the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Jul 2026
A Novel Modeling Framework to Track Contaminant Sources and Transport in a Groundwater-Dominated Watershed
Abstract. Backwards particle-tracking is widely used in groundwater modeling to gain insights into transport processes in terms of source areas, travel paths, and travel times. In this study we develop a novel modeling framework to apply this approach to SWAT-MODFLOW, a coupled surface water-groundwater model, using the particle-tracking model MODPATH. We tested the framework by applying MODPATH to a SWAT-MODFLOW model of a groundwater-dominated watershed and tracking particles backwards from a river reach with significant spring discharges to the locations where they entered the MODFLOW domain. By combining the source area information of the particles with the nitrate loading from the SWAT model, we were able to simulate nitrate concentrations in the discharged groundwater as well as to simulate changes in nitrate concentrations resulting from different land use and management scenarios. Critically, our approach developed here permits simulation of travel time distributions and source areas of water volumes and nitrate masses. We show how these source and age components can be used to guide the prioritization of water quality mitigation strategies that consider both intrinsic (i.e., hydrogeological) vulnerability as well as specific management actions that could be taken to improve surface water and groundwater quality.
- Preprint
(2057 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2828', Anonymous Referee #1, 14 Jul 2026
-
AC1: 'Reply on RC1', Rob De Rooij, 01 Nov 2026
The Introduction needs to include a broader and more critical review of previous studies. The current discussion does not clearly show the research gap or the novelty of the proposed framework. Please better explain the relationship between this study and previous particle-tracking, nitrate-source identification, and coupled surface water-groundwater modeling studies.
We thank the reviewer for highlighting the need to clarify the research gap and novelty of our framework. In response, we will revise the Introduction to better distinguish our approach from previous particle-tracking, nitrate-source identification and other related modeling studies. Specifically, we will emphasize how linking nitrate loading simulated as a function of land use and climate with backward particle tracking enables quantification of source contributions and travel times of both groundwater and nitrate, which is a novel outcome. We will also clarify how this information supports spatial prioritization of mitigation actions and assessment of their expected response times.
The claimed computational efficiency should be quantitatively demonstrated. Please compare the runtime, memory requirement, and numerical performance of the proposed framework with SWAT-MODFLOW-RT3D under similar model settings. The authors should also discuss whether the computational savings are achieved at the cost of accuracy because of particle tracking and temporally averaged flow fields.
We thank the reviewer for this comment. In response, we will describe in more detail when and how our approach is advantageous in terms of modeled outcomes as well as potential computational efficiency. We will also qualify our efficiency claims to acknowledge that we have not quantitatively compared runtime, memory requirements, or numerical performance. However, we note that since SWAT-MODFLOW-RT3D cannot be used to simulate source areas and travel-times directly (in a single run), a one-to-one comparison focusing solely on efficiency with our approach is not straightforward. Although it may be possible to compare them when simulating concentrations, such a comparison would address only part of the functionality considered in this study, where the key objective is to gain insights into travel times and source areas. We therefore do not propose a direct benchmarking comparison but will expand the discussion of potential accuracy tradeoffs associated with neglecting dispersion and using temporally averaged flow fields, including the conditions under which these assumptions are expected to be reasonable.
The Upper Floridan Aquifer is strongly karstified, but it is represented as an equivalent porous medium in the model. Preferential pathways and karst conduits may significantly affect capture zones, groundwater travel times, and mitigation responses. Please test alternative conceptual models, such as high-conductivity preferential pathways, or at least evaluate the sensitivity of the main conclusions to the possible influence of conduit flow.
In our study, we acknowledge this limitation in the discussion, and we thank the reviewer for highlighting its potential implications for our conclusions. However, the role of preferential pathways along conduits on flow and transport is difficult to evaluate without additional understanding of their spatial distribution and hydraulic properties. Explicit representation of conduit flow would require at least some knowledge of the location and geometry of conduit networks. In addition, we would need to develop a model code that can track particles through a coupled conduit-matrix flow model. While simplified high-conductivity pathways could be tested within an equivalent porous medium representation, their placement and properties would remain hypothetical given the available data. A systematic evaluation of alternative conceptual models is thus beyond the scope of this framework-development study. In response to this comments, we will expand the Discussion to explain how unresolved preferential pathways could affect capture zones, travel-time distributions, and predicted mitigation responses. We will also further clarify that the robustness of these predictions to alternative representations of karst flow could be an avenue for future work.
The adjustment of horizontal hydraulic conductivity needs further clarification. Although the authors maintained the same effective horizontal conductivity across the model layers, changing its distribution among different hydrostratigraphic units may affect interlayer flow, particle pathways, and travel-time distributions. Please provide more hydrogeological justification for these modifications and evaluate their influence on the main results.
We thank the reviewer for highlighting the need to clarify the rationale and consequences of these adjustments. In our study, this adjustment is supported by comparison of the particle-tracking results with available groundwater-age estimates and the hydrogeological understanding of the aquifer. As described in Appendix B, the NFSEG model domain’s original conductivity values produced a median travel time of approximately 150 years, with substantial particle movement through the Lower Floridan Aquifer. The adjusted values produced travel times more consistent with the estimates reported by Katz et al. and increased flow through the Upper Floridan Aquifer, consistent with its greater karstification, particularly in unconfined areas. We note that we already evaluate the influence of changing the hydraulic conductivities on simulated travel times in Appendix B. We will make this justification and the comparison in Appendix B more explicit in the main text. We will also clarify that maintaining the effective horizontal conductivity does not preserve interlayer flows or particle pathways, and that the adjustment is not a unique solution.
