the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Jun 2024
Assessing the long-term effectiveness of nitrogen management for groundwater protection in the agricultural crop production sector in Wallonia, Belgium
Abstract. Current nitrogen management programs within the agricultural crop production sector aim at optimizing crop productivity while minimizing environmental externalities, in particular groundwater contamination with nitrates. However, the effectiveness of these programs has been varied, with many studies indicating mixed or minimal results. Understanding the drivers of nitrate concentration in groundwater and its change is crucial for evaluating nitrogen regulations and guiding policy and management in the agricultural sector.
In this context, our study focused on assessing the effectiveness of the sustainable nitrogen management program for agriculture in Wallonia (PGDA), Belgium, on groundwater protection against nitrate contamination. We analysed nitrate concentration time series over the period 2002–2020 from 36 locations across four groundwater bodies within the Walloon nitrate vulnerable zones, situated in the agricultural belt. To capture the extent and dynamics of nitrate pollution, we developed and applied six indicators, providing a detailed view of both the current state and temporal trends of nitrate levels. Additionally, we computed spatially-explicit variables for each monitoring point to describe potential nitrate sources and their migration potential towards groundwater, and we examined their explanatory power in relation to the six nitrate pollution indicators.
Our findings indicate a modest overall improvement in average nitrate concentrations post-PGDA implementation. However, a closer examination at the individual site level reveals encouraging trends, with some locations showing pronounced decreased nitrate levels and with a decline in the average rate of change in nitrate concentration in 2020 indicating a slowdown in the rate of increase (or an acceleration in the rate of decrease) compared to 2002. Our results also underscore a complex array of factors influencing nitrate pollution and trends, with land use patterns and aquifer characteristics identified as key determinants. The study suggests that the absence of desired changes in certain areas could be attributed to a time lag between the introduction of regulatory measures and the observable impact on groundwater quality. This research highlights the intricate relationship between environmental regulation, land use, and groundwater quality, emphasizing the need for continued monitoring and adaptive management strategies to effectively address nitrate pollution in groundwater.
- Preprint
(1147 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-173', Anonymous Referee #1, 28 Jun 2024
1. General remarks:
The study described in this manuscript is important in the context of the need to assess the effects of actions taken to reduce nitrate pollution of groundwater and, consequently, rivers.
This requires a good methodology for determining the temporal trend of nitrate concentrations in groundwater and assessing the reversal of this trend. The study addresses this problem, and the adopted methodology for examining trend reversal seems to be appropriate. However, the methodological aspect is not properly highlighted in the manuscript. I would suggest increasing the emphasis and proportional share of the description of methodological issues of trend reversal analysis (Lowess and CP methods) throughout the text.
The second research goal, i.e. "identify the factors controlling the nitrate concentration and temporal trend", seems to be less important for the development of science (it is 2024). Moreover, in the case of this study, it is not universal, but local significance. Therefore, I would suggest moving parts of the manuscript regarding factors controlling the nitrate concentration and temporal trend to the Supplementary Material.
Instead, I would suggest going deeper into the details of the methodology description. Especially, the choice of the moment of trend reversal, number and “locations” of change points, the length of the period of examining a given trend, etc. These are universal issues, interesting for the reader involved in the analysis of water quality monitoring results.
2. Detailed remarks:
- Text and Table 1 - indicator I4 - should be described rather as "slope between 20...- 20.." instead "slope in 2002". Similarly, correction should be in column “Usage and interpretation". Similarly in respect to indicator I5, as well as I6. After all, the trend lasts for some time, not 1 year. The text should precisely explain how the period (length?) of examining the upward trend was assumed (calculated?), e.g. 2001-2003, instead of e.g. 1996-2002. Or vice versa - because it is visible differently in Fig. 3 and Fig. 4. Similarly, in the case of the time period for analysing the downward trend - whether it is 2019-2021 or for example 2010-2021 (or maybe even for example 2002-2020 ?). What are the ranges analysed in this study and why? After all, it directly affects the final result of assessing the effectiveness of measures to reduce nitrate pollution. Maybe it is even the only important factor on which everything depends. This should be described in detail.
- In text, no referral to Fig. 3, and there is no information on the basis of which data the time series was created.
- lines 148-149 - "The window length..." - no explanation of the reason for this approach. Why eight years?
