasons using the time-series groundwater level monitoring data that we used for our study. This is explained in the methodology part of the manuscript. However, since we have Lake Tana in our study area, there is a possibility of surface water storage change. For example, the evaporation from the open lake water may cause the summation of annual AET + Runoff + Recharge to be greater than the annual RF i.e. RF < AET + runoff + groundwater recharge. Hence, there is a possibility that the storage term changes. However, change in storage in the WetSpass model is not as important as the other hydrological models, which can simulate water balance components at fine-time series (such as a daily or hourly). For such models, storage amount (variation) at each time-series highly affects the water balance components of the following time-series value. E.g. the storage change of day 1 affects the water balance values for day 2. But in the WetSpass model applied in this study (seasonal), there are only two time-series whose storage difference (especially the groundwater storage) by an independent water table map.
L142: Which evaporation equation was used?
Reply: the Penman equation is used for evaporation in WetSpass. It is discussed already in the result and discussion section.
L144: I think it would be good to state which physically-based equations (e.g., Darcy) were used and elaborate a bit on the model (e.g., finite differences).
Reply: In general, the WetSpass model is based on known hydrological equations (conceptually and physically meaningful equations), and also some empirical equations developed from works of literature. Yes, a better-elaborated explanation of the model background will be added in the revised version of the manuscript.
L147-156: This description is rather vague, what was done exactly? How were the values changed?
Reply: In the revised manuscript, detailed changed values of the land use parameters are explained, as it is already explained above. Basically, the changes are in the percentage of bare land and vegetation area percentage of a given land use type (bush and shrubland and sub-afro-alpine grassland), and also root depths of forest land (it is already explained in the paragraph).
L157-162: This could be elaborated and made more specific. How many parameters were calibrated, how many data points were used for calibration, was there a formal objective function used or only visual goodness of fit?
Reply: there is no formal objective function used, only the visual goodness of fit has been used. Basically, about five global parameters have been used for calibration (discussed above). The final optimal values (after intensive manual calibration trials) have been presented in a table in the revised manuscript. The goodness of fit was compared between the accumulated mean annual river discharge amount (simulated surface runoff +baseflow) vs the measured river discharge amount of the seven major rivers.
L164-164: “to validate the recharge estimates from WetSpass.”
Reply: validation is changed by evaluation, the word might have been confusing.
L165: I was a bit confused here, perhaps change to “During model calibration”.
Reply: thank you, changed accordingly.
L174: remove “our” or mention the operating organization
Reply: thank you. It is changed by the BDU-IUC project funding
L177: remove “relatively”
Reply: removed
L178-179: Add a reference for this statement
Reply; thank you, added.
L183: Can mostly be specified?
Reply: Yes, it has been done in the revised one.
L187-188: I think at least a reference would be appropriate here.
Reply: Ok, will be done.
L191: The naming “so-called first-class stations” seems inappropriate here, I suggest rephrasing, also in the Figures.
Reply: Ok, corrected accordingly!!
L198-201: How many monitoring wells were used? What period were the groundwater tables observed? Does the period overlap with the discharge data and the meteorological data? Is this the same data as that mentioned in L306?
Reply: Yes, they are the same data listed on L306. The monitoring wells were 65 in number. The measurements were taking place from 2017 to 2019: some using automatic water level measuring device (measuring every half an hour), and others were manually (per each day).
About data overlap: partly it overlaps with meteorological data. However, the meteorological data at most of the stations are long-term averages (2005 to 2018), generally, it is impossible to say the data exactly overlaps. However, since it is long-term average seasonal values that we are finally used as input, it will not be a major problem. The intention of the modeling work here is not primarily to see the long-term temporal variation of recharge and other water balance terms (but limited to mean seasonal variations). Similarly, the river discharge data used for calibration is not in the same time interval (2012-2013, mentioned on line 204). Nevertheless, as it is the average seasonal values that are applied, it will/would not have significant variation.
L205-206: One value per year, does that mean the model is calibrated to just a very few data points? How many rivers were in the dataset?
Reply: using 7 river gauging data at 7 major rivers of the basin (shown in Fig. 8). It is the seasonal values which are used for calibration (summer: June – September, and winter: October to May).
L231-232: Unclear what “appropriate” is here.
Reply: It is to mean code matching between the model parameter tables and land use classes has been made so that the model reads the right land use.
L279: How many pumping and slug tests were done? What values for Sy were found, are these reported or available somewhere?
Reply: This pumping test analysis is our next paper (already prepared and will be submitted to a journal soon). A test on 31 hand-dug wells was successfully executed and analyzed (the statement is added in the new revised version). Yes, the results per aquifer type were reported already on lines 507 to 515 of the old version.
L282: Perhaps a separate subheading for the CMB.
Reply: Ok, accepted.
L306: So the recharge estimates from the WTF method are only representative for that period. Is this considered when comparing to the other methods?
Reply: As we have mentioned earlier, the recharge estimated in all of the methods is mean annual. It is true that the meteorological variables used in the WetSpass model, and perhaps the land use have variations over the years. However, the annual values of these meteorological variables used as input for WetSpass are more or less constant. However, day-to-day variations of these parameter values especially the rainfall is high and would have resulted in high daily variations of water balance components such as recharge. Unfortunately, the WetSpass cannot simulate such fine time-series variations except the long-term seasonal variations.
Comparing the long-term recharge amount simulated by the WetSpass and the WTF for these periods would not be much of a problem (as it is the mean annual which is being compared). Similarly, the CMB is calculated only for the groundwater and rainfall sampling years (2017-2018).
The effect of land use change is evaluated in this paper and found small.
Thank you, we will put this fact as one of the limitations of this research.
L325: Perhaps this could be combined with Fig. 2. with the hydrogeological setting?
Reply: Yes it is possible to combine. However, it would be too earlier to put it here for the readers to follow the methodology section.
L336: Where is 4.1? Also, I suggest starting with 4.3 (verification), before discussing the water balance components.
Reply: Thank you so much, we corrected both according to your suggestions.
L342: At this point, I do not know for what period WetSpass computes the recharge. Perhaps write: “The annual recharge over the period 20XX-20XX..”
Reply: The WetSpass model computes from 2005 to 2018. We added the phrase in the manuscript.
L348: minor typo.
Reply: corrected accordingly.
L351: Change “Next to” to “Apart from”.
Reply: thank you, have done.
L360: minor typo
Reply: thank you, corrected accordingly.
L366: values
Reply: corrected.
L379: Exclude the lake in figure 7B. Perhaps add subplot with the precipitation. The legend in Fig.7a has a mistake in the values. Why was a continuous coloring scheme not used (e.g., from yellow to blue), this would make the figures much easier to interpret. In the figure caption, specify “long term” by mentioning the exact years.
Reply: thank you for the comments. It is not the lake which is mentioned in Fig.7b, rather the value. There is by far a higher AET value than in the other areas (since it is a water body where open water evaporation operates). That is why the contrast between the lake and the other land uses is higher, which causes the values at the lake to have a similar colour.
Yes, we will remove the lake in figure 7c. The mistake in the values of the figure 7a legend might be the 0 value (I guess) over the water body. WetSpass assumes that there is no recharge for the aquifers under the floor of the water body.
We have checked and corrected the maps according to the suggestions.
L407-447: I do not think this is a proper “Model verification”, as the same data is used for calibration! We cannot verify a model using the same data that was used for calibration. Perhaps this section can be renamed to “Calibration results”.
Reply: in the WetSpass model, there is no way that we insert calibration data (river discharge data) in the model itself during calibration. We were running the model (more than a hundred times) by changing the global model parameters, and we have been cross-checking the river discharge values of the basin. After doing it for several times, the optimal values (i.e. the closest values between the simulated and measured discharge data) were reached, and changing the values for the modeling parameters reached to a point that changing values no more improved the R2 between measured and simulated values. However, this is mostly between the total annual river discharges of all the rivers together. Hence, comparing the discharge values of the simulated and the measured after extracting the values (in raster map) by each river catchment can be partly considered as validation. However, as you mentioned, it will give full confidence if we discuss it under calibration rather than validation. For validation, the one estimated by the WTF and CMB can be still kept on.
L427-429: A high R2 was obtained after the calibration, which may be interpreted as that the model can explain a large part of the observed discharge variation. However, Fig. 8 also clearly shows a large systematic error between modelled and observed discharge, which I think could be more clearly stated. Contrary to what is stated in the text, the simulated discharge is always higher compared to the observations. This would be a good point to get back to in the discussion.
Reply: Thank you for your concern over the issue. As you stated, generally simulated values are higher than the observed ones. The possible reasons have been discussed in this section. The possible reason would be the assumption of the method that we followed to calculate the total river discharge: as it is already stated, we assumed the total discharge is a summation of surface runoff (SR) plus groundwater recharge (GWR) that the model gives both in spatial maps and as a single total value. However, theoretically, river discharge can be lower than this summation when there is significant deep groundwater flow that flows through the subsurface (without emerging to the surface) i.e. part of the water goes without being measured by the river gauging station. In other words, baseflow is not always equal to recharge, there are cases that baseflow is much lower than the recharge. Hence, there is a probability that the river discharge value is less than the sum of SR and GWR. The real geological conditions for the different catchments are already mentioned with reference to previous studies in the discussion section.
In general, it does not mean that the WetSpass model overestimates the river discharge but it maybe because of the assumption of the baseflow amount to be equal to groundwater recharge in the total river discharge calculation of this paper. As the WetSpass model does not give total river flow amount as an output, we followed this procedure to equate it so that we can compare with the measured river discharge amount measured at the river outlets.
L448: Here the section “Model verification” starts in my opinion.
Reply: Ok, we edited accordingly.
L448-463: Is there a reason the estimates from the CMB method were not used here?