Minor comments
Lines 1-2: The title uses the broad term “contaminant,” whereas the framework is mainly applied to nitrate and includes nitrate-specific loading and attenuation assumptions. Please consider replacing “Contaminant” with “Nitrate” or more clearly explain whether the proposed framework can also be applied to other contaminants.
We will explain more clearly that while our study focuses on nitrate , the SWAT-MODLFOW-MODPATH framework we developed currently allows simulation of Nitrogen solutes (nitrate, nitrite, ammonium and organic nitrogen), Phosphorous solutes (SRP), Pesticides , and major salt ions (sulfate, calcium, magnesium, sodium and potassium, chloride and carbonate).
Lines 13-15: Please replace “Backwards particle-tracking” with “backward particle tracking.” The terminology and hyphenation should also be consistent throughout the manuscript. This will be corrected.
Line 98: Typo. “SWAT-MODLFOW” should be corrected to “SWAT-MODFLOW.” This will be corrected
Line 181: “Travel time probably density function” should be corrected to “travel time probability density function.” This will be corrected
Line 268: The labels in Figure 4b are difficult to read. Please increase the font size or improve the figure resolution. This will be corrected
Lines 272-279: Please include a figure showing the vertical model discretization, hydrostratigraphic units, and spatial distribution of the main MODFLOW parameters.
We will add such a figure in a new supplementary materials section.
Line 280: Please replace “variably” with “variable.” This will be corrected
Lines 378-384 and 417-420: The fonts and legends in Figures 6, 7, and 10 are relatively small. In addition, the large number of source components makes these figures difficult to understand. Please enlarge the text and legends to improve their readability. This will be corrected
Lines 394-395 and 414-415: The legends or color bars in Figures 8 and 9 are missing clear names and units. Please clearly indicate the variables shown and provide units where applicable. This will be corrected
Lines 414-415: Figure 9b should include its own land-use legend instead of asking readers to refer back to Figure 5. This will be corrected
Lines 409-411: In this case, the NNC is achieved…
We will correct this sentence
Line 556: Should “Figure A1b” be corrected to “Figure B1b”? This will be corrected
Citation: https://doi.org/10.5194/egusphere-2026-2828-RC1
Citation: https://doi.org/10.5194/egusphere-2026-2828-AC1
-
AC1: 'Reply on RC1', Rob De Rooij, 01 Nov 2026
-
RC2: 'Comment on egusphere-2026-2828', Anonymous Referee #2, 01 Sep 2026
GENERAL COMMENTS
This manuscript addresses a topic with potentially significant implications for water resources and agricultural management. It also raises interesting considerations regarding the importance of understanding the timescales over which the effects of designed or implemented mitigation strategies develop.
However, I feel that the manuscript often lacks some of the information needed to motivate and justify the methodological choices described. I have therefore identified several potential “gaps” throughout the manuscript. It is not strictly necessary to address all of them, but I strongly believe that addressing at least some of these specific points would make the manuscript substantially clearer and more transparent.
SPECIFIC COMMENTS
1) Lines 33-36. In my opinion, this sentence (from “Assessing the downstream […]” to the end of the paragraph) could be connected more clearly to the preceding part of the text. For example, it might be useful to explain why the “downstream response to a change in nitrate loading” is particularly relevant for the design of mitigation strategies.
2) Lines 46-48. Why does SMR not provide direct information on source areas and travel times? How is the modelling performed by the software involved?
3) Line 49. For greater clarity, it might be useful to briefly explain why the development of strategies to improve water quality requires information on source areas and travel times.
4) Line 60. It is not entirely clear to me whether the “particle-tracking” mentioned at the beginning of the line refers to the “backward particle-tracking” mentioned shortly before. If so, it might be useful to state this explicitly for clarity. If not, what is the connection with the backward particle tracking discussed in the preceding text?
5) Line 70. For greater clarity, it might be useful to specify after the full stop that the particle tracking is performed independently (perhaps mentioning MODPATH), and that, as stated in the following sentence, the results are then combined with those concerning nitrate loading.
6) Line 96. What is the purpose of developing multiple SWAT-MODFLOW models? This is mentioned later in the manuscript, but for greater clarity it might be useful to briefly mention it here as well.
7) Line 97. Is a single stand-alone MODFLOW model sufficient, or is one required for each SWAT-MODFLOW model?
8) Lines 97-99. How and why is the stand-alone MODFLOW model sufficient? Why is a coarser temporal timescale used? Moreover, why is a new model necessary would it not be sufficient to include the MODPATH component in the SWAT-MODFLOW model?
9) Line 120. Is the “step” mentioned here a time step?
10) Lines 121-124. Which steps would be the most effective and which the least effective? Moreover, how long should the period covered by the flow fields be in order to fully simulate particle trajectories? Can this be determined a priori?
11) Lines 128-130. I understand that this is already explained in Appendix A, but for greater clarity it might be worth briefly mentioning here how the original code worked and why the new approach is an improvement.
12) Lines 134-135. Do the assumptions mentioned here therefore apply in this case? If so, why? How could the validity of these assumptions be assessed in other case studies?
13) Lines 141-143. How can one determine how far back in time to go in cases such as this? Can this be defined a priori?
14) Lines 156-158. It is not clear to me why the method is more efficient if all tracked particles contribute to the calculations. Perhaps this could be explained in greater detail to facilitate understanding.