- lines 152-152 - Helsel & Hirsch, 2002 - no in References.
- In text, no referral to Fig. 4, and there is no information on what data this time series is based on.
- Fig. 5 - no explanation of how to understand the meaning of the descriptions next to the "more desirable / less desirable" arrows in the context of histogram results.
- Fig. 6 - no explanation of how to understand the meaning of the descriptions next to the arrows "desirable effect / undesirable effect" in Kendall rank correlation.
- lines 404-406 - if precipitation is a variable in this research, it should be called "precipitation" or "rainfall", instead of "recharge". "Recharge" means that evapotranspiration is already included in this value. By the way, in Table 2 is "Rainfall", not "Recharge" - ?.
- References seems to be incomplete –
regarding the base materials, for example:
Hirsch et al. 1991 - Selection of methods for the detection and estimation of trends in water quality.
Esterby, 1996 - Review of methods for the detection and estimation of trends with emphasis on water quality applications.
Grath et al. 2001 - The EU Water Framework Directive: statistical aspects of the identification of groundwater pollution trends, and aggregation of monitoring results
Craig & Daly, 2010 - Methodology for establishing groundwater threshold values and the assessment of chemical and quantitative status of groundwater, including an assessment of pollution trends and trend reversal.
In relation to relatively new articles, for example:
Frollini et al., 2021 - Groundwater quality trend and trend reversal assessment in the European Water Framework Directive context: an example with nitrates in Italy.
Comments on at least these indicated publications would be needed.
3. Technical/editorial remarks
- lines 165-169 - incorrect paragraph formatting (font, line spacing, etc.)
- line 460 - no data about the publisher/institution and publication place.
Citation: https://doi.org/10.5194/hess-2024-173-RC1 -
AC1: 'Reply on RC1', Elise Verstraeten, 19 Aug 2024
We would like to thank the reviewer for his insightful comments on our manuscript. We have carefully considered his.her feedback and would like to address the main points as follows:
We acknowledge the importance of robust methods for identifying the timing of trend reversals, the number, and the "locations" of change points, and recognize that a dedicated paper on this subject would be valuable to the scientific community. However, methodological development is not the primary focus of our study. We employed two different trend and break point detection methods to ensure that our conclusions were not dependent on the choice of method. As the results were consistent across both methods, we propose to present the results from one method in the main text and include the results from the other method in the Supplementary Material to avoid any potential confusion on the main purpose of the paper.
Regarding the relevance of the second research goal, we acknowledge that there is an extensive body of studies that focused on identifying the factors controlling nitrate contamination. However, we believe that the added value of this study lies in two key aspects: (i) it is conducted on a regional scale and (ii) it aims to identify the drivers of temporal change in nitrate concentration using as well long-term as high-resolution time series data, unlike most studies based on t either short-term data or low-resolution long-term data. This originality has also been recognized by the second reviewer who stated: "In-depth explanatory studies of the response of nitrate in deeper groundwater for drinking water are quite rare." Overall, we also believe that such case studies are valuable as they can be integrated into the pool of similar studies conducted elsewhere, facilitating cross-site comparative research. For these reasons, we believe that pursuing this objective contributes significantly to enriching the overall body of knowledge. To put more emphasis on the temporal contribution of our work, we suggest to reformulate the objective 2 to “To identify the factors controlling the temporal changes in nitrate concentrations” (initially, it was “To identify the factors controlling the nitrate concentration and temporal trend.”)
2. Detailed remarks:
Text and Table 1 - indicatorI4 - should be described rather as "slope between 20...- 20.." instead "slope in 2002". Similarly, correction should be in column “Usage and interpretation". Similarly in respect to indicator I5, as well as I6. After all, the trend lasts for some time, not 1 year. The text should precisely explain how the period (length?) of examining the upward trend was assumed (calculated?), e.g. 2001-2003, instead of e.g. 1996-2002. Or vice versa - because it is visible differently in Fig. 3 and Fig. 4. Similarly, in the case of the time period for analysing the downward trend - whether it is 2019-2021 or for example 2010-2021 (or maybe even for example 2002-2020 ?). What are the ranges analysed in this study and why? After all, it directly affects the final result of assessing the effectiveness of measures to reduce nitrate pollution. Maybe it is even the only important factor on which everything depends. This should be described in detail.