Reply: Because the CMB is done for the different river catchments (by subdividing the river catchments into eastern, southern, northern, and western catchments, based on the similarity and difference in the runoff characteristics) rather than calculating at each sampling point. We took the average chloride concentration sampled and analyzed for the different major river catchments (Table 4) (chloride for both groundwater and rainfall), for doing the calculation (as mentioned already in the methodology section). It is impossible to get effective precipitation (rainfall minus surface runoff at each sampling point, which is one of the important variables in the CMB calculation). Hence, it is not easy to extract values from the recharge raster map produced by WetSpass at each sampling point like for the WTF method. However, we have a discussion part about it in the section (comparing the result with other methods) “summarizing groundwater recharge estimations by the different methods and comparing with other similar studies”.
In the validation section, we added a discussion with CMB as well in the new revised version.
L465: Figure 9 could be condensed/smaller.
Reply: accepted, we have done it.
L500-549: This section could be placed earlier, as the results described in this section were previously used to compare to WetSpass. Add “method” after WTF throughout the section.
Reply: Ok, thank you, rearranging has been done.
L506: “has been taken”
Reply: inserted
L526: Change “to catch up” to “to capture”.
Reply: changed
L557: Perhaps the higher range of values could be reported here.
Reply: It is to mean if the CMB calculation had been made at each groundwater sampling point (if it was possible to do so), the spatial variation of recharge would have been more than what has been calculated and reported i.e. average values over each catchment.
L560-591: Just a suggestion. Perhaps the Authors can come up with a nice plot that visualizes the different recharge estimates from all the other studies, and those from this study.
Reply: yes, we will show it either in a table or cross-plot by summarizing the studies done at different sub-catchments in the Lake Tana basin.
L595: minor typo.
Reply: corrected
L612: Perhaps I misinterpret, but I added the percentages of runoff (29%), recharge (22%), and evapotranspiration (53%), and these do not add up to 100% (=104%). Why is this, are there model errors or changes in storage?
Reply: This is already discussed above as a reply to your general comments. This might happened due to two reasons:
1) some of the recharged water has also transpired especially in the forest area (deep-rooted eucalyptus trees considered in the root depth parameter table (5.5 m depth), which is below the water level, especially for the summer season). The WetSpass adds this transpiration amount from the groundwater to the calculated total AET value, it is why the water table raster map is needed as an input file.
2) The AET at the open water (the Lake Tana and the Wetlands) is equal to the PET, which is higher than the RF, hence there might be some imbalance at, for example, the Lake Tana. This is also reported by different authors as a closure error.
L640: Some recommendation/implications for future studies and work could be added at the end.
Reply: Yes, recommendations suggesting how researchers and water management practitioners use this research output, and limitations of the study mainly method assumptions and limitations are added in the revised version.
-
AC3: 'Reply on RC2', Alemu Yenehun Beyene, 22 Feb 2022
Status: closed
-
RC1: 'Comment on hess-2021-527', Anonymous Referee #1, 23 Dec 2021
The paper reports regional study based on physically-based water balance WetSpass model, which has been applied to simulate the water balance components, including runoff, evapotranspiration and groundwater recharge, at Lake Tana basin in Ethiopia. Further, the spatial groundwater recharge map by WetSpass was validated through water table fluctuation and chloride mass balance methods. The topic is interesting. However, the paper under its current form suffers from several points concerning its content and structure. I’m not English native, but it looks that the paper needs bit English improvements to reach high written level required by high-tier international journal like HESS. The Introduction section looks too modest. Usually, the Introduction should provide a thorough literature review on the topic, starting from the general to the specific and clearly identifying the gap(s) in current knowledge that the paper addresses. This allows to formulate explicitly research questions, hypothesis, and the contribution of the study. The authors give general info on the recharge at the beginning followed by introducing the characteristics of the Ethiopian volcanic basin. However, it is not clear what is the issue. Novelty and justification seem the main issue of the paper. There is unclear scientific rationale on why the authors are pursuing the proposed study. The literature review seems does not specify clearly research gap(s) of previous studies. The authors stated that ‘‘Different methods are developed for recharge estimation. However, choosing appropriate methods is often challenging’’. This directs the reader to consider that novelty will be ‘‘developing recharge choosing tool’’, but it is not the case. Rather, the paper is oriented towards regional study focusing on groundwater recharge estimating and mapping. It is not clear what is the paper contribution compared to previous recharge studies conducted on the study area (e.g., Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005). Other points when the authors mention that ‘‘the area has a climate with long dry winters and short rainy summer seasons. Hence, the groundwater recharge-discharge processes are expected to vary highly both spatially and temporally’’. I believe that with only two seasons and same type of precipitation (rainfall) at the study area, there are other areas around the world presenting more complicate highly spatiotemporal variation such as the case of the humid northern regions dominated by variety of meteorological inputs or the subhumid regions submitted to many different seasons. My question, where is the situation of the present study compared to other recharge studies undertaken under different climatic conditions. Also, the authors indicated that ‘‘One of the challenges for the point recharge estimation methods is their incapability to estimate it in a spatially distributed way’’. I respectfully disagree with this statement as one of the main advantages stemming from integrating remote sensing with GIS within recharge mapping is in the capability to investigate the recharge at unprecedented levels of spatiotemporally variability. However, I agree that only limited studies have incorporated the spatial distribution of recharge with corresponding rates. Among them the references cited by the authors (e.g., Batelaan and De Smedt 2007). However, there is unclear scientific rationale on why the authors are pursuing the proposed study when they adopt the WetSpass model. The literature review seems does not specify clearly research gap(s) of previous studies related to the WetSpass model and the novelty compared to the previous WetSpass studies. If just using this model, the proposed study should be simple case study that can be fit with other journals like Hydrology Journal: Regional Studies. This seems confirmed with the main objectives cited at the end of the Introduction section. When the authors mention ‘‘studying the recharge mechanisms’’, what is the contribution relative to the existing work of Yenehun (2020) cited in lines 119-120 of the paper. For the methodology (Section 3), I raised some comments which are integrated in the attached document. My main concern is the structure of this section which seems confusing between methods and results. The authors introduce some results (e.g., developed maps) that should be placed in the Results section. Also, many methods are suffering from less details and explanations, and often the authors introduce some information without previous definition. This approach complicates the understanding of the methodology, while it looks appropriate to introduce a flowchart for the model development and explain how each equation, input or parameter has been adopted, calculated, or assumed. In many places of the text, the authors based on their own knowledge, but less detail is introduced. This cannot allow reader to judge this knowledge and understand the background of this knowledge. The authors provided criticism about interpolation in the Introduction section ''it is unwise to extrapolate or regionalized the result by the conventional point recharge estimation techniques''. However, many concepts (e.g., maps) of the proposed paper are based on the interpolation approach. In separate locations, the authors mention some ways that have been adopted for the calibration phase, it will be better to provide a sole section introducing all the calibration process rather to be distributed over several places in the text. I formulated many specific comments in the attached document, but I feel the paper suffers from a lack of novelty and structure. Regarding the raised comments, I believe the paper is not suitable for publication and I suggest directing it to other focusing journals.
-
AC1: 'Reply on RC1', Alemu Yenehun Beyene, 25 Jan 2022
Response to the general comments forwarded by the referee:
English language improvements and corrections have been made including suggestions made by you in the revised version.
In the general comment as well as in the specific comments (in the attached manuscript of the referee), clarity about the objectives has been raised as the main limitation of the manuscript. This could be due to problems of us in explaining the main objectives as explicitly as possible which we thought we have done in the revised version that we will attach.
The main aims of this paper are:
Each recharge estimation technique has its own limitations, depending on the recharge mechanisms, the aquifer topographical setting, aquifer type (for example fracture vs porous soil aquifers, confined vs unconfined), aquifer geometry, etc. Hence, the effectiveness of recharge estimation techniques is varying, and thus in this study evaluation of the hydrological model (WetSpass) has been made (i.e. validation with measured point values though some variables are estimated in these methods too!!) for the Lake Tana basin having high topographic, variable slope, and comprising different volcanic aquifers. Such aquifer types are found in many parts of Ethiopia and is the major groundwater resources for the global water need. The WetSpass model applied in this study is a physically-based distributed hydrological model which gives spatially distributed spatial water balance terms including recharge. Yes, there are statistical interpolations for some of its input variables but the final spatial recharge and other water balance output maps are different from simply measured point value interpolations, rather they have resulted after optimal global model parameters are set through the calibration process. Hence, one of the objectives of the paper is evaluating the hydrological model, and thus giving suggestions for recharge methods to be better applied for different aquifers lying at different topographical settings: this has been clearly put in the discussion and conclusion section of this manuscript. The paper put concluding remarks, where physical hydrological models, would be effectively applied. Why they are less effective for some aquifers and more for others. This helps to take into account the accuracy of recharge values in the area-specific water management plans and decisions being made at different levels.
The next aim of this manuscript is to produce a spatially distributed groundwater recharge rate map for the Lake Tana basin, which is the source of the transboundary Blue Nile River basin where a high tension of hydropolitics is currently affecting the entire region. The spatially distributed recharge rate map produced in this study for the Lake Tana basin is an important output for further groundwater modeling, and water management issues of the basin.
The other one is identifying the most important hydrometeorological and physical factors and prioritizing them for groundwater recharge distribution (qualitatively) for the study basin, and so for similar study areas, including the effect of land use change on the recharge and other hydrological terms. These have been discussed in the result and discussion section and concluded in the conclusion part.