15) Lines 160-162. It is not entirely clear to me how it is ensured that each particle contributes equally to the travel time distribution. This may not be necessary, but I would consider explaining this aspect in greater detail.
16) Lines 162-164. I do not fully understand how the random number generator contributes to the distribution of the particles. I would therefore suggest providing some additional details to clarify this aspect.
17) Equation 2. I assume that x and y are the spatial coordinates within area A. For greater precision, I would consider including them in the description of the variables provided in the text below.
18) Lines 179-180. What does τ (tau) represent the travel time of? It might be useful to specify this in the description.
19) Lines 180-181. How was pv derived from the probability density function? I would also suggest including a reference to the equation mentioned.
20) Line 182. This may be unnecessary for readers with a strong background in this field, but I would consider briefly explaining what is meant by a “first-order rate process” and why Equation 2 entails this assumption.
21) Lines 199-200. How is the HRU percolation rate scaled?
22) See comment on lines 162-164 about the random numbers generator.
23) Lines 232-233. I would consider specifying why the source areas and travel times are mapped at the same resolution as the HRUs in SWAT.
24) It would be useful to add to section 3.1 (or elsewhere, depending on what is considered most appropriate) information on the extent of the study area and the areas modelled by the different models, if these differ. This is not present in the current version, but would provide a greater context for the modelling framework described, particularly in terms of assessing the efficiency of the methods presented and their potential applicability to other studies.
25) Line 255. I see that the full name of the NFSEG acronym is provided on line 260. However, I would consider introducing it here or moving the full definition to this point, as this is the first occurrence of the acronym in the text.
26) Lines 263-276. I would consider briefly mentioning in the text the reasons for choosing these particular values for storage coefficients, specific yield, and porosity.
27) Line 298. Is it appropriate to refer to the conditions reported for 2017 as “current” conditions, given that this was almost ten years ago? Are they still representative of the current situation? Are they also representative of the 80-year period over which they are applied (see lines 320-321)?
28) Lines 302-304. Why were these particular conditions selected for the scenario? What are they representative of? I suggest providing some justification in the text to better contextualize this choice.
29) Line 307. What does the acronym NLDAS stand for? It would be preferable to define it in the text.
30) Lines 314-315. For greater clarity, I suggest specifying here which modelling years correspond to the three different stress periods.
31) Lines 322-323. Why were one million particles used? Please consider briefly motivating this choice in the text.
32) Lines 337-339. On what basis is the strategy described in these lines defined? What motivated the choices regarding the removal of septic tanks and the conversion of agricultural land?
33) Lines 339-345. The rationale behind these methodological choices is not entirely clear to me. I would recommend providing a few additional details to make the reasoning easier to follow.
34) Lines 366-369. Please consider specifying in the text the components to which reference is being made at each stage.
35) Line 402. It does not seem immediately clear to me how it can be inferred that “the most significant contribution areas correspond to agricultural land use”. Please consider adding a brief explanation.
36) Lines 409-411. Please check the sentence after the comma. It seems that either some information is missing or there is an unintended repetition.
37) Line 555. What is meant by “CFC measurements”? I recommend providing at least the full definition of the acronym.Figures:
38) All figures: I suggest ensuring that the figures are also clear and accessible to readers with different types of color-vision deficiency. Some of the color scales currently used could perhaps be replaced with alternatives that provide better accessibility.
39) Figure 1. The text within the box in the lower-left corner (“Run MODPATH […]”) seems somewhat schematic to me. Please consider making it clearer and more fluent.
40) Figures 3 and 4. It would be useful to add a numerical scale bar to the maps to provide a better indication of the dimensions of the areas shown.
41) Figure 4. I recommend increasing the font size of the spring names, particularly because the springs are referred to in the text (e.g., in section 3.3, page 19).
42) Figure 6 and Figure 7. I recommend increasing the font size of the axis labels and legends, as they are currently rather difficult to read.
43) Figure 6. I would consider indicating the median values of the variables mentioned in the text on Figures 6a and 6b, namely 32 years and 19 years, respectively.
44) Figure 7b. Please consider indicating on the graph the three periods referred to on line 371, for example using thicker vertical lines. It might also be useful to indicate the point at which NNC is reached (approximately 60 years; see line 375).
45) Figure 9. I think it would be preferable to include the legend here rather than simply referring the reader to the legend in Figure 5.
46) Figure 10. Also in this case, I would also consider indicating here the point at which NNC is reached (50 years; see line 409).
TECHNICAL CORRECTIONS
47) References to the literature in text are sometimes enclosed in square brackets and sometimes in parentheses. I suggest choosing one type of bracket and using it consistently throughout the manuscript.
48) Line 13. The manuscript uses the form “backward propagation”, rather than “backwards propagation” as used here. I recommend maintaining consistency and using “backward”.
49) I would remove some commas to improve the flow of the text. For example, the comma on line 37 after “[Wei et al., 2019]”; the comma on line 73 between “loadings” and “can”; and the comma on line 424 after “(Wei et al., 2019)”.
50) Line 111. I believe the correct form is “HRUs” rather than “HRU’s”, as currently written.
51) Line 141. It seems that an “r” is missing from “though” at the end of the line; and that should read “through”.
52) Line 181. Please check whether “probably” should instead be “probability”.
53) Lines 212-213. There appears to be a repetition, and I believe one of the two expressions “for particles” and “for distributing particles” could be removed.
54) Line 262. There is an unnecessary space after the parenthesis “(Durden et al., 2019)”.
55) Line 268. In the legend of Figure 4a, “hydaulic” appears instead of “hydraulic”.