- We believe it is appropriate to maintain the names “slope in 2002/2020,” as the slopes depend on specific time points (01-01-2002 and 01-01-2020). When using the change point method, the indicators are the slope of the linear regression between the two change points before and after the time points (01-01-2002 and 01-01-2020). When using the lowess method, the indicators are the slopes of the tangent lines to the lowess regression at 01-01-2002 and 01-01-2020. To make this entirely clear, we suggest modifying the descriptions of I4 and I5 to “Rate of change in nitrate concentration on 01-01-2002. [...]” and clarifying the calculation methods it in the text.
In text, no referral to Fig. 3, and there is no information on the basis of which data the time series was created.
- The purpose of figure 3 is to illustrate the six indicators, as mentioned in lines 132-134, where the figured is mentioned. We used the time series of one of the monitoring station as an example. We will specify that in the legend. For time series shorter than eight years, we took a window of 2/3 of the data, ensuring the
Lines 148-149 - "The window length..." - no explanation of the reason for this approach. Why eight years?
- As all time series have different lengths, we decided to use a fixed length for the window instead of a fraction. This approach ensures consistency across different time series, allowing for a more uniform comparison of trends. We chose a window length of eight years to capture meaningful trends while keeping long-term fluctuations. For time series shorter than eight years, we took a window length of 2/3, as a visual analysis showed using eight years resulted in insufficient sensitivity.
Lines 152-152 - Helsel & Hirsch, 2002 - no in References.
- Thank you for bringing that to our attention. We will add the reference.
In text, no referral to Fig. 4, and there is no information on what data this time series is based on.
- Thank you for bringing that to our attention. The time series data is based on one specific monitoring station, this will be specified in the legend. Given the decision to present and discuss the CP method in the supplementary material, we will move Figure 4 there as well and remove it from the main paper for the sake of clarity.
Fig. 5 - no explanation of how to understand the meaning of the descriptions next to the "more desirable / less desirable" arrows in the context of histogram results.
Fig. 6 - no explanation of how to understand the meaning of the descriptions next to the arrows "desirable effect / undesirable effect" in Kendall rank correlation.
- We will provide more detailed captions for figures 5 and 6, to help the reader’s interpretation of the results.
Lines 404-406 - if precipitation is a variable in this research, it should be called "precipitation" or "rainfall", instead of "recharge". "Recharge" means that evapotranspiration is already included in this value. By the way, in Table 2 is "Rainfall", not "Recharge" - ?
- The used variable is indeed ‘precipitation’. Thank you for bringing this to our attention, we will change the names.
References seems to be incomplete –
regarding the base materials, for example:
Hirsch et al. 1991 - Selection of methods for the detection and estimation of trends in water quality.
Esterby, 1996 - Review of methods for the detection and estimation of trends with emphasis on water quality applications.
Grath et al. 2001 - The EU Water Framework Directive: statistical aspects of the identification of groundwater pollution trends, and aggregation of monitoring results
Craig & Daly, 2010 - Methodology for establishing groundwater threshold values and the assessment of chemical and quantitative status of groundwater, including an assessment of pollution trends and trend reversal.
In relation to relatively new articles, for example:
Frollini et al., 2021 - Groundwater quality trend and trend reversal assessment in the European Water Framework Directive context: an example with nitrates in Italy.
Comments on at least these indicated publications would be needed.
- We thank the reviewer for the insightful references concerning groundwater quality trend assessment. We will review these references and cite them where relevant.
3. Technical/editorial remarks
lines 165-169 - incorrect paragraph formatting (font, line spacing, etc.)
line 460 - no data about the publisher/institution and publication place.
- Thank you for bringing this to our attention. We will take care of that.
Citation: https://doi.org/10.5194/hess-2024-173-AC1
-
RC2: 'Comment on hess-2024-173', Anonymous Referee #2, 23 Jul 2024
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2024-173/hess-2024-173-RC2-supplement.pdf
-
AC2: 'Reply on RC2', Elise Verstraeten, 19 Aug 2024
We sincerely thank the reviewer for taking the time to carefully review our manuscript and for providing such insightful comments. We have thoughtfully considered the feedback provided, and we hope that our responses adequately demonstrate the value of our paper and its relevance for publication in HESS.