With due you respect, these objectives, especially the first one are important for the international readers. Many hydrological models for different areas (small catchment-scale to large basins) are developed for estimating groundwater recharge. However, evaluation of the method with point estimations based on direct water level measurement (WTF) and chemical tracer (CMB) gives a good insight for future recharge estimation techniques for similar aquifer types wherever they are located in the world. In this study, identifying topographical and geological characteristics, and thus the recharge mechanisms have found an important factor and starting point in selecting recharge estimation methods. The study also pointed out that the common approach that is being implemented i.e. calculating recharge by a multitude of methods and averaging out the results of the different methods is found unreliable. Rather selecting appropriate one or few technique/(s) and considering that as the optimal result is recommended.
Evaluating the groundwater recharge estimation mechanism for the volcanic aquifers lying at different topographical settings, that represents a vast major part of the world groundwater aquifers, and able to suggest more appropriate method/(s) will benefit different further similar studies and researchers. Hence, this study can be seen as a dual purpose paper: evaluating the physical-based hydrological model (and so other similar models), and giving a spatial recharge rate map for the important Lake Tana basin.
Thank you (the reviewer) for your valuable comments, in the new version, shortcomings in the introduction section have been improved as per the suggestions. The research questions are outlined and the contribution of the research output is mentioned in the new version.
The reason why we talk about “choosing appropriate methods” is to direct our reader that evaluation of a method will be one of the objectives of this manuscript. The reason why different methods were developed for recharge estimation rather than using the simplest and cost-effective method is due to the fact that different aquifers have different recharge mechanisms for which some methods are more effective than others. It is why we did evaluation and validation of the existing method. Furthermore, when we say point recharge estimation, we are referring to recharge estimation techniques that estimate at a given point (e.g. on groundwater well, spring, etc.) such as WTF and CMB methods. Otherwise, yes indeed, we agree with the advance of Remote sensing and GIS, and with the developing capability of the grid-based spatially simulating hydrological models, producing spatially distributed water balance maps is so possible and being widely applied. The applied WetSpass model for this study is one of such kinds, and we only applied an existing model (nothing new is done in this study as far as new methodology is concerned). As we have aforementioned, we pursued this study, for evaluating the physically-based hydrological models for our study basin so that areas having similar hydrogeological characteristics will take into account the suggestions of our evaluation result for future application. Besides, the recharge rate map for the basin, which is an important research product for groundwater exploitation practices and plans, will help groundwater managers and policymakers in this important area which is regarded as one of the growth corridors of the country. Besides, the recharge map is also an important input for further groundwater modeling work.
With due respect, the studies by (Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005) are not all about recharge estimations studies. It is only Demlie et al., 2008, who did a recharge estimation only using the CMB method on a volcanic aquifer for the Akaki catchment located at about 550 km from the Lake Tana basin. The others had made different hydrogeological studies but had put general statements about the recharge variability across the volcanic aquifers. Thus, we can make clear that the presented study in this paper is different from the studies made by these authors.
Some maps in the methodology are shifted to result section as suggested. The shortened explanation of the methodology section is for the reason of shorting the paper. We thought citing the papers that develop the original methodologies and latter modifications will be enough and give detailed background knowledge including the different mathematical equations. However, parameter value modifications that have been done in this paper are discussed and justified in the new version of this paper.
Citation: https://doi.org/10.5194/hess-2021-527-AC1 -
AC2: 'Reply on RC1', Alemu Yenehun Beyene, 22 Feb 2022
- Response to reviewer 1 review comments
Response to the general comments:
English language improvements and corrections have been made including suggestions made by you in the revised version.
In the general comment as well as in the specific comments (in the attached manuscript of the referee), clarity about the objectives has been raised as the main limitation of the manuscript. This could be due to problems of us in explaining the main objectives as explicitly as possible which we thought we have done in the revised version that we will attach.
The main aims of this paper are: - Each recharge estimation technique has its own limitations, depending on the recharge mechanisms, the aquifer topographical setting, aquifer type (for example fracture vs porous soil aquifers, confined vs unconfined), aquifer geometry, etc. Hence, the effectiveness of recharge estimation techniques is varying, and thus in this study evaluation of the hydrological model (WetSpass) has been made (i.e. validation with measured point values though some variables are estimated in these methods too) for the Lake Tana basin having high topographic, variable slope, and comprising different volcanic aquifers. Such aquifer types are found in many parts of Ethiopia and is the major groundwater resources for the global water need. The WetSpass model applied in this study is a physically-based distributed hydrological model which gives spatially distributed water balance terms including groundwater recharge. Yes, there are statistical interpolations for some of its input variables but the final spatial recharge and other water balance output maps are different from simply estimated point value interpolations, rather they have resulted after optimal global model parameters are set through the calibration process (detail explanation about this issue can be found in the response to the line by line comments). Hence, one of the objectives of the paper is to evaluate the hydrological model, and thus give suggestions for recharge methods to be better applied for different aquifers lying at different topographical settings: this has been clearly put in the discussion and conclusion section of this manuscript. The paper put concluding remarks, where physical hydrological models, would be effectively applied. Why they are less effective for some aquifers and more for others. This helps to take into account the accuracy of recharge values in the area-specific water management plans and decisions being made at different levels.
- The next aim of this manuscript is to produce a spatially distributed groundwater recharge rate map for the Lake Tana basin, which is the source of the transboundary Blue Nile River basin where a high tension of hydropolitics is currently affecting the entire region. The spatially distributed recharge rate map produced in this study for the Lake Tana basin is an important output for further groundwater modeling, and water management issues of the basin.
- The other one is identifying the most important hydrometeorological and physical factors and prioritizing them for groundwater recharge distribution (qualitatively) for the study basin, and so for similar study areas, including the effect of land use change on the recharge and other hydrological terms. These have been discussed in the result and discussion section and concluded in the conclusion part.
With due respect, these objectives, especially the first one are important for international readers. Many hydrological models for different areas (small catchment-scale to large basins) are developed for estimating groundwater recharge. However, evaluation of the method with point estimations based on direct water level measurement (WTF) and chemical tracer (CMB) gives a good insight for future recharge estimation techniques for similar aquifer types wherever they are located in the world. In this study, identifying topographical and geological characteristics, and thus the recharge mechanisms have found an important factor and starting point in selecting recharge estimation methods. The study also pointed out that the common approach that is being implemented i.e. calculating recharge by a multitude of methods and averaging out the results of the different methods is found unreliable. Rather selecting an appropriate one or few technique/(s) and considering that as the optimal result is recommended.
Evaluating the groundwater recharge estimation mechanism for the volcanic aquifers lying at different topographical settings, that represents a vast major part of the world groundwater aquifers, and able to suggest more appropriate method/(s) will benefit different further similar studies and researchers. Hence, this study can be seen as a dual purpose paper: evaluating the physical-based hydrological model (and so other similar models), and giving a spatial recharge rate map for the important Lake Tana basin.
Thank you (the reviewer) for your valuable comments, in the new version, shortcomings in the introduction section have been improved as per the suggestions. The research questions are outlined and the contribution of the research output is mentioned in the new version.The reason why we talk about “choosing appropriate methods” is to direct our reader that evaluation of a method will be one of the objectives of this manuscript. The reason why different methods were developed for recharge estimation rather than using the simplest and cost-effective method is due to the fact that different aquifers have different recharge mechanisms for which some methods are more effective than others. It is why we did an evaluation and validation of the existing method. Furthermore, when we say point recharge estimation, we are referring to recharge estimation techniques that estimate at a given point (e.g. on groundwater well, spring, etc.) such as WTF and CMB methods. Otherwise, yes indeed, we agree with the advance of Remote sensing and GIS, and with the developing capability of the grid-based spatially simulating hydrological models, producing spatially distributed water balance maps is so possible and being widely applied. The applied WetSpass model for this study is one of such kind, and we only applied an existing model (nothing new is done in this study as far as new methodology is concerned). As we have aforementioned, we pursued this study, for evaluating the physically-based hydrological models for our study basin so that areas having similar hydrogeological characteristics will take into account the suggestions of our evaluation result for future application. Besides, the recharge rate map for the basin, which is an important research product for groundwater exploitation practices and plans, will help groundwater managers and policymakers in this important area which is regarded as one of the growth corridors of the country. Besides, the recharge map is also an important input for further groundwater modeling work.
With due respect, the studies by (Alemayehu and Kebede, 2011; Ayenew et al., 2008; Demlie et al., 2008, 2007b; Kebede et al., 2005) are not all about recharge estimations studies. It is only Demlie et al., 2008, who did a recharge estimation only using the CMB method on a volcanic aquifer for the Akaki catchment located at about 550 km from the Lake Tana basin. The others had made different hydrogeological studies but had put general statements about the recharge variability across the volcanic aquifers. Thus, we can make clear that the presented study in this paper is different from the studies made by these authors.
Some maps in the methodology are shifted to the result section as suggested. The shortened explanation of the methodology section is for the reason of shorting the paper. We thought citing the papers that develop the original methodologies and later modifications will be enough and give detailed background knowledge including the different mathematical equations. However, parameter value modifications that have been done in this paper are discussed and justified in the new version of this paper (some of the details can be found in the following response to your line-by-line comments).
Response to detail line by line comments
Comment1: Language editing such as deletion, addition, and rewriting are suggested.
Reply: kindly accepted, and language edition has been done throughout the body of the manuscript, according to the suggestions.
Comment 2: The different correction suggestions for the location map (Fig. 1)
Reply: accepted, and appropriate corrections have been done in the new revised manuscript.
Lines 128-126: “That means years of seasonal or monthly time series data are averaged into single seasons or months.” How this limitation is considered in this study?
Reply: The objective of this study (as far as WetSpass is concerned) is to produce long-term average spatial recharge and other water balance component maps showing the spatial variation of the components. Furthermore, it is to see the seasonal variations of these water balance terms. It is not to see the fine-time scale temporal variability of them. It would indeed be nice if we, for example, were able to see the change of recharge over the years, but the WetSpass model cannot do that.