56) Line 269. There is a missing space between “springsheds,” and “stream networks”.
57) Line 280. “Variable” should probably be used here instead of “variably”.
58) Line 286. For stylistic consistency, the parentheses around “Rath, 2021) should be removed.
59) Line 288. The manuscript generally uses the form “land use(s)”, rather than “land-uses” as it appears here. I therefore suggest removing the hyphen between the two words to maintain consistency with the rest of the text.
60) Line 344. It seems that the “the” before “grassland values” should be removed.
61) Line 359. I believe “function as weather” should be corrected to “function of weather”.
62) Line 508. The manuscript generally uses the form “particle-tracking” rather than “particle tracking” as it appears here. I therefore suggest adding the hyphen between the two words to maintain consistency throughout the text.Citation: https://doi.org/10.5194/egusphere-2026-2828-RC2 -
AC2: 'Reply on RC2', Rob De Rooij, 01 Nov 2026
SPECIFIC COMMENTS
1) Lines 33-36. In my opinion, this sentence (from “Assessing the downstream […]” to the end of the paragraph) could be connected more clearly to the preceding part of the text. For example, it might be useful to explain why the “downstream response to a change in nitrate loading” is particularly relevant for the design of mitigation strategies.We will rephrase this sentence.
2) Lines 46-48. Why does SMR not provide direct information on source areas and travel times? How is the modelling performed by the software involved?We will clarify the differences between SWAT-MODFLOW-RT3D (SMR) and SWAT-MODFLOW-MPDPATH and describe briefly how SMR does not provide information on source areas and travel times directry byt could provide this information indirectly by varying nitrate loadings in a series of simualtions..
3) Line 49. For greater clarity, it might be useful to briefly explain why the development of strategies to improve water quality requires information on source areas and travel times.We will add text that explicitly states why this information is important,
4) Line 60. It is not entirely clear to me whether the “particle-tracking” mentioned at the beginning of the line refers to the “backward particle-tracking” mentioned shortly before. If so, it might be useful to state this explicitly for clarity. If not, what is the connection with the backward particle tracking discussed in the preceding text?We will rewrite some parts in the corresponding paragraph to avoid any confusion.
5) Line 70. For greater clarity, it might be useful to specify after the full stop that the particle tracking is performed independently (perhaps mentioning MODPATH), and that, as stated in the following sentence, the results are then combined with those concerning nitrate loading. We will explicitly state that particle-tracking is performed independently.
6) Line 96. What is the purpose of developing multiple SWAT-MODFLOW models? This is mentioned later in the manuscript, but for greater clarity it might be useful to briefly mention it here as well. We will revise this section to explain that multiple models are required to simulate different land use scenarios that result in different recharge and nitrate loading ratess.
7) Line 97. Is a single stand-alone MODFLOW model sufficient, or is one required for each SWAT-MODFLOW model? We will describe more clearly that it is a single stand-alone model.
8) Lines 97-99. How and why is the stand-alone MODFLOW model sufficient? Why is a coarser temporal timescale used? Moreover, why is a new model necessary would it not be sufficient to include the MODPATH component in the SWAT-MODFLOW model?In the original manuscript with began section 2 with a very brief summary of the framework to introduce figure 1.. In the revised manuscript we will indicate more clearly in this section that each that each step is described in more detail in subsequent sections. later on.
9) Line 120. Is the “step” mentioned here a time step? We will rewrite this paragraph to avoid confusion of what is meant by step. It is a single computational step to solve a semi-analytical solution for the path within a cell given that the flow field does not change.
10) Lines 121-124. Which steps would be the most effective and which the least effective? Moreover, how long should the period covered by the flow fields be in order to fully simulate particle trajectories? Can this be determined a priori?In the revised manuscript we will separate the 2 issues more clearly (transient hourly flow fields vs time period covered). We will also explain how the needed time period covered could be determined (measurements or particle-tracking results)
11) Lines 128-130. I understand that this is already explained in Appendix A, but for greater clarity it might be worth briefly mentioning here how the original code worked and why the new approach is an improvement. We will revise this section to briefly explain the error in the original code that resulted in incorrect computation of river stages and the rational for the new approach..
12) Lines 134-135. Do the assumptions mentioned here therefore apply in this case? If so, why? How could the validity of these assumptions be assessed in other case studies? In the revised manuscript we will justify these assumptions in more detail, and discuss how the validity of these assumptions could be tested more rigorously for this and other case studies.
13) Lines 141-143. How can one determine how far back in time to go in cases such as this? Can this be defined a priori? We will describe that in the absence of measured values one could test this based on simulation results with solely current land use.
14) Lines 156-158. It is not clear to me why the method is more efficient if all tracked particles contribute to the calculations. Perhaps this could be explained in greater detail to facilitate understanding. We will explain this in greater detail
15) Lines 160-162. It is not entirely clear to me how it is ensured that each particle contributes equally to the travel time distribution. This may not be necessary, but I would consider explaining this aspect in greater detail. We will add that this means that they contribute with the same weight. We either do it this way (with equal weights) or we must compute different weights for the particles. Either way, we must know the flow rates in the discharge cells.
16) Lines 162-164. I do not fully understand how the random number generator contributes to the distribution of the particles. I would therefore suggest providing some additional details to clarify this aspect. In the revised manuscript we will explain that this methodology is is similar to Latin hypercube sampling for large numbers.
