the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 May 2024
Nitrate-nitrogen dynamics in response to forestry harvesting and climate variability: Four years of UV nitrate sensor data in a shallow, gravel aquifer
Abstract. The leaching of inorganic nitrogen can adversely affect groundwater and hydrologically connected streams and rivers. Traditionally, these effects have been assessed using discrete water quality measurements. However, it is difficult to characterise the complex biogeochemical processes that control nitrate-nitrogen dynamics in groundwater using temporally sparse data. In this study, we installed a continuous UV nitrate sensor, downgradient of forestry land use in a shallow, gravel aquifer to understand nitrate-nitrogen dynamics in groundwater. We found that there were two mechanisms of nitrate-nitrogen pulses in groundwater from the upgradient forestry land use. The most prevalent were nutrient losses during winter months when plant uptake is lower. Outside of winter months, we observed a higher nitrate-nitrogen concentration (12 mg L-1) as a result of changing biogeochemical conditions after trees were harvested, compared to 5.9 mg L-1 when there was no harvesting. We used a novel hysteresis approach, comparing nitrate-nitrogen concentrations and groundwater levels after rainfall recharge to understand event scale variability. First flush events in winter had a larger average area (more hysteresis) of 0.65 compared to 0.35 (less hysteresis) for subsequent events. Peak concentrations occurred earlier in events during 2021 (wetter) compared to 2020 (dryer), highlighting slower drainage pathways in years with less recharge. Through this analysis we also found evidence that the mobilisation of nitrate-nitrogen shifted from rainfall recharge to rising groundwater levels after the surface supply was depleted from successive recharge events. Finally, the nitrate-nitrogen load analysis indicates that the leaching and export occurs in pulses, that discrete sampling cannot accurately characterise. For example, in 2021, over 80 percent of the exported load occurred during a quarter of the year and discharged when there were base flow conditions in the nearby Hurunui River. These findings have implications for forestry land management, the understanding of inorganic nitrogen dynamics in groundwater in response to rainfall recharge and can be applied to future climate projections where periods of drought and storm events are more frequent.
- Preprint
(1835 KB) - Metadata XML
-
Supplement
(149 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-964', Anonymous Referee #1, 18 Jun 2024
General comments
The manuscript titled "Nitrate-nitrogen dynamics in response to forestry harvesting and climate variability: Four years of UV nitrate sensor data in a shallow, gravel aquifer" provides valuable insights into the nitrate-nitrogen dynamics influenced by forestry cutting, variation in rainfall (recharge) and groundwater level. The authors effectively used continuous UV nitrate sensors to provide 4 years of data at 15-mins intervals, allowing for detailed temporal analysis in quantifying the amount and timing of the groundwater nitrate-nitrogen leaching and its mobilization and transport mechanisms with hysteresis analysis. However, there are areas that need further clarification and improvement. In general,
- Clear research objectives are addressed, but there is no leading research question(s) presented.
- The introduction provides a good overview of the problems and methods but could benefit from highlighting the consequences of disturbance on forestry nutrient leaching, e.g., concluding or comparing the difference in nitrate-nitrogen concentration before and after ecosystem change for the six disturbed forests listed in Table 1 (if data is available in your references).
- Although only the hysteresis analysis for winter months is shown in your supplement, do you have any findings or insights for the correlation between groundwater level and nitrate-nitrogen leaching during droughts?
Specific comments
- Line 160: Add the coordinates to the monitoring site.
- Line 170: How do you account for the effects of nitrogen leaching from irrigated beef on the nitrate-nitrogen concentration in the Hurunui River?
- Line 173, Figure 1: Is it possible to show the location of the lysimeter on the map?
- Line 179: How does the path length setting of the nitrate sensor affect its measurement accuracy? What might be the causes for the overestimation of high nitrate-nitrogen concentrations?
- Line 190: What’s the precision of the TriOS NICO sensor during the study period?
- Line 207-208: How did you adjust/calibrate the UV nitrate sensors?
- Line 213-215: Which land surface recharge model did you use? Is it a 1-D model? What is its principle for calculating land surface recharge? How did you define the values of parameters? From your reference, I learnt that the model is named the soil-water balance model (GDA-LSR). It’s better to introduce a bit of the model and justify model performance because rainfall recharge is one of your main studied variables.
- Line 225: At what depth was the UV nitrate sensor installed in the Hurunui River?
- Line 227: What is the difference between “rainfall recharge” here and that calculated from the land surface model in Line 212?
- Line 241: What’s the HARP (full name?) algorithm, and what do values represent in Line 348, e.g., 0.65, 0.35 and 50% and 55%?
- Line 259: Does “the volume of water moving in a horizontal direction from the aquifer to the Hurunui Rive” refer to the same as groundwater discharge (qi) in Eq. (1)? Can you introduce the equation used for calculating qi?
- Line 275: Add units to each variable used in the equations and provide clear definitions.
- Line 291: Why did the groundwater level remain low and stable from 2020 to June 2021?
- Line 306: Include a statement clarifying the two distinct climate types and how they influenced the rates of nitrate-nitrogen leaching.
- Line 355: What is the residual analysis for and what is meant by “not changed system state” in Fig. 4?
- Line 383: How did you end up with the conclusion of “The annual and quarterly measurements over or underestimated the load depending on whether they captured the winter pulses of nitrate-nitrogen” from the data in Table 3?
- Line 385: “As a sensitivity analysis” for groundwater discharge calculation?
- Line 402: “We infer that the reduced nutrient demand, increased rainfall and favourable soil water balance conditions induced higher nutrient losses from the forest soils during winter”, where the “favourable soil water balance conditions” is a general indication, could you specify what kind of soil water balance conditions based on your findings in this study?
Technical comment
- Line 128: Add a punctuation mark before “Conversely”.