Line 139: how this parameter is calculated
Reply: thank you for the comment. The parameter is the amount of rainfall that reaches the surface (total RF minus the one intercepted by the plant canopy). The interception is calculated by the model using the land use at each modeling pixel.
In the WetSpass model, depending on the type of vegetation, the interception fraction represents a constant percentage of the annual precipitation value. Thus, the fraction decreases with an increase in the annual total rainfall amount (since the vegetation cover is assumed to be constant throughout the simulation period). The detail about it is found in Batelaan and De Smedt (2007, 2001).
Line 142: how EP is calculated? What do you mean by open water?
Reply: Open water used here to mean a water body (could be lake water, dam reservoir, etc.).
The description how evapotranspiration is calculated in WetSpass is mentioned as follows.
For the calculation of seasonal evapotranspiration, a reference value of transpiration is obtained from open-water evaporation and a vegetation coefficient:Trv= c Eo
Trv = the reference transpiration of a vegetated surface [LT-1];
Eo = potential evaporation of open water [LT-1] and
c = vegetation coefficient [–].
This vegetation coefficient can be calculated as the ratio of reference vegetation transpiration as given by the Penman-Monteith equation to the potential open-water evaporation, as given by the Penman equation
For vegetated groundwater discharge areas, the actual transpiration (Tv) is equal to the reference transpiration as there is no soil or water availability limitation:Tv= Trv if (Gd - ht)≤ Rd
Gd, is groundwater depth [L];
ht is the tension saturated height [L] and
Rd is the rooting depth [L].
For vegetated areas where the groundwater level is below the root zone the actual transpiration is given by:
Tv= f(θTrv if (Gd- ht ) < Rdf(θ) is a function of the water content
the detail about it is found in the original model developing literature by Batelaan and De Smedt (2007, 2001).
Line 147: give detail explanation about the parameter modifications
Reply: Detailed explanation is added in the revised one. We mean some adjustment for the parameters’ values on the default land use parameter table has been made, for example, the root depth of forest land is changed. In the model parameter table, each land use type has given impervious, bare land, vegetation, and open water percentages. In the default parameter of the WetSpass model, the vegetation area coverage for the Bush and Shrubland is 100%. This is based on the observation of the land use class type in the temperate zone (the Netherlands and Belgium). However, the Bush and Shrubland land use type in the Ethiopian (tropics) context is different: the vegetation is sometimes sparsely distributed, and is with some bare land component, during our field verification on the land use type, we tried to guess (using parcels land on which we do some measurement), and came up with about 10% is bare land and the rest 90% is consisting of vegetation. Similarly, adjustments have been made by Gebreyohannes et al. (2013) during their application of the model for Geba catchment, in northern Ethiopia. Similarly, the sub-afro-alpine vegetation land use type found in our area (consisting of about 0.3% of the total area coverage) is not present in the default land use classes of the WetSpass model. However, we made an equivalent with wet meadow land use type, and following a similar procedure, we modified the land use percentage to 80% vegetation and 20% bare land (it was 100% vegetation for wet meadow land cover type in the model parameter table).
Line 165: Paragraph above does not include land use as a parameter for calibration??
Reply: Yes, indeed land use is not used for calibration. It is to mean the land use which is used for model calibration is the one made using 2014 satellite images. The model parameters were optimized using this land use and then after the developed model was rerun using other land uses (1986 and 2000), just to see the change of the water balance components as a result of the change in land use over these years. Some paraphrasing has been done to make clear the issue.
Line 166: why these years (186, 2000, 2014) are chosen?
Reply: Because we have existing land use maps made using the satellite images taken in those years.
Line 170: It is unclear how the different model parameters serve the model?
Reply: as it is mentioned earlier in our response to your general comments, putting every detail of the model here would make the paper too long as the objective of this manuscript is neither modifying the model nor testing it. It is rather for applying the model and comparing (validating) with other methods, see how effective are such models (physically-based hydrological models) in estimating groundwater recharge (given the different assumptions these models have though they give fine spatially distributed maps), and identifying the controlling physical and meteo(hydro)logical factors, etc. Hence, detailed equations are better to be cited and be read from the published papers (containing model development and latter modifications). However, thank you to you, we found it is short and we added some more elaborations in the revised version.
Line 179: why??
Reply: Because the source for all water balance components including recharge is rainfall. For the months with no rainfall at all (most of the winter months) is less important. However, it does not mean that knowing at what month of summer (June, July, August, September ) is the most important recharge is happening, and at which month is the less, and why that is happening is insignificant for the hydrological knowledge of the area. It is so important and other works dedicated to it should focus on and study it.
Line 183: Unclear
Reply: it is to the rainfall amount, corrected accordingly.
Line 184: provide the number 8??
Reply: 8 of the stations are established by the project that funds this study. It is BDU-IUC project (funded by VLIR-UOS, by the Flemish Government of Belgium).
Line 187: In the introduction section you criticize the interpolation techniques.
Reply: Yes, in general, interpolation has drawbacks as it is estimations at several places given real measurement at some spatial points of an area. However, it depends on the property of the parameter (the matter) for which that interpolation is being executed for. For example for our specific case, interpolation for RF and other meteorological parameters has been done as the WetSpass model needs spatial maps as an input. However, interpolation of some parameters e.g. groundwater recharge (which is mentioned in the introduction section) is highly dependent on the ground hydrogeological condition of the area and is highly variable irrespective of spatial proximity for which interpolation mostly depends. Hence, instead of interpolating estimated results at measuring points of an area (e.g. by the WTF method), it is highly better to produce it using some method which can take into account all those hydrogeological factors (which use them as input). Hence, playing with spatially distributed input factors in the mathematical calculations (though some are prepared by interpolation) is not the same as final point value interpolation. It is true that we need spatially distributed groundwater values (maps) for example for groundwater modeling, and further management works, and hence, preparing it by interpolation of point values is unwise and will have significant errors and uncertainties.
Line 192: some detail on this method is required, at least the equation
Reply: Penman-Monteith method is a well-known method in hydrology. We think citing it would be enough for the reader.
Line 199: How these data on the groundwater level are collected??
Reply: the detail about it is discussed in WTF section, but it is true that we have to put some information on how we collect them. We have given some details in the revised one.
Line 201: is it possible to show them in your map?
Reply: Yes, they are shown on Fig. 6.
Line 204: separate section for calibration
Reply: thank you, yes, we agree, and hence, we put the calibration details under the heading model calibration in the revised manuscript.
Line 206: is there no climate change over the years 1986-2014, and 2012 to 2013? Selection should be justified here.
Reply: yes, there might be climate change over 1986 to 2014. However, the objective here is to see how the land use change affects the major water balance components using an already optimized WetSpass model. yes, in reality, the values of the actual water balance terms for 186, 2000, and 2014 might be more or less than what is simulated in this paper due to climate change or due to variabilities in meteorological parameters, however, it is common to keep other factors constant to evaluate the effect one influencing factor. Thus, the effect of land use on the hydrological variables was evaluated keeping other variables constant. The reason why these specific years were used for the evaluation is already mentioned above (it is because of the availability of land uses, mapped based on the satellite images taken in those years).
Line 222: the limitation factor of what?
Reply: corrected, it is to mean one of the limitation of this study.
Line 231-232: should be explained
Reply: It is to mean the same codes that the WetSpass model uses for each land use class has been given for each class of land use map used in the model as an input variable. Otherwise, the model would not read it. Corrected accordingly in the revised version.
Line 247: is there fieldwork that has been done? should be detail in this case
Reply: There was no fieldwork for this part, however, the authors use their general area knowledge.
Line 278: what about Johnson (1967)? Provide some detail.
Reply: It is a literature compilation of different possible specific yield values for different geological materials (aquifers in this case). We added some explanation about it in the revised one.
Line 279: where is the formula?
Reply: the formula, how it is developed, and detail about it is found in the paper cited. It is an empirical formula. The equation is included in the revised version.
Line 281: is it possible to provide it a supplementary material?
Reply: Unfortunately, the pumping test analysis paper is written and being submitted to a journal as a continuous part of this study. Hence, providing here as a supplementary material may not be possible. Sorry, for the inconvenience in the follow-up of the studies. the pumping test paper has to be published before this manuscript submission.
Line 293: period of what?
Reply: the period that the total RF is collected. Corrected accordingly.
Line 296: how many groundwater samples are considered, sampling map is required.
Reply: 138 groundwater samples distributed all over the basin have been used. Ok, the groundwater sample map is included in the revised version.
Line 297: there is some confusion related to runoff characteristics and coefficients. Could you provide us more details.
Reply: As it is shown in equation 5, the CMB method takes into account the runoff amount because what is needed in the equation is the amount of the infiltrated water i.e. the rainfall minus the surface runoff amount. As it is difficult to get runoff amount at each groundwater sampling point, we subdivided the area based on studied runoff characteristics. Dessie et al. (2015), studied the runoff characteristics of the Lake Tana basin using a conceptual based hydrological model and calculated average runoff coefficients ( = surface runoff amount divided by rainfall amount) for each area. In their study, they classified the basin into southern, northern, eastern, and western catchments, based on the runoff coefficient values and characteristics. We used those coefficients and calculated the runoff amount needed in the CMB equation. Accordingly, the recharge amount for each sub-area is shown in table 4.
It was possible to get runoff values at each groundwater sampling point using our distributing runoff map (from the WetSpass model output) and perhaps calculation of recharge at each sampling point would seem possible. The reason that we did not do it is because our objective i.e. we need to compare the different methods' results independently. Besides, as we have taken average chloride values of rainfall samples for each sub-area, the chloride values at each groundwater sampling point would be less certain (our rainfall samples were not sampled at each groundwater sampling point).