17) Equation 2. I assume that x and y are the spatial coordinates within area A. For greater precision, I would consider including them in the description of the variables provided in the text below. This will be corrected
18) Lines 179-180. What does τ (tau) represent the travel time of? It might be useful to specify this in the description. We will specify clearly it is travel time of water.
19) Lines 180-181. How was pv derived from the probability density function? I would also suggest including a reference to the equation mentioned.We thank the reviewer for identifying this unclear wording. We will replace “derived from” with “provided by” to make clarify that pv is provided directly by the travel-time probability density function rather than calculated through an additional derivation. We will also make the equation reference explicit and clarify that the supporting references are provided immediately before Equation 2.
20) Line 182. This may be unnecessary for readers with a strong background in this field, but I would consider briefly explaining what is meant by a “first-order rate process” and why Equation 2 entails this assumption. We will briefly state the equation (pde) for a first-order rate process to make this clearer.
21) Lines 199-200. How is the HRU percolation rate scaled?We think that is explained: “such that the total percolation for each MODFLOW cell equals the
22) See comment on lines 162-164 about the random numbers generator. We will make clear that more information can be found in the references for transfer probability functions.
23) Lines 232-233. I would consider specifying why the source areas and travel times are mapped at the same resolution as the HRUs in SWAT. We will add that this the same resolution for which we have simulated nitrate loadings.
24) It would be useful to add to section 3.1 (or elsewhere, depending on what is considered most appropriate) information on the extent of the study area and the areas modelled by the different models, if these differ. This is not present in the current version, but would provide a greater context for the modelling framework described, particularly in terms of assessing the efficiency of the methods presented and their potential applicability to other studies.We will add in section 3.2 that the extent of the study area is 12077 km2.
25) Line 255. I see that the full name of the NFSEG acronym is provided on line 260. However, I would consider introducing it here or moving the full definition to this point, as this is the first occurrence of the acronym in the text.This will be corrected
26) Lines 263-276. I would consider briefly mentioning in the text the reasons for choosing these particular values for storage coefficients, specific yield, and porosity.We will clarify in the revised manuscript that we based our choices information in the given references.
27) Line 298. Is it appropriate to refer to the conditions reported for 2017 as “current” conditions, given that this was almost ten years ago? Are they still representative of the current situation? Are they also representative of the 80-year period over which they are applied (see lines 320-321)?We acknowledge that 2017 is no longer current. We will revise the manuscript to indicate “2017 conditions” rather than current conditions. . We also clarify that we assume that 2017 land use and practices go back 80 years and will acknowledge the reasons for, and limitations of, this assumption.
28) Lines 302-304. Why were these particular conditions selected for the scenario? What are they representative of? I suggest providing some justification in the text to better contextualize this choice.We will add something like: This scenario is useful to check if we can reach the nutrient criteria under these conditions but also to construct a fourth scenario which is discussed in detail later on in which we apply the replacement of current agricultural and forestry lands with low-density longleaf pines only in a selected region.
29) Line 307. What does the acronym NLDAS stand for? It would be preferable to define it in the text.We will define this acronym in the revised manuscript.
30) Lines 314-315. For greater clarity, I suggest specifying here which modelling years correspond to the three different stress periods.This will be corrected
31) Lines 322-323. Why were one million particles used? Please consider briefly motivating this choice in the text.We will revise the manuscript to explain the motivation for, and justify, the use of one million particles.
32) Lines 337-339. On what basis is the strategy described in these lines defined? What motivated the choices regarding the removal of septic tanks and the conversion of agricultural land?We will add that this strategy removes the largest contributors to nitrate loading but uses the travel times to limit the area affected by mitigation
33) Lines 339-345. The rationale behind these methodological choices is not entirely clear to me. I would recommend providing a few additional details to make the reasoning easier to follow.We will add that this methodology is also more flexible as it is relatively straightforward to use another travel time contour.
34) Lines 366-369. Please consider specifying in the text the components to which reference is being made at each stage.This will be corrected
35) Line 402. It does not seem immediately clear to me how it can be inferred that “the most significant contribution areas correspond to agricultural land use”. Please consider adding a brief explanation.We agree this wording is not clear. We will revise the manuscript to indicate that agricultural land uses are characterized by relatively high loadings compared to other land uses..
36) Lines 409-411. Please check the sentence after the comma. It seems that either some information is missing or there is an unintended repetition.This will be corrected
37) Line 555. What is meant by “CFC measurements”? I recommend providing at least the full definition of the acronym.This will be corrected
Figures:
38) All figures: I suggest ensuring that the figures are also clear and accessible to readers with different types of color-vision deficiency. Some of the color scales currently used could perhaps be replaced with alternatives that provide better accessibility.We thank the reviewer for this suggestion. We will use online tools (e.g., the Coblis Color Blindness Simulator; https://www.color-blindness.com/coblis-color-blindness-simulator/) to assess the accessibility of our figures for readers with different types of color-vision deficiency and revise the color schemes where needed. For figures with many source categories, we will improve contrast and use additional visual distinctions, such as line styles or direct labels where practical, to reduce reliance on color alone.
39) Figure 1. The text within the box in the lower-left corner (“Run MODPATH […]”) seems somewhat schematic to me. Please consider making it clearer and more fluent.This will be corrected
40) Figures 3 and 4. It would be useful to add a numerical scale bar to the maps to provide a better indication of the dimensions of the areas shown.Scale bars will be added.