We would like to start our response by emphasizing that our study aimed to use a data-driven approach instead of a more traditional modelling-based process-oriented approach. We used in-situ and satellite data to investigate the drivers of groundwater nitrate concentrations and their temporal changes. This approach intentionally avoids model-based data, as models are based on hypotheses and assumptions that we think could bias our findings. We will make sure to clarify our message on that aspect in the revised manuscript.
The other comments are addressed below.
I missed several relevant and available explanatory data, for example the nitrogen surplus (or Gross Nitrogen Balance) and proof of insight in the nitrogen cycle.
- We considered using gross nitrogen balance (in particular, the indicator “Nitrogen with potential to leach” APL (Azote Potentiellement Lessivable)) data, but ran into the issue that for the Walloon region, APL is an indicator available at punctual locations, not on a consistent regional-basis. While regional data are available, as pointed out by the reviewer, they are predicted from models, which is against our approach.
While aiming to support and contribute to the Walloon nitrate policies, the paper is not providing any detail on the policy measures and how these could relate to observed nitrate trends. Finally, I also found that the introduction of nitrate issue missed recent insights. The statistical analysis, using the six indicators and the two approaches to detect inflection points is sound and quite original, but unfortunately fails to deliver policy relevant conclusions.
- We agree that policy measures are important. However, this study aimed to consolidate the scientific basis. While such a study can be foundational for policy-making, we believe it is not the scope nor role of this paper to formulate policy recommendations. Our results provide insights into the drivers of nitrate concentrations and their temporal changes. Nonetheless, the complexity and variability of nitrate contamination across different regions and timescales mean that our findings, while valuable, are not conclusive enough to generate specific policy recommendations. Instead, they serve as a step in building a robust knowledge base that policymakers can use in the future.
Results and discussion hardly transcend to the level of describing the results of the statistical analysis and miss a translation to relevant, new conclusions about the effectiveness of the Walloon implementation of the Nitrates Directive or about the mechanism of response of nitrate in aquifers to inputs and other factors.
- We respectfully disagree with the assertion that our study did not generate any new relevant conclusions regarding the effectiveness of the PGDA (Walloon implementation of the Nitrate Directive). Firstly, our research is the first data-driven study in the Walloon region to explicitly demonstrate the lag time between the implementation of the PGDA in 2002 and the subsequent long-term improvements in water quality, specifically in terms of nitrate concentrations across four different water bodies.
- Moreover, our trend and slope analysis provides valuable insights into the rate of improvement. Understanding the time lag between the implementation of the PGDA and the observed improvements in water quality is crucial for accurately interpreting monitoring results and evaluating the policy’s effectiveness. This information is also vital for reporting to the EU, especially given that the EU imposes specific deadlines for achieving water quality targets, which may not always align with the natural temporal dynamics of dilution and attenuation mechanisms within groundwater bodies.
- Furthermore, we believe our study underscores the importance of establishing and maintaining long-term, consistent longitudinal monitoring of water quality and the drivers of water pollution. Such efforts are essential for effectively evaluating the impact of environmental policies.
- We will make sur to better highlight these conclusions and their implication.
I get the impression this works is not well connected to ongoing work and evaluations of the Walloon implementation of the Nitrates directive (which is mandatory every 4 years)1.
- We will make sure to revise the reports related to this work and make a better job in using them to contextualizing our objectives and connecting and discussing our results.
I also would expect that reconstructing or projecting spatially explicit trends of nitrate concentrations in aquifers would typically require process oriented numerical simulation models (like for example MODLFOW). While the authors make some reference to this type of modelling in the discussion, they not clearly explain why they choose for the data driven regression models in this study. While I don’t want to downplay the investment in using numerical groundwater simulation models (preferably you would team up with experts which already operationalized the model), the chosen statistical approach also appears to have been quite time consuming while (in my opinion) not providing clear answers about effectiveness of policy are legacy effects.
- As mentioned above, we followed in our study a purely data driven approach, as the major objective of the study is to focus on cause-impact relations. This avoids introducing bias in the data analysis related to the hypotheses that underpin the mechanistic models like MODFLOW, if such models would have been used. We agree that spatially discretised modelling could give additional insights in the spatial trends, but this was not the objective of the study which only focusses on temporal trends.