Line 298: I do not have any idea about these four catchments. Should be detail.
Reply: Thank you, some descriptions about these catchments grouped based on runoff characteristics are added in the revised version of the manuscript. Also few descriptions about how runoff is calculated have been also added.
Line 330: maybe in different periods as you have only 60 wells.
Reply: About 138 groundwater samples for chemical analysis (chloride concentration) were considered in this study (not 60). Equation 5, shows chloride concentration in the rainfall is needed in the CMB method. However, the problem here is the chemistry of the rainfall is not constant temporally and spatially. The rainfall may have different sources: seawater or maybe also fresh water for which chloride concentration reaching at a specific area in different seasons or months or periods would be variable. In this study, we tried to consider such possible causes of uncertainty to a certain extent, for the spatial variation (as stated earlier, we grouped the catchments), and for the possible temporal variation, we tried to sample both the groundwater and the rainfall in the similar period so the chloride concentration in the groundwater is mostly from the sampled rainfall that infiltrated recently. At least being able to sample the groundwater which percolates from the rainfall that we sampled too will give a reasonable recharge rate. But if we, for example, sample the groundwater in June and the rainfall in August, perhaps there will be a high error in the recharge value estimated.
Line 341: it is better to start by Evap, runoff and after R. Including the precipitation is valuable.
Reply: the suggestion is accepted, and corrected accordingly
Line 346: why so include in the result and discussion part.
Reply: thank you for the comment, moved to the methodology section.
Line 352: is there analysis (e.g. statistical analysis) to prove this.
Reply: Unfortunately, there is no such a solid analysis for all the factors rather we used our physical observation of the maps. But we extract the recharge map by the different class combinations of soil, land use, slope, and rainfall amounts, and tried to see the attribute table of extracted raster maps of class combinations vs recharge rate. In doing so, we tried to judge which factor is the most important and gave rank in controlling the recharge amount. It is impossible to explain it in quantification.
Line 355: some confusing with this value of 0 mm.
Reply: at the water body recharge is assumed 0 (there is no infiltration beneath the floor of the water body).
Line 360: it looks no appropriate to consider the 0 mm at water bodies for determing an average.
Reply: I think we have to consider rather. Because we are reporting the average value of the basin including the lake body. It is one of the land use classes where the other water balance components are influenced on. Since we are comparing it with RF that rains over all parts of the basin, the one (even 0 value) should be also included in the average.
Line 386: As winter is the dry season, we expect to have evp more important than summer, but this is not the case of the present study. Why?
Reply: yes, winter is the dry season, and hence there is no enough rainfall as the water source for evapotranspiration, the less value in winter is reasonable. The source water for evapotranspiration is basically coming from rainfall plus the evaporation on the lake. The lake evaporation is happening throughout the year irrespective of being summer or winter (it only depends on the amount of potential evaporation). Even compared to the low amount of rainfall, the 44% of winter evapotranspiration is high. This is due to the evaporation taking place from the lake water surface during the long dry winter.
Line 395: validation section
Reply: Ok, we will shift it to the validation section.
Line 397: It is better to compare with other methods.
Reply: Yes, it is also compared with other literature values made by other methods (shown in table 2, and has been discussed in the text preceding the table).
Line 402: coarser time scale??
Reply: It means not for fine time scale like for daily or hourly but it is per seasonal. The term is common in GIS, RS, and other spatio-temporal modeling works.
Line 428: where is R2 in the figure?
Reply: It will be added in the revised one.
Line 443: 1:1 line??
Reply: It is called an 'identity line' or 'line of equality'. It is standardizing the axis to compare measured data with predicted data, or two different models. The starting and ending point of both axes should be the same (through the axis origin in our case).
Line 465: about name the caption preceding or following the description?
Reply: thank you, we corrected it according to the suggestions.
Line 473: rephrasing suggestion on the subtopic??
Reply: accepted, and corrected as per suggested.
Line 500: you used WTF for validation in the previous section. So, it is better to introduce this result before validation.
Reply: accepted, rearrangement is made.
Line 511: what do you mean by degree of weathering?
Reply: It means the rate of weathering. A term used to mention the different grades of rock weathering which in turn refers to the different physical strengths of lithological materials from rock to soil. Hydraulic properties like porosity, hydraulic property, transmissivity, specific yield, etc. depend on weathering effect.
Line 542: It is better to include the chemical analyses of Cl both for rainfall and the selected groundwater samples. The effective precipitation values used for calculation are also required.
Reply: kindly accepted, a map consisting of the groundwater sampling location (for 138 GW samples), with a label of its chloride values is presented in the revised version. Also, the Cl amount of the RF is presented in table 4.
Line 552: I do not understand what represents this value as you have several periods and catchments.
Reply: accepted, and corrected accordingly. These values are the mean annual values for the whole Lake Tana basin. They are the average values for the whole basin estimated by the three methods.
Line 616: This R2 is good but as can be seen in figure, there is some lag between simulated and observed values.
Reply: yes, there are overestimations of river discharge by the WetSpass model (except for three of the major rivers: Gumara, Gilgel Abay, and to a certain extent, Gelda rivers). The possible reasons have been discussed in Lines 426-441 of the old version. The possible reason would be the assumption of the method that we followed to calculate the total river discharge: as it is already stated, we assumed the total discharge is a summation of surface runoff (SR) plus groundwater recharge (GWR) that the model gives both spatial maps and as a single total average value for the whole basin. However, theoretically, river discharge can be lower than this summation when there is significant deep groundwater flow that flows through the subsurface (without emerging to the surface) i.e. part of the water goes without being measured by the river gauging station. Hence, there is a possibility that the measured river discharge value is less than the sum of SR and GWR. In other words, it is with a special assumption that baseflow is equal to groundwater recharge (when all the recharge water emerges to the surface and flows through the river flow measuring station). The real geological conditions for the different catchments are already mentioned with reference to previous studies in the discussion section.
Hence, it does not mean that the WetSpass model overestimates the river discharge but it maybe because of the assumption of the baseflow amount to be equal to groundwater recharge in the total river discharge calculation of this paper. As the WetSpass model does not give total river flow amount as an output, we have followed this procedure to equate it so that we can compare it with the measured river discharge amount measured at the river outlets.Citation: https://doi.org/10.5194/hess-2021-527-AC2 - Response to reviewer 1 review comments
-
AC1: 'Reply on RC1', Alemu Yenehun Beyene, 25 Jan 2022
-
RC2: 'Comment on hess-2021-527', Anonymous Referee #2, 26 Jan 2022
General
In this study Alemu Yenehun et al. estimate groundwater recharge and study the spatial and temporal recharge patterns for the Lake Tana region in Ethiopia. Three (well-) established recharge estimation methods are applied, a model approach using the physically based WetSpass model, the groundwater table fluctuation method, and the chloride mass balance method. I applaud the Authors for conducting such an extensive study in a research area where data acquisition I assume can often be challenging. I think the main contributions from this paper are 1) an improved understanding of the groundwater recharge fluxes in the case study area, and 2) the comparison of different recharge estimation methods. These contributions are valuable, as groundwater is an important source of drinking water in the region and a better understanding of the resource in this region would serve many. I found the manuscript generally well written (minor textual improvements are required), but some restructuring and changes to the figures might be required to improve the readability of the manuscript. The main concern I have for this manuscript are related to the methodology and the description thereof, discussed in detail below and in separate line comments. Additionally, while I think the study is worth publishing, I am unsure whether the contribution fits the scope of HESS (“substantial new concepts, ideas, methods, or data”) in its current form. In my opinion, it would fit much better in other journals where it can be submitted as a case study (e.g., Hydrogeology journal), and reviewed as such. If the Authors wish to publish in HESS, I think a more elaborate analysis and discussion of the uncertainties in recharge estimation could be done to better fit within the scope of the journal.
Methods description
From the description of the methodology and data following in this manuscript, it is hard to reproduce the results and gain a full understanding of the modeling procedure. This could partly be solved by sharing the scripts, input, and output data used in this study. However, often a more detailed description of the modeling procedure is also required. The WetSpass model requires substantial data input, which is not always clearly described in the manuscript. From the description of the data, it seems that different time periods were used to generate the input data for WetSpass. Perhaps this could be clarified using a table that summarizes the different data sources and time series characteristics (e.g., measurement interval, period, operator). Given the high temporal variability of the different hydrometeorological variables, I assume the same time periods for all variables are used, but I could not verify this from the current manuscript. A few times it is mentioned “expert judgement” is used, but it remains unclear what values were applied and why. The calibration process is only briefly described, and the calibrated parameters are not reported. I general, I think it would be good to rewrite the methods section with reproducibility in mind.
Uncertainties in recharge estimates
The Authors mention in the introduction that it is important to take uncertainty in recharge estimation into account (line 47), and thus I was expecting a more elaborate analysis or discussion of the uncertainty in recharge estimation methods applied in this manuscript. As the WetSpass model is manually calibrated, no parameter uncertainties are available. It therefore remains unknown how uncertain these recharge estimates are. This could be addressed by a sensitivity analysis or a more elaborate uncertainty analysis. A discussion of the limitations of the different methods and the uncertainties of the recharge estimates at the end of section 4 would be a welcome addition to the current manuscript.
Title
The title suggests a more general study on volcanic aquifers, while a case study is presented. I think it would be better if the title reflects the fact that it is a case study. Additionally, the title suggests that there are “point measurements” of groundwater recharge. In my view it be better to refer to these as ‘point estimates’, as they are empirical estimates from a recharge estimation method and not real measurements.
Line comments
L31: add a comma after floodplain
L72: regionalize
L77: were all these studies in a specific study area? If so, good to mention that.