41) Figure 4. I recommend increasing the font size of the spring names, particularly because the springs are referred to in the text (e.g., in section 3.3, page 19).This will be corrected
42) Figure 6 and Figure 7. I recommend increasing the font size of the axis labels and legends, as they are currently rather difficult to read.This will be corrected
43) Figure 6. I would consider indicating the median values of the variables mentioned in the text on Figures 6a and 6b, namely 32 years and 19 years, respectively.This will be corrected
44) Figure 7b. Please consider indicating on the graph the three periods referred to on line 371, for example using thicker vertical lines. It might also be useful to indicate the point at which NNC is reached (approximately 60 years; see line 375).This will be corrected
45) Figure 9. I think it would be preferable to include the legend here rather than simply referring the reader to the legend in Figure 5.This will be corrected
46) Figure 10. Also in this case, I would also consider indicating here the point at which NNC is reached (50 years; see line 409).This will be corrected
TECHNICAL CORRECTIONS
47) References to the literature in text are sometimes enclosed in square brackets and sometimes in parentheses. I suggest choosing one type of bracket and using it consistently throughout the manuscript.This will be corrected
48) Line 13. The manuscript uses the form “backward propagation”, rather than “backwards propagation” as used here. I recommend maintaining consistency and using “backward”.This will be corrected
49) I would remove some commas to improve the flow of the text. For example, the comma on line 37 after “[Wei et al., 2019]”; the comma on line 73 between “loadings” and “can”; and the comma on line 424 after “(Wei et al., 2019)”.This will be corrected
50) Line 111. I believe the correct form is “HRUs” rather than “HRU’s”, as currently written. This will be corrected
51) Line 141. It seems that an “r” is missing from “though” at the end of the line; and that should read “through”.This will be corrected
52) Line 181. Please check whether “probably” should instead be “probability”.This will be corrected
53) Lines 212-213. There appears to be a repetition, and I believe one of the two expressions “for particles” and “for distributing particles” could be removed.This will be corrected
54) Line 262. There is an unnecessary space after the parenthesis “(Durden et al., 2019)”.This will be corrected
55) Line 268. In the legend of Figure 4a, “hydaulic” appears instead of “hydraulic”.This will be corrected
56) Line 269. There is a missing space between “springsheds,” and “stream networks”.This will be corrected
57) Line 280. “Variable” should probably be used here instead of “variably”.This will be corrected
58) Line 286. For stylistic consistency, the parentheses around “Rath, 2021) should be removed.This will be corrected
59) Line 288. The manuscript generally uses the form “land use(s)”, rather than “land-uses” as it appears here. I therefore suggest removing the hyphen between the two words to maintain consistency with the rest of the text.This will be corrected
60) Line 344. It seems that the “the” before “grassland values” should be removed.This will be corrected
61) Line 359. I believe “function as weather” should be corrected to “function of weather”.This will be corrected
62) Line 508. The manuscript generally uses the form “particle-tracking” rather than “particle tracking” as it appears here. I therefore suggest adding the hyphen between the two words to maintain consistency throughout the text.This will be corrected
Citation: https://doi.org/10.5194/egusphere-2026-2828-AC2
-
AC2: 'Reply on RC2', Rob De Rooij, 01 Nov 2026
-
RC3: 'Comment on egusphere-2026-2828', Anonymous Referee #3, 04 Sep 2026
This manuscript presents a modeling framework that integrates SWAT-MODFLOW with backward MODPATH particle tracking to characterize groundwater source areas, travel-time distributions, nitrate source contributions, and potential responses to land-use management. The framework is potentially useful for groundwater-dominated watersheds because it combines SWAT-derived nitrate loading with particle-based source and age information that is not readily available from concentration-based transport simulations alone. The Santa Fe River Basin application provides a reasonable demonstration of the workflow.
Because the manuscript is primarily a framework paper, I do not think the case study needs to represent all complexities of karst nitrate transport. However, several aspects of the framework's novelty, demonstration, and interpretation should be clarified before publication.
Major comments
1. The novelty of this framework appears to lie in using SWAT-MODFLOW to provide spatially and temporally varying recharge and nitrate loading for MODPATH. The authors could more clearly distinguish this contribution from previous particle-tracking nitrate studies and, if possible, demonstrate the added value of SWAT-derived loading through its effects on source attribution or predicted nitrate responses.
2. The nitrate predictions from the new framework are not directly compared with observations. Even a simple modeled-versus-observed nitrate comparison would strengthen the demonstration of the framework.
3. The effect of averaging flow conditions on source areas and travel-time distributions should be further demonstrated or justified.
4. Computational efficiency is claimed but not quantified. A representative runtime comparison would strengthen this claim.Minor comments
Lines 98 and 311: Please correct “MODLFOW” to “MODFLOW.”
Line 324: “Sante Fe River” should be “Santa Fe River.”
Lines 335–345 and 417–420: The Methods describe conversion of agricultural land within the 30-year contour, whereas the Fig. 10 caption states that both agricultural and production-forestry lands are converted. Please clarify the scenario actually simulated.
Lines 409–410: The sentence beginning “163507 hectares of the current condition 196315 hectares...” is hard to understand.
Lines 445–451: Please check the figure references. The land-use-specific TTDs appear to be shown in Fig. 7a, rather than Fig. 8a, and the temporal nitrate response appears in Fig. 7b, rather than Fig. 8b.
Throughout: Use “CDFs” and “TTDs” rather than “CDF's” and “TTD's” for plurals.Figures:
Figures 6–7: Please increase the sizes of the legends and labels.