Recommendations for revision:
Adding more information and insight on the N cycle and N budget of the Walloon region and the consequence for the N loading of Walloon aquifers
Connect and refer to the publication of “Bilan d'azote en agriculture et flux d'azote des sols vers les eaux”( 21 décembre 2022 by État de l'environnement Wallon and show the added value your work. This includes showing the added value of your statistical regression approach versus a simulation approach (http://etat.environnement.wallonie.be/contents/indicatorsheets/SOLS%204.html).
- We propose to add a section “Nitrogen cycle and budget in Wallonia” after the introduction that provides insights into the nitrogen cycle and nitrogen budget of the Walloon Region, along with the consequences for nitrogen loading in the aquifers. Additionnally, we will consolidate our introduction with results from assessment studies and reports recommended by the reviewer, such as "Etat de l’Environnement Wallon" and other referenced materials recommended in the Review Remarks. We will also consolidate our discussion to interpret and discuss our results in light of this information.
Review remarks in more detail:
L50-54: The list of standard measures for NVZs sometime aim at protection of surface water (e.g., closed period for fertilizer and manure application on steep slopes, frozen ground) some more for groundwater (cover crops, balanced fertilization). I would suggest providing more detail and use these for your statistical analysis. I also wonder if in Wallonia additional measures (beyond requirements for Nitrates Directive) for regions where groundwater (or surface water) is used for drinking water production (in my country, The Netherlands, we have groundwater protection zones – “ grondwater-beschermings-gebieden” with more restrictions than in NVZs).
- We will provide more detail on the measures for NVZ. Additionally, we will introduce the legal groundwater protection zones early on, as they indeed have more restrictions. This is relevant because the delineation of the influence zones of the monitoring points is based on these protection zones.
L63-66: Suggest formulating this as a (or a few) clear research hypothesis. Suggest also to delete “landscape elements” and not only refer to accumulation but also to retardation of nitrate by chemical transformation to N2 and N2O (denitrification).
- We will consider the formulation of one or more research hypotheses and add the notion of retardation/denitrification.
L68-72: In my experience from the Netherlands Action Programs (or Plans) for implementation of the Nitrates Directive need to be evaluated and renewed every 4 years. Is this not the case for Wallonia, and if so, there must a history of evaluation report available. Please check and add.
- We will do so.
L111-117 You explain that you exclude points with anoxic groundwater using a certain criterion. By this you focus on sites with higher risks of nitrate leaching to deeper aquifers. Why this is valid I would think that denitrification potential still can be an important explanatory variable? For example, for the upper soils it is the depth of the groundwater table and presence of organic material (to deliver the nitrate reduction) and deeper also the presence of pyrite in geological formations is an important factor for the denitrification potential. Can you explain, or justify why denitrification potential is not included or considered?
- We worked with nitrates time series from station located high-risk sites (NVZ). We did exclude points with anoxic groundwater to keep this focus. We will indicate this in the manuscript. Only one location was detected as in anoxic conditions and hence removed. The nitrate concentrations in this point were very low (< 2 mg/l). Keeping this point would have highly affected our results. We would have liked to add variables representing the factors affecting denitrification, but again information on organic matter content, or stratification of DOM in groundwater was unavailable. This actually partly motivated a conclusion of our paper: the need for more data and continued monitoring if we want to apply a data-based approach.
L184-186. What do you mean by the “we used the depth of (the bottom part of) the water intake structures as a proxy of the depth to the groundwater table”? In the Netherlands, the depth of groundwater intake for drinking water production (50-100 m) is much deeper than the phreatic groundwater table (1-2 m)
- We would have like to use both the depth of the phreatic groundwater table and the depth of the water intake structures. However, this was not possible because the water producer did not measure the table depth. Additionally, the available observations from the regional institution are sometimes quite far from our monitoring points (https://piezometrie.wallonie.be/home/observations/niveau-deau-souterraine.html?station=DESO%2FPZ475). Using these observations is irrelevant, as the groundwater table depth is spatially variable in the study region.
L186: Why not use rainfall and not precipitation surplus? I assume that this information is available (like in the Netherlands and also in maps).
- Precipitation surplus has been modeled by the EPICgrid project (https://orbi.uliege.be/bitstream/2268/153076/1/Poster_SPGE_DGARNE_121010.pdf). We did not use this information at first as we did not have access to it. As the explanatory power of precipitation was low, we did not ask for accessing this information.