L123: Perhaps refer to the GitHub repo: https://github.com/WetSpass
L123: No reference to Wang et al (1996) required?
L132/135/139: Sentences introducing these equations would be nice.
L133: Should there not be a change in storage term?
L142: Which evaporation equation was used?
L144: I think it would be good to state which physically based equations (e.g., Darcy) were used and elaborate a bit on the model (e.g., finite differences).
L147-156: This description is rather vague, what was done exactly? How were the values changed?
L157-162: This could be elaborated and made more specific. How many parameters were calibrated, how many data points were used for calibration, was there a formal objective function used or only visual goodness of fit?
L164-164: “to validate the recharge estimates from WetSpass.”
L165: I was a bit confused here, perhaps change to “During model calibration”.
L174: remove “our” or mention the operating organization
L177: remove “relatively”
L178-179: Add a reference for this statement
L183: Can mostly be specified?
L187-188: I think at least a reference would be appropriate here.
L191: The naming “so-called first-class stations” seems inappropriate here, I suggest rephrasing, also in the Figures.
L198-201: How many monitoring wells were used? What period were the groundwater tables observed? Does the period overlap with the discharge data and the meteorological data? Is this the same data as that mentioned in L306?
L205-206: One value per year, does that mean the model is calibrated to just a very few data points? How many rivers were in the dataset?
L231-232: Unclear what “appropriate” is here.
L279: How many pumping and slug tests were done? What values for Sy were found, are these reported or available somewhere?
L282: Perhaps a separate subheading for the CMB.
L306: So the recharge estimates from the WTF method are only representative for that period. Is this considered when comparing to the other methods?
L325: Perhaps this could be combined with Fig. 2. with the hydrogeological setting?
L336: Where is 4.1? Also, I suggest starting with 4.3 (verification), before discussing the water balance components.
L342: At this point I do not know for what period WetSpass computes the recharge. Perhaops write: “The annual recharge over the period 20XX-20XX..”
L348: minor typo.
L351: Change “Next to” to “Apart from”.
L360: minor typo
L366: values
L379: Exclude the lake in figure 7B. Perhaps add subplot with the precipitation. The legend in Fig7a has a mistake in the values. Why was a continuous coloring scheme not used (e.g., from yellow to blue), this would make the figures much easier to interpret. In the figure caption, specify “long term” by mentioning the exact years.
L407-447: I do not think this is a proper “Model verification”, as the same data is used for calibration! We cannot verify a model using the same data that was used for calibration. Perhaps this section can be renamed to “Calibration results”.
L427-429: A high R2 was obtained after the calibration, which may be interpreted as that the model can explain a large part of the observed discharge variation. However, Fig. 8 also clear shows a large systematic error between modelled and observed discharge, which I think could be more clearly stated. Contrary to what is stated in the text, the simulated discharge is always higher compared to the observations. This would be a good point to get back to in the discussion.
L448: Here the section “Model verification” starts in my opinion.
L448-463: Is there a reason the estimates from the CMB method were not used here?
L465: Figure 9 could be condensed/smaller.
L500-549: This section could be placed earlier, as the results described in this section were previously used to compare to WetSpass. Add “method” after WTF throughout the section.
L506: “has been taken”
L526: Change “to catch up” to “to capture”.
L557: Perhaps the higher range of values could be reported here.
L560-591: Just a suggestion. Perhaps the Authors can come with a nice plot that visualizes the different recharge estimates from all the other studies, and those from this study.
L595: minor typo.
L612: Perhaps I misinterpret, but I added the percentages of runoff (29%), recharge (22%), and evapotranspiration (53), and these do not add up to 100% (=104%). Why is this, are there model errors or changes in storage?
L640: Some recommendation/implications for future studies and work could be added at the end.
Citation: https://doi.org/10.5194/hess-2021-527-RC2 -
AC3: 'Reply on RC2', Alemu Yenehun Beyene, 22 Feb 2022
Response to Referee 2
General
In this study Alemu Yenehun et al. estimate groundwater recharge and study the spatial and temporal recharge patterns for the Lake Tana region in Ethiopia. Three (well-) established recharge estimation methods are applied, a model approach using the physically-based WetSpass model, the groundwater table fluctuation method, and the chloride mass balance method. I applaud the Authors for conducting such an extensive study in a research area where data acquisition I assume can often be challenging. I think the main contributions from this paper are 1) an improved understanding of the groundwater recharge fluxes in the case study area, and 2) the comparison of different recharge estimation methods. These contributions are valuable, as groundwater is an important source of drinking water in the region and a better understanding of the resource in this region would serve many. I found the manuscript generally well written (minor textual improvements are required), but some restructuring and changes to the figures might be required to improve the readability of the manuscript. The main concern I have for this manuscript are related to the methodology and the description thereof, discussed in detail below and in separate line comments. Additionally, while I think the study is worth publishing, I am unsure whether the contribution fits the scope of HESS (“substantial new concepts, ideas, methods, or data”) in its current form. In my opinion, it would fit much better in other journals where it can be submitted as a case study (e.g., Hydrogeology journal), and reviewed as such. If the Authors wish to publish in HESS, I think a more elaborate analysis and discussion of the uncertainties in recharge estimation could be done to better fit within the scope of the journal.
Reply: Thank you so much. We have tried to do an uncertainty analysis of the model parameters and have added it in the revised version.
Methods description
From the description of the methodology and data following in this manuscript, it is hard to reproduce the results and gain a full understanding of the modeling procedure. This could partly be solved by sharing the scripts, input, and output data used in this study. However, often a more detailed description of the modeling procedure is also required. The WetSpass model requires substantial data input, which is not always clearly described in the manuscript. From the description of the data, it seems that different time periods were used to generate the input data for WetSpass. Perhaps this could be clarified using a table that summarizes the different data sources and time series characteristics (e.g., measurement interval, period, operator). Given the high temporal variability of the different hydrometeorological variables, I assume the same time periods for all variables are used, but I could not verify this from the current manuscript. A few times it is mentioned “expert judgement” is used, but it remains unclear what values were applied and why. The calibration process is only briefly described, and the calibrated parameters are not reported. I general, I think it would be good to rewrite the methods section with reproducibility in mind.
Reply: Yes, indeed, a table containing the different data sources and time period of the WetSpass model input variables will be added in the revised final manuscript. It is true that for some of the model input variables, the data for different time periods have been considered. As the model is using long-term averages of the different variables as an input (in our case: long-term average summer and winter), hence, observing long-term time-series temporal variations of the water balance components with the WetSpass model is impossible. The model is limited as it is simulating only long-term monthly averages as a fine time scale (time series results are not possible). In other words, the model is basically developed to simulate long-term average monthly/seasonal conditions, and most importantly the spatial variations. The model is basically developed for spatial simulation of groundwater recharge (which is later coupled with groundwater modeling-MODFLOW) whose short -time (day to day or hour to hour variations) is less important compared to the other water balance terms.
Except for the groundwater level map (which is based on recently collected data), meteorological variables (rainfall, average temp., wind speed, all meteorological parameters used to calculate PET using the Penman-Monteith method) were more or less in similar time periods. Yes, it is true that the meteorological variables have high temporal variations (most importantly the seasonal or monthly, or daily variations); year-to-year variation is relatively small. In other words, there is a small variation for a given season or month for different years. Given the objective of this paper i.e. the spatial variation of recharge and to a certain extent other water balance terms, these would not be a major problem.
The other non-meteorological (physical) parameter that changes with time is the land use. For the long-term average data (2005 to 2018), we used the land use map using the 2014 satellite image and ground truth data, for the modeling; land use of 1980 and 2000 for the evaluation of the effect of land use change on the water balance terms. The use of one-time data for the land use and long-term average for the other parameters might have slightly departed from reality and may have caused some uncertainties. However, the model is not robust enough to use different land uses for calibration. Furthermore, the change in land use is not so fast, and hence, the land use of 2014 might have represented more or less the possible land use status of the area during 2005-2018.
Explicit statements for what do we mean by “expert judgment” will be added in the revised version. We mean some adjustment for the parameters’ values on the default land use parameter table has been made, for example, the root depth of forest land is changed. In the model parameter table, each land use type has given impervious, bare land, vegetation, and open water percentages. In the default parameter of the WetSpass model the vegetation area coverage for the bush and shrubland is 100%. This is based on the observation of the land use class type in the temperate zone (the Netherlands and Belgium). However, the Bush and Shrubland use in the Ethiopian (tropics) context is different: the vegetation is sometimes sparsely distributed, and is with some bare land component, during our field verification on the land use type, we tried to guess (using parcels of land on which we do some measurement), and came up with about 10% is bare land and the rest 90% is consisting of vegetation. Similarly, adjustments have been made by Gebreyohannes et al. (2013) during their application of the model for Geba catchment, in northern Ethiopia. Similarly, the sub-afro-alpine vegetation land use type found in our area (consisting of about 0.3% of the total area coverage) is not present in the default land use classes of the WetSpass model. However, we made an equivalent with wet meadow land use type, and following a similar procedure we modified the land use percentage to 80% vegetation and 20% bare land (it was 100% vegetation for wet meadow land cover type in the model parameter table.
<< insert table1: the model land use parameter table here, attached in the supplement file>>
The steps followed in the calibration process, and the table for the calibrated parameters will be added in the revised version of the manuscript.
The modeling section of the methodology is more elaborately written in the revised version. However, to reduce the pages of the manuscript, all the detailed equations, and the empirical formulas developed and followed in the WetSpass model are better not to be included, we would rather cite the papers that developed the model first and improved further.