Figure 9: Consider including the land-use legend directly in the figure rather than requiring readers to refer back to Fig. 5.Citation: https://doi.org/10.5194/egusphere-2026-2828-RC3 -
AC3: 'Reply on RC3', Rob De Rooij, 01 Nov 2026
Major comments
1. The novelty of this framework appears to lie in using SWAT-MODFLOW to provide spatially and temporally varying recharge and nitrate loading for MODPATH. The authors could more clearly distinguish this contribution from previous particle-tracking nitrate studies and, if possible, demonstrate the added value of SWAT-derived loading through its effects on source attribution or predicted nitrate responses.We will revise the manuscript to more clearly explain that the provision of a simulated nitrate loading as a function of land use and weather conditions is a key advantage of our modelling approach vis-à-vis previous particle-tracking studies in which this loading must be specified as a boundary condition.
2. The nitrate predictions from the new framework are not directly compared with observations. Even a simple modeled-versus-observed nitrate comparison would strengthen the demonstration of the framework.We will revise the discussion to include measuremed values of nitrate in the springs in the study region to illustrate that our simulated values fall within a reasonable range.
3. The effect of averaging flow conditions on source areas and travel-time distributions should be further demonstrated or justified.We will revise the manuscript to add further discussion regarding when our assumptions are valid.
4. Computational efficiency is claimed but not quantified. A representative runtime comparison would strengthen this claim.We will revise the manuscript to more precisely explain that the SMR model simulates both flow and transport with a daily timestep whereas our framework utilizes steady-state flow fields across much larger time (i.e. multi-decadal) intervals corresponding to particular fixed land uses. .
Minor comments
Lines 98 and 311: Please correct “MODLFOW” to “MODFLOW.”This will be corrected
Line 324: “Sante Fe River” should be “Santa Fe River.”This will be corrected
Lines 335–345 and 417–420: The Methods describe conversion of agricultural land within the 30-year contour, whereas the Fig. 10 caption states that both agricultural and production-forestry lands are converted. Please clarify the scenario actually simulated.This will be corrected\
Lines 409–410: The sentence beginning “163507 hectares of the current condition 196315 hectares...” is hard to understand.We will rephrase this sentence
Lines 445–451: Please check the figure references. The land-use-specific TTDs appear to be shown in Fig. 7a, rather than Fig. 8a, and the temporal nitrate response appears in Fig. 7b, rather than Fig. 8b.This will be corrected
Throughout: Use “CDFs” and “TTDs” rather than “CDF's” and “TTD's” for plurals.This will be corrected
Figures:
Figures 6–7: Please increase the sizes of the legends and labels.This will be corrected
Figure 9: Consider including the land-use legend directly in the figure rather than requiring readers to refer back to Fig. 5.This will be corrected
Citation: https://doi.org/10.5194/egusphere-2026-2828-AC3
-
AC3: 'Reply on RC3', Rob De Rooij, 01 Nov 2026
-
CC1: 'Comment on egusphere-2026-2828', Giacomo Medici, 10 Sep 2026
General comments
Very good research on groundwater hydrology with angle on contaminant transport. Please, follow my specific comments to improve the final version of the manuscript.
Specific comments
Line 35. “Hydrological processes vary widely in time”. Insert conceptual link between contaminant transport of nitrate, and climate change which is clearly time dependant.
Lines 53-55. “particle-tracking is widely regarded as a useful and relatively computationally efficient tool”. Insert recent review papers on advective flow and particle tracking with an evident worldwide angle:
- Lupi, F., Agbotui, P. Y., Medici, G. 2026. Hydraulic Conductivity in the Mesozoic units of the Umbria-Marche succession (Italy); insights towards a sustainable management of carbonate aquifers worldwide. Sustainability, 18, 9297; https://doi.org/10.3390/su18189297.
- Çallı, K. Ö., Chiogna, G., Bittner, D., Sivelle, V., Labat, D., Richieri, B., ... & Hartmann, A. (2025). Karst water resources in a changing world: Review of solute transport modeling approaches. Reviews of Geophysics, 63(1), e2023RG000811.
Line 86. Which is the general aim/goal of your hydrological research on contaminant transport?
Line 86. I would call these three points “specific objectives”.
Line 99. You need to justify the use of MODPATH that only incorporates the advection.
Lines 242-253. Please, provide more detail on the hydro-stratigraphy. Presence of carbonates and sandstones? Specify lithology and depositional paleoenvironment.
Lines 242-253. What about presence of high-angle faults that can affect the groundwater flow?
Lines 436-477. These two sub-paragraphs of the discussion should report references to engage the international audience.
Figures and tables
Figure 1. Increase the font size.
Figure 2. Spatial scale unclear.
Figure 3. Increase the graphic resolution of the maps.
Figure 4. Increase the graphic resolution.
Figure 5. Spatial scale unclear.
Figure 7. Details on the axes are not readable. Please, increase the graphic resolution.
Citation: https://doi.org/10.5194/egusphere-2026-2828-CC1 -
AC4: 'Reply on CC1', Rob De Rooij, 01 Nov 2026
General comments
Very good research on groundwater hydrology with angle on contaminant transport. Please, follow my specific comments to improve the final version of the manuscript.
Specific comments
Line 35. “Hydrological processes vary widely in time”. Insert conceptual link between contaminant transport of nitrate, and climate change which is clearly time dependant.
Our recent paper (using the same SWAT-MODFLOW model) looks into climate change (Lee, D., Graham, W., de Rooij, R., Reaver, N., & Kaplan, D. Implications of land use and climate change for meeting environmental flows and water quality targets in a groundwater-dominated watershed. Available at SSRN 6439518.). However, that is not the focus of this paper and is also not mentioned in the references in this line. We feel that a deviation into the role of climate change will confuse the reader as it is never mentioned again in the remainder of this manuscript.