- Line 168: Add a comma before “sections”.
- Line 192: Add a punctuation mark after “(Fig. 2)”.
- Line 215: Add the unit to “soil water capacity (86.0 mm)”.
- Line 252: Add a comma after the “bore”.
- Line 262: Add a comma after the “…bgl)”.
- Line 301: Just some suggestions for better visualisation of Figure 4, (1) Between two consecutive years on x-axis, add scale bars with monthly timestep will help readers to identify the winter and summer months easier; (2) Label each plot with (a)-(g) for future reference; (3) Combine the last two graphs, e.g., plotting the point series data as the secondary y-axis with different colour scheme on the continuous dataset, to compare lab measurements and UV sensor data while saving space.
- Line 311: Figure 5, the well count for red groups seems to be in the reserve order, then it should range from [0%, 100%] from the top to the bottom at the y-axis on the right hand side.
Good luck and look forward to your replies!
Citation: https://doi.org/10.5194/egusphere-2024-964-RC1 -
AC1: 'Reply on RC1', Ben Wilkins, 04 Jul 2024
Thank you very much for your insightful comments on this paper. I think they will improve the paper and make it more readable. I appreciate the time that has been spent reviewing this paper.
General comments
Clear research objectives are addressed, but there is no leading research question(s) presented.
Thank you for pointing this omission out and we agree that clarification on the leading research question is needed. The manuscript will be improved to include a leading research question on utilising high frequency nitrate-nitrogen data to elucidate the drivers of changes in concentration, groundwater level and concentration relationships and more accurately estimate the exported loads.
The introduction provides a good overview of the problems and methods but could benefit from highlighting the consequences of disturbance on forestry nutrient leaching, e.g., concluding or comparing the difference in nitrate-nitrogen concentration before and after ecosystem change for the six disturbed forests listed in Table 1 (if data is available in your references).
I agree that it would be ideal to compare the difference in nitrate-nitrogen before and after ecosystem changes. However, ecosystem change does not always lead to a step change in nitrate-nitrogen concentrations. Results in other published studies are somewhat inconsistent with some studies reporting large changes before and after disturbance, while others report no change or even a decrease in nitrate-nitrogen concentrations. This is not surprising considering the complex biogeochemical processes controlling leaching and varied monitoring frequencies during studies. It does make commenting on the consequences of disturbance on forestry leaching difficult.
Perhaps I could further highlight the idea that the discrete sampling may not capture the dynamic nitrate-nitrogen changes after ecosystem disturbance, so it is difficult to know how representative the before and after concentrations are. Therefore, it is difficult to determine the consequences of disturbance on forestry nutrient leaching. This highlights the benefits of continuous monitoring.
Although only the hysteresis analysis for winter months is shown in your supplement, do you have any findings or insights for the correlation between groundwater level and nitrate-nitrogen leaching during droughts?
I agree that it would be great to have a comparative analysis between hysteresis in winter months and during periods of low recharge. This is somewhat covered in the discussion where hysteresis in 2020 was found to be more varied, perhaps due to the combination of faster travel times after rainfall recharge and slower pathways because of low groundwater levels. Hysteresis type is also discussed. The supplement shows all hysteresis events during the study period.
Specific comments
Line 160: Add the coordinates to the monitoring site.
Agree. Will add well coordinates.
Line 170: How do you account for the effects of nitrogen leaching from irrigated beef on the nitrate-nitrogen concentration in the Hurunui River?
We explain in the discussion that the effect of the nitrate-nitrogen leaching from the irrigated beef and the forestry harvesting cannot be separated in the observed Hurunui River concentration. We know from previous studies that the nutrient leaching from irrigated beef will be greater than forestry leaching and will therefore contribute more to the observed Hurunui River concentration.
Line 173, Figure 1: Is it possible to show the location of the lysimeter on the map?
Yes, we will show the location of the lysimeter on the map.
Line 179: How does the path length setting of the nitrate sensor affect its measurement accuracy? What might be the causes for the overestimation of high nitrate-nitrogen concentrations?
The TriOS NICO specifications indicate that the accuracy of a 5 mm path length is approximately ± 5% + 0.5 mg/L. Thus, at higher nitrate-nitrogen concentrations there is potentially a larger absolute difference in actual and observed concentrations and therefore less accuracy.
There are a number of reasons why the UV nitrate sensor might not be accurate at higher concentrations:
- It is difficult to have a matrix algorithm that represents the constituents that cause interference during pulses of nitrate-nitrogen because of the short time frame that these pulses occur over.
- There is a trade-off of accuracy between (path length and matrix algorithm) accounting for low concentrations/reduced interference effects during most of the year or accounting for high concentrations/higher interference effects during pulses of nitrate-nitrogen. The current set up has higher accuracy during most of the year and loses accuracy during pulse events.
Overall, the variability of the site means that the accuracy of the UV nitrate sensor cannot be optimised for both high and low concentrations of nitrate-nitrogen (see: Pellerin et al., 2013).
Line 190: What’s the precision of the TriOS NICO sensor during the study period?
In-situ, the precision of the nitrate sensor varies based on different conditions and concentrations. Regular maintenance can also preserve precision (Pellerin et al., 2013).
Compared to the nitrate-nitrogen standards where there are no interference effects, the UV nitrate sensor was most precise at lower nitrate-nitrogen concentrations, similar to Figure 3.
Because of the range of factors controlling precision and the difficulty in measuring real world precision, I prefer not putting a number on precision. I think noting that it is something we controlled is sufficient for the deployment.
Line 207-208: How did you adjust/calibrate the UV nitrate sensors?
Propose changing the start of this sentence to: The recorded UV nitrate sensor measurements.
This sentence is about the adjustment of the UV nitrate sensor data.
Some recorded data from the UV nitrate