Uncertainties in recharge estimates
The authors mention in the introduction that it is important to take uncertainty in recharge estimation into account (line 47), and thus I was expecting a more elaborate analysis or discussion of the uncertainty in recharge estimation methods applied in this manuscript. As the WetSpass model is manually calibrated, no parameter uncertainties are available. It therefore remains unknown how uncertain these recharge estimates are. This could be addressed by a sensitivity analysis or a more elaborate uncertainty analysis. A discussion of the limitations of the different methods and the uncertainties of the recharge estimates at the end of section 4 would be a welcome addition to the current manuscript.
Reply: The WetSpass model has different parameters and needs input variables to give a reasonable output for the understanding of water balance components of a given river basin/watershed. Yes, to evaluate the efficiency of the WetSpass model applied in this study, the sensitivity analyses of all parameters and input variables would be a good addition of the paper. A simple sensitivity analysis is carried out and added in this revised version of the manuscript. Since the objective of doing this sensitivity analysis is to know the possible uncertainties that might be incorporated into the estimated water balance values, we did a sensitivity analysis for the calibration of the global parameters of the WetSpass model. These are alfa coefficients (it is a parameter adjusting soil moisture value), interception coefficients (adjusting interception by the plant canopy), LP coefficient (parameter adjusting evapotranspiration), rainfall intensity, and x coefficient (runoff delay factor). The other physical factors (called local parameters) such as the contribution of the slope (40%), landuse (30%) and soil (30%) in determining the balance components share, are originally developed from literature (experimental research). Thus, we kept them as they are (we have taken the default values of the model), and hence, we have not done a sensitivity analysis though their local variation might have caused some uncertainties in the estimated recharge and other water balance terms. Furthermore, we do not perform sensitivity analysis of the meteorological variables used in the model perhaps they might cause some uncertainties during measurement, interpolation, etc. They were fixed variables in our calibration similar to the physical parameters mentioned above.
Among the global model parameters calibrated in the model, the most sensitive global model parameter is RF intensity. On average, for every 1 mm/hr increase of the intensity, there is a mean annual recharge increment of 38.2 mm. Fortunately, we measured the rainfall intensity at some of the meteorological stations that the BDU-IUC project (the project that sponsors this study) established (using the automatic rain gauges), and hence, we used the range of these data in our calibration.
<<insert Figure 1 attached in the supplementary file here>>
Figure 1. Mean annual groundwater recharge (mm) versus average rainfall intensity (mm/hr.) for the different possible intensity values. The model was running for these RF intensity values by keeping the other model parameters constant.
In general, except for the RF intensity (which highly depends on the rainfall characteristics of an area), the other global model parameters have given a possible range of values in the WetSpass model (LP coefficient and x coefficient: 0.1 to 1; and alfa coefficient and a interception 1 to 10). Hence, the sensitivity testing for values within these ranges has been performed.
For every variation in LP coefficient, there is a 4 mm increase or decrease of the annual recharge.
<<insert Figure 2 attached in the supplementary file here>>
Figure 2. Mean annual groundwater recharge (mm) versus LP coefficient for the different possible LP coefficient values.
On average, for every 1 value variation for alfa coefficient there is only about 2.3 mm variation in annual recharge rate.
<<insert Figure 3 attached in the supplementary file here>>
Figure 3. Mean annual groundwater recharge (mm) versus Alfa coefficient for the different possible Alfa coefficient values.
The variation in a interception is small and only for the initial values. For the relatively higher values, the recharge is almost not changing, and hence, the recharge can be considered as insensitive for the variation of a interception. This has been also shown with the low R2 value.
<<insert Figure 4 attached in the supplementary file here>>
Figure 4. Mean annual groundwater recharge (mm) versus a interception coefficient for the different possible a interception coefficient values.
The groundwater recharge is not changing for the possible changing intervals of the x-coefficient, and hence, the x-coefficient is the most insensitive parameter among all the global model parameters for the Lake Tana basin.
<<insert Figure 5 attached in the supplementary file here>>
Figure 5. Mean annual groundwater recharge (mm) versus x coefficient for the different possible x coefficient values.
The limitation and the sensitivity analysis which would bring uncertainties to the reported values have been added in the revised version.
Title
The title suggests a more general study on volcanic aquifers, while a case study is presented. I think it would be better if the title reflects the fact that it is a case study. Additionally, the title suggests that there are “point measurements” of groundwater recharge. In my view it be better to refer to these as ‘point estimates’, as they are empirical estimates from a recharge estimation method and not real measurements.
Reply: Thank you so much. Yes, indeed, they are point estimates. For example, the variables in the calculation of the WTF method are the change in the water level and the specific yield (Sy). The water level is based on fine-time scale measurement, however, there is estimation for Sy though it is also analyzed from slug and pumping tests (real measurements), estimation from Johnson (1967) literature compilation has been also used. Given the high aquifer variability (compared to the representativity of each test well), it is true the WTF values are also estimations. Thus, the topic is corrected accordingly.
Yes, indeed the methods are tested in the case study area, however, we do believe the evaluation part that recommends the effective application of the methods for aquifers lying at different topographical and hydrogeological conditions would make the study beyond being case study, and will attract global readers.
Line comments
L31: add a comma after floodplain
Reply: thank you, added.
L72: regionalize
Reply: accepted, and corrected accordingly.
L77: were all these studies in a specific study area? If so, good to mention that.
Reply: mentioned, and thus corrected accordingly.
L123: Perhaps refer to the GitHub repo: https://github.com/WetSpass
Reply: thank you so much. Cited accordingly.
L123: No reference to Wang et al (1996) required?
Reply: Wang et al (1996) is about the WetSpa model. Maybe it is a good idea to add it as WetSpa is the starting point and the basis for the development of WetSpass. Hence, it is added in the new revised version.
L132/135/139: Sentences introducing these equations would be nice.
Reply: thank you, added.!
L133: Should there not be a change in storage term?
Reply: Yes, in the WetSpass model, the storage term is neglected. There are two ways of incorporating change in storage to the model on a seasonal basis. The first instance is that the plant available soil moisture reservoir in the summer is assumed to be filled up while it can be depleted in the model in the winter (for areas like Lake Tana basin); the second case is that in the model a different groundwater depth can be used for winter and summer. In our methodology, we followed the second option, we had prepared the groundwater table map for the summer and winter seasons using the time-series groundwater level monitoring data that we used for our study. This is explained in the methodology part of the manuscript. However, since we have Lake Tana in our study area, there is a possibility of surface water storage change. For example, the evaporation from the open lake water may cause the summation of annual AET + Runoff + Recharge to be greater than the annual RF i.e. RF < AET + runoff + groundwater recharge. Hence, there is a possibility that the storage term changes. However, change in storage in the WetSpass model is not as important as the other hydrological models, which can simulate water balance components at fine-time series (such as a daily or hourly). For such models, storage amount (variation) at each time-series highly affects the water balance components of the following time-series value. E.g. the storage change of day 1 affects the water balance values for day 2. But in the WetSpass model applied in this study (seasonal), there are only two time-series whose storage difference (especially the groundwater storage) by an independent water table map.
L142: Which evaporation equation was used?
Reply: the Penman equation is used for evaporation in WetSpass. It is discussed already in the result and discussion section.
L144: I think it would be good to state which physically-based equations (e.g., Darcy) were used and elaborate a bit on the model (e.g., finite differences).
Reply: In general, the WetSpass model is based on known hydrological equations (conceptually and physically meaningful equations), and also some empirical equations developed from works of literature. Yes, a better-elaborated explanation of the model background will be added in the revised version of the manuscript.
L147-156: This description is rather vague, what was done exactly? How were the values changed?
Reply: In the revised manuscript, detailed changed values of the land use parameters are explained, as it is already explained above. Basically, the changes are in the percentage of bare land and vegetation area percentage of a given land use type (bush and shrubland and sub-afro-alpine grassland), and also root depths of forest land (it is already explained in the paragraph).
L157-162: This could be elaborated and made more specific. How many parameters were calibrated, how many data points were used for calibration, was there a formal objective function used or only visual goodness of fit?
Reply: there is no formal objective function used, only the visual goodness of fit has been used. Basically, about five global parameters have been used for calibration (discussed above). The final optimal values (after intensive manual calibration trials) have been presented in a table in the revised manuscript. The goodness of fit was compared between the accumulated mean annual river discharge amount (simulated surface runoff +baseflow) vs the measured river discharge amount of the seven major rivers.
L164-164: “to validate the recharge estimates from WetSpass.”
Reply: validation is changed by evaluation, the word might have been confusing.
L165: I was a bit confused here, perhaps change to “During model calibration”.
Reply: thank you, changed accordingly.
L174: remove “our” or mention the operating organization
Reply: thank you. It is changed by the BDU-IUC project funding
L177: remove “relatively”
Reply: removed
L178-179: Add a reference for this statement
Reply; thank you, added.
L183: Can mostly be specified?
Reply: Yes, it has been done in the revised one.
L187-188: I think at least a reference would be appropriate here.
Reply: Ok, will be done.
L191: The naming “so-called first-class stations” seems inappropriate here, I suggest rephrasing, also in the Figures.
Reply: Ok, corrected accordingly!!
L198-201: How many monitoring wells were used? What period were the groundwater tables observed? Does the period overlap with the discharge data and the meteorological data? Is this the same data as that mentioned in L306?
Reply: Yes, they are the same data listed on L306. The monitoring wells were 65 in number. The measurements were taking place from 2017 to 2019: some using automatic water level measuring device (measuring every half an hour), and others were manually (per each day).
About data overlap: partly it overlaps with meteorological data. However, the meteorological data at most of the stations are long-term averages (2005 to 2018), generally, it is impossible to say the data exactly overlaps. However, since it is long-term average seasonal values that we are finally used as input, it will not be a major problem. The intention of the modeling work here is not primarily to see the long-term temporal variation of recharge and other water balance terms (but limited to mean seasonal variations). Similarly, the river discharge data used for calibration is not in the same time interval (2012-2013, mentioned on line 204). Nevertheless, as it is the average seasonal values that are applied, it will/would not have significant variation.