L197-2001. Why use land cover and land use as proxies for nitrogen load while there is trend information on N input and N surplus available, regarding input, also per crop for Wallonia, and also mapped. See e.g.
- Our goal is not to create the best possible model of groundwater nitrate concentrations, but to identify the primary risk factors. We thus used the risk factors we wanted to investigate as explanatory variables. We selected the crop classes based on the available estimations of nitrogen surplus: meadows have the lowest mean value, while potatoes have the highest. We did not consider all crop classes for which the mean N surplus is available, as doing so would have significantly increased the collinearity in our dataset. We also did not use the existing maps N-input and N surplus maps directly, because these maps use inter- and extrapolation procedures to generated regionalized maps which potentially may introduce bias. Since our approach is a purely data driven approach, we used data that were directly regionally available (this is the case for land cover and land use and not for N input and N surplus).
The evolution of the (modelled) N load to groundwater in the figure below looks quite similar to that of the observed nitrate concentration in your Figure 4, but with an apparent delay of 10-20 years. This indicates that a more process-oriented approach is better and possible.
- Figure 4 shows a time series of nitrate concentrations measured at a single location that is not necessarily illustrative of the others. That being said, we will make sure to use the information in this graph to interpret our results.
Also, Eurostat provides time trends for N input and surplus, but only for Belgium as a whole. https://ec.europa.eu/eurostat/databrowser/view/aei_pr_gnb__custom_12226009/default/table?la ng=en
See below the evolution of the Gross Nitrogen Balance (kg N/ha). This type of data is available and provides direct insight in the evolution of nitrogen loads on the aquifer. I found it very surprising that this type of data is not used for this paper, as it obviously is an important explaining variable.
- We will integrate that information in the new section “Nitrogen cycle and budget in Wallonia” and use it to interpret and discuss our results. Yet, the information of N input and N surplus is partially based on inter- and extrapolation modelling procedure which may introduce bias in a pure data driven methodology. These data were therefore not used in our study.
I am pretty sure there is also data on N loads form other sources than agriculture which could help to provide insight into the relative importance of different sources of nitrates and distinction of diffuse (non-point) and point sources (see e.g. the European Nitrogen Assessment4, 2011, Chapter 16; Leip, A., Achermann, B., Billen, G., Bleeker, A., Bouwman, A., de Vries, W., ... & Winiwarter, W. (2011). Integrating nitrogen fluxes at the European scale). I miss an overview of the relative strengths of all nitrogen sources as well as their potential contribution to polluting groundwater resources for drinking water production.
- We will integrate the different nitrogen fluxes in the new section ‘'Nitrogen cycle and budget in Wallonia”
Table 2 Elaborating on the previous point, I found it surprising that the authors did not include any variable in Table 2 related to N loading of the aquifer and specific measures as in the PGDA, and how these change over time. The included variables related to land use, and, in case of meadow, the area trends are, in my view, very coarse proxies, especially for these N loads. I can understand that your choice for land use as an explanatory variable allows a spatially explicit approach, however the website of the “État de l'environnement Wallon” also shows a map of modelled nitrate (http://etat.environnement.wallonie.be/contents/indicatorsheets/SOLS%204.html). This official publication by the Walloon government appears more advanced than yours.
- As previously mentioned, we aim for a pure data-driven approach and avoid proxies which are based on modelling results and subjected to possible bias. The modelled N at the bottom of the soil profile was therefore not considered as an appropriate proxy for N load in our study. It would be interesting to compare the space-time dynamics of N load estimates inferred from our proxies with those modelled with a spatially distributed N model as published on the “etat.environnement.wallonie” web site, but this was not in the scope of our study. In addition, the purpose of the Walloon government publication and our study are not the same: the first one aims to assess the N budget of the agricultural sector, among which percolation to the groundwater, whereas we aim to assess the evolution of groundwater nitrate concentrations in rural and semi-rural areas, and the factors affecting the observed changes. Confusion might arise from the title of our study, which we suggest to change to “Assessing the long-term effectiveness of nitrogen management for groundwater protection in nitrate vulnerable zones in Wallonia, Belgium"
I also suspect that several of the “potential explanatory variables” are correlated, e.g., land use by meadow, crops and forest and green (nature?) must add up. Before you start the correlation and regression analysis for observed nitrate it is wise to check for these autocorrelations. This information perhaps also can be used to reduce the number of variables.