Lines 53-55. “particle-tracking is widely regarded as a useful and relatively computationally efficient tool”. Insert recent review papers on advective flow and particle tracking with an evident worldwide angle:
Line 86. Which is the general aim/goal of your hydrological research on contaminant transport?
Line 86. I would call these three points “specific objectives”.
We will revise the manuscript to emphasize that point 1 is our general aim and points 2 and 3 are our specific objectives
Line 99. You need to justify the use of MODPATH that only incorporates the advection.
That MODPATH is purely advective and that this is a serious limitation is mentioned in the discussion.
Lines 242-253. Please, provide more detail on the hydro-stratigraphy. Presence of carbonates and sandstones? Specify lithology and depositional paleoenvironment.
We will revise the manuscript to icnlude more details on the hydro-stratigraphy of the region and will include A figure showing the hydrostratigraphic units and vertical model discretization in a new supplementary materials section.
Lines 242-253. What about presence of high-angle faults that can affect the groundwater flow?
The manuscript will be revised to include the fact that there is no evidence of major faults that affect groundwater flow in the area.
Lines 436-477. These two sub-paragraphs of the discussion should report references to engage the international audience.
Figures and tables
Figure 1. Increase the font size. This will be corrected
Figure 2. Spatial scale unclear. This figure is intended to be a schematic showing a hypothetical example of a HRU intersected by grid cells. We will state explicitly in the revised manuscript that it is a schematic to avoid confusion.
Figure 3. Increase the graphic resolution of the maps. This will be corrected
Figure 4. Increase the graphic resolution. This will be corrected
Figure 5. Spatial scale unclear. We will add a scale bar
Figure 7. Details on the axes are not readable. Please, increase the graphic resolution. This will be corrected
Citation: https://doi.org/10.5194/egusphere-2026-2828-AC4
-
AC4: 'Reply on CC1', Rob De Rooij, 01 Nov 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 321 | 170 | 57 | 548 | 42 | 44 |
- HTML: 321
- PDF: 170
- XML: 57
- Total: 548
- BibTeX: 42
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study aims to develop a new modeling framework by integrating SWAT-MODFLOW with MODPATH to identify nitrate source areas, quantify groundwater travel-time distributions, and evaluate the responses of spring-fed waters under different land-use and management scenarios. Overall, the topic is interesting and relevant, and the proposed framework has methodological novelty and potential value for groundwater quality management. However, several major issues related to model assumptions, validation, parameterization, and interpretation of the results need to be further clarified before the manuscript can be considered for publication. Below, I have listed comments, hoping they may help improve the manuscript’s quality.
Major comments
The Introduction needs to include a broader and more critical review of previous studies. The current discussion does not clearly show the research gap or the novelty of the proposed framework. Please better explain the relationship between this study and previous particle-tracking, nitrate-source identification, and coupled surface water-groundwater modeling studies.
The claimed computational efficiency should be quantitatively demonstrated. Please compare the runtime, memory requirement, and numerical performance of the proposed framework with SWAT-MODFLOW-RT3D under similar model settings. The authors should also discuss whether the computational savings are achieved at the cost of accuracy because of particle tracking and temporally averaged flow fields.
The Upper Floridan Aquifer is strongly karstified, but it is represented as an equivalent porous medium in the model. Preferential pathways and karst conduits may significantly affect capture zones, groundwater travel times, and mitigation responses. Please test alternative conceptual models, such as high-conductivity preferential pathways, or at least evaluate the sensitivity of the main conclusions to the possible influence of conduit flow.
The adjustment of horizontal hydraulic conductivity needs further clarification. Although the authors maintained the same effective horizontal conductivity across the model layers, changing its distribution among different hydrostratigraphic units may affect interlayer flow, particle pathways, and travel-time distributions. Please provide more hydrogeological justification for these modifications and evaluate their influence on the main results.
Minor comments
Lines 1-2: The title uses the broad term “contaminant,” whereas the framework is mainly applied to nitrate and includes nitrate-specific loading and attenuation assumptions. Please consider replacing “Contaminant” with “Nitrate” or more clearly explain whether the proposed framework can also be applied to other contaminants.
Lines 13-15: Please replace “Backwards particle-tracking” with “backward particle tracking.” The terminology and hyphenation should also be consistent throughout the manuscript.
Line 98: Typo. “SWAT-MODLFOW” should be corrected to “SWAT-MODFLOW.”
Line 181: “Travel time probably density function” should be corrected to “travel time probability density function.”
Line 268: The labels in Figure 4b are difficult to read. Please increase the font size or improve the figure resolution.
Lines 272-279: Please include a figure showing the vertical model discretization, hydrostratigraphic units, and spatial distribution of the main MODFLOW parameters.
Line 280: Please replace “variably” with “variable.”
Lines 378-384 and 417-420: The fonts and legends in Figures 6, 7, and 10 are relatively small. In addition, the large number of source components makes these figures difficult to understand. Please enlarge the text and legends to improve their readability.
Lines 394-395 and 414-415: The legends or color bars in Figures 8 and 9 are missing clear names and units. Please clearly indicate the variables shown and provide units where applicable.
Lines 414-415: Figure 9b should include its own land-use legend instead of asking readers to refer back to Figure 5.
Lines 409-411: In this case, the NNC is achieved…
Line 556: Should “Figure A1b” be corrected to “Figure B1b”?