L205-206: One value per year, does that mean the model is calibrated to just a very few data points? How many rivers were in the dataset?
Reply: using 7 river gauging data at 7 major rivers of the basin (shown in Fig. 8). It is the seasonal values which are used for calibration (summer: June – September, and winter: October to May).
L231-232: Unclear what “appropriate” is here.
Reply: It is to mean code matching between the model parameter tables and land use classes has been made so that the model reads the right land use.
L279: How many pumping and slug tests were done? What values for Sy were found, are these reported or available somewhere?
Reply: This pumping test analysis is our next paper (already prepared and will be submitted to a journal soon). A test on 31 hand-dug wells was successfully executed and analyzed (the statement is added in the new revised version). Yes, the results per aquifer type were reported already on lines 507 to 515 of the old version.
L282: Perhaps a separate subheading for the CMB.
Reply: Ok, accepted.
L306: So the recharge estimates from the WTF method are only representative for that period. Is this considered when comparing to the other methods?
Reply: As we have mentioned earlier, the recharge estimated in all of the methods is mean annual. It is true that the meteorological variables used in the WetSpass model, and perhaps the land use have variations over the years. However, the annual values of these meteorological variables used as input for WetSpass are more or less constant. However, day-to-day variations of these parameter values especially the rainfall is high and would have resulted in high daily variations of water balance components such as recharge. Unfortunately, the WetSpass cannot simulate such fine time-series variations except the long-term seasonal variations.
Comparing the long-term recharge amount simulated by the WetSpass and the WTF for these periods would not be much of a problem (as it is the mean annual which is being compared). Similarly, the CMB is calculated only for the groundwater and rainfall sampling years (2017-2018).
The effect of land use change is evaluated in this paper and found small.
Thank you, we will put this fact as one of the limitations of this research.
L325: Perhaps this could be combined with Fig. 2. with the hydrogeological setting?
Reply: Yes it is possible to combine. However, it would be too earlier to put it here for the readers to follow the methodology section.
L336: Where is 4.1? Also, I suggest starting with 4.3 (verification), before discussing the water balance components.
Reply: Thank you so much, we corrected both according to your suggestions.
L342: At this point, I do not know for what period WetSpass computes the recharge. Perhaps write: “The annual recharge over the period 20XX-20XX..”
Reply: The WetSpass model computes from 2005 to 2018. We added the phrase in the manuscript.
L348: minor typo.
Reply: corrected accordingly.
L351: Change “Next to” to “Apart from”.
Reply: thank you, have done.
L360: minor typo
Reply: thank you, corrected accordingly.
L366: values
Reply: corrected.
L379: Exclude the lake in figure 7B. Perhaps add subplot with the precipitation. The legend in Fig.7a has a mistake in the values. Why was a continuous coloring scheme not used (e.g., from yellow to blue), this would make the figures much easier to interpret. In the figure caption, specify “long term” by mentioning the exact years.
Reply: thank you for the comments. It is not the lake which is mentioned in Fig.7b, rather the value. There is by far a higher AET value than in the other areas (since it is a water body where open water evaporation operates). That is why the contrast between the lake and the other land uses is higher, which causes the values at the lake to have a similar colour.
Yes, we will remove the lake in figure 7c. The mistake in the values of the figure 7a legend might be the 0 value (I guess) over the water body. WetSpass assumes that there is no recharge for the aquifers under the floor of the water body.
We have checked and corrected the maps according to the suggestions.
L407-447: I do not think this is a proper “Model verification”, as the same data is used for calibration! We cannot verify a model using the same data that was used for calibration. Perhaps this section can be renamed to “Calibration results”.
Reply: in the WetSpass model, there is no way that we insert calibration data (river discharge data) in the model itself during calibration. We were running the model (more than a hundred times) by changing the global model parameters, and we have been cross-checking the river discharge values of the basin. After doing it for several times, the optimal values (i.e. the closest values between the simulated and measured discharge data) were reached, and changing the values for the modeling parameters reached to a point that changing values no more improved the R2 between measured and simulated values. However, this is mostly between the total annual river discharges of all the rivers together. Hence, comparing the discharge values of the simulated and the measured after extracting the values (in raster map) by each river catchment can be partly considered as validation. However, as you mentioned, it will give full confidence if we discuss it under calibration rather than validation. For validation, the one estimated by the WTF and CMB can be still kept on.
L427-429: A high R2 was obtained after the calibration, which may be interpreted as that the model can explain a large part of the observed discharge variation. However, Fig. 8 also clearly shows a large systematic error between modelled and observed discharge, which I think could be more clearly stated. Contrary to what is stated in the text, the simulated discharge is always higher compared to the observations. This would be a good point to get back to in the discussion.
Reply: Thank you for your concern over the issue. As you stated, generally simulated values are higher than the observed ones. The possible reasons have been discussed in this section. The possible reason would be the assumption of the method that we followed to calculate the total river discharge: as it is already stated, we assumed the total discharge is a summation of surface runoff (SR) plus groundwater recharge (GWR) that the model gives both in spatial maps and as a single total value. However, theoretically, river discharge can be lower than this summation when there is significant deep groundwater flow that flows through the subsurface (without emerging to the surface) i.e. part of the water goes without being measured by the river gauging station. In other words, baseflow is not always equal to recharge, there are cases that baseflow is much lower than the recharge. Hence, there is a probability that the river discharge value is less than the sum of SR and GWR. The real geological conditions for the different catchments are already mentioned with reference to previous studies in the discussion section.
In general, it does not mean that the WetSpass model overestimates the river discharge but it maybe because of the assumption of the baseflow amount to be equal to groundwater recharge in the total river discharge calculation of this paper. As the WetSpass model does not give total river flow amount as an output, we followed this procedure to equate it so that we can compare with the measured river discharge amount measured at the river outlets.
L448: Here the section “Model verification” starts in my opinion.
Reply: Ok, we edited accordingly.
L448-463: Is there a reason the estimates from the CMB method were not used here?
Reply: Because the CMB is done for the different river catchments (by subdividing the river catchments into eastern, southern, northern, and western catchments, based on the similarity and difference in the runoff characteristics) rather than calculating at each sampling point. We took the average chloride concentration sampled and analyzed for the different major river catchments (Table 4) (chloride for both groundwater and rainfall), for doing the calculation (as mentioned already in the methodology section). It is impossible to get effective precipitation (rainfall minus surface runoff at each sampling point, which is one of the important variables in the CMB calculation). Hence, it is not easy to extract values from the recharge raster map produced by WetSpass at each sampling point like for the WTF method. However, we have a discussion part about it in the section (comparing the result with other methods) “summarizing groundwater recharge estimations by the different methods and comparing with other similar studies”.
In the validation section, we added a discussion with CMB as well in the new revised version.
L465: Figure 9 could be condensed/smaller.
Reply: accepted, we have done it.
L500-549: This section could be placed earlier, as the results described in this section were previously used to compare to WetSpass. Add “method” after WTF throughout the section.
Reply: Ok, thank you, rearranging has been done.
L506: “has been taken”
Reply: inserted
L526: Change “to catch up” to “to capture”.
Reply: changed
L557: Perhaps the higher range of values could be reported here.
Reply: It is to mean if the CMB calculation had been made at each groundwater sampling point (if it was possible to do so), the spatial variation of recharge would have been more than what has been calculated and reported i.e. average values over each catchment.
L560-591: Just a suggestion. Perhaps the Authors can come up with a nice plot that visualizes the different recharge estimates from all the other studies, and those from this study.
Reply: yes, we will show it either in a table or cross-plot by summarizing the studies done at different sub-catchments in the Lake Tana basin.
L595: minor typo.
Reply: corrected
L612: Perhaps I misinterpret, but I added the percentages of runoff (29%), recharge (22%), and evapotranspiration (53%), and these do not add up to 100% (=104%). Why is this, are there model errors or changes in storage?
Reply: This is already discussed above as a reply to your general comments. This might happened due to two reasons:
1) some of the recharged water has also transpired especially in the forest area (deep-rooted eucalyptus trees considered in the root depth parameter table (5.5 m depth), which is below the water level, especially for the summer season). The WetSpass adds this transpiration amount from the groundwater to the calculated total AET value, it is why the water table raster map is needed as an input file.
2) The AET at the open water (the Lake Tana and the Wetlands) is equal to the PET, which is higher than the RF, hence there might be some imbalance at, for example, the Lake Tana. This is also reported by different authors as a closure error.
L640: Some recommendation/implications for future studies and work could be added at the end.
Reply: Yes, recommendations suggesting how researchers and water management practitioners use this research output, and limitations of the study mainly method assumptions and limitations are added in the revised version.
-
AC3: 'Reply on RC2', Alemu Yenehun Beyene, 22 Feb 2022
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
837 | 449 | 51 | 1,337 | 37 | 34 |
- HTML: 837
- PDF: 449
- XML: 51
- Total: 1,337
- BibTeX: 37
- EndNote: 34
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Cited
3 citations as recorded by crossref.
- Comparative evaluation of SWAT and WTF techniques for recharge estimation in the Vea catchment, Ghana C. Okrah et al. 10.1007/s40899-023-00813-6
- Impact of Climate and Land-Use Change on Groundwater Resources, Study of Faisalabad District, Pakistan M. Sajjad et al. 10.3390/atmos13071097
- Contribution of on-farm avocado ( Persea americana ) tree-based agroforestry practice on selected soil physical and chemical properties of Inguti small watershed, in the highlands of North-Western Ethiopia Z. Kindie et al. 10.1080/27658511.2023.2289702