- We did check for collinearity using the variance inflation factor before doing the regression analysis. Because of this we removed some variables of the dataset which showed high collinearity. This will be reemphasized when discussing the results.
L203-206. I assume by “punctual” you “mean” point sources? But more importantly, I miss a clear motivation. Common sense is that in northwest Europe or the EU in general, non-point pollution from agriculture is the dominant source of soil and groundwater pollution. When focusing on specific regions for drinking water production this can be different as the groundwater abstraction areas are smaller and often protected areas. For example, I would be surprised when (active) graveyards or large sceptics tanks are located (or allowed) in drinking water abstraction areas. Please explain.
- Mattern et al. (2011) showed that for the Brusselian sands groundwater body, one of the four groundwater bodies studied in our paper , point sources contribute to its pollution. While new graveyards are forbidden in the legal protection areas, there is no regulation on existing ones. Septic tanks must be rainproof, however we do not know too which extent this is the case. Regulations are applicable to the legal protection zone, we consider a bigger influence zone, including the watershed, hence including zones in which these regulations are not applicable.
Mattern, S., Sebilo, M., & Vanclooster, M. (2011). Identification of the nitrate contamination sources of the Brusselian sands groundwater body (Belgium) using a dual-isotope approach. Isotopes in Environmental and Health Studies, 47(3), 297–315. https://doi.org/10.1080/10256016.2011.604127
Table 4: Why do you average descriptive statistics for all 4 sites, while their characteristics are quite different?
- This approach provides a general overview of the statistics without overloading the paper. Figure 5 discriminates the data across the four sites. We suggest adding the descriptive statistics for each site separately in the Supplementary Material.
The histograms in Figure 5, are quite original but also a bit unusual and the “count” is lumped way of presenting the difference in nitrate levels and trends between the four groundwater bodies.
L313-217: Why “trend likely due to higher conductivity and renewal rate”? Don’t trend and legacy effects also depend on other factors, like denitrification, the volume of the aquifer. Intuitively you may expect that sandy aquifers respond more quicky to changes in total N load than chalk aquifers, but if the volumes are very different, I would not be so sure.
- We agree with this and will modify this statement in the revision.
Figure 5 to some extent confirms that monitoring sites in the sandy groundwater bodies show more decreasing nitrate trends.
- Yes, we will insist on that in the text.
L325. Indeed, potato is known for having high N surpluses, while grassland is known to have low N surpluses. But the nitrate effect also strongly depends on the denitrification potential of the soil and aquifer. There is quite a history of literature on this in the Netherlands and I guess also the Flemish region (see e.g. Dico Fraters, Ton van Leeuwen, Leo Boumans & Joan Reijs (2015) Use of long-term monitoring data to derive a relationship between nitrogen surplus and nitrate leaching for grassland and arable land on well-drained sandy soils in the Netherlands, Acta Agriculturae Scandinavica, Section B — Soil & Plant Science, 65:sup2, 144-154, DOI: 10.1080/09064710.2014.956789). Also keep in mind that potato is always cultivated as part of a rotation of crops (e.g., a grain and sugar beet). Furthermore, in the previously given website for the Walloon nitrogen data (http://etat.environnement.wallonie.be/contents/indicatorsheets/SOLS%204.html ) you can find data for this (see below), which show both that potato is a crop with higher nitrate risks but differences with other major crops (cereals, sugar beet, maize) change over time. Why did you not use this type of information or team up with people collecting and interpreting this data?
- As explained above, we use a pure data-driven approach and avoid using data that are based on inter- and extrapolation procedure.
Citation: https://doi.org/10.5194/hess-2024-173-AC2 - We considered using gross nitrogen balance (in particular, the indicator “Nitrogen with potential to leach” APL (Azote Potentiellement Lessivable)) data, but ran into the issue that for the Walloon region, APL is an indicator available at punctual locations, not on a consistent regional-basis. While regional data are available, as pointed out by the reviewer, they are predicted from models, which is against our approach.
-
AC2: 'Reply on RC2', Elise Verstraeten, 19 Aug 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 303 | 63 | 23 | 389 | 11 | 11 |
- HTML: 303
- PDF: 63
- XML: 23
- Total: 389
- BibTeX: 11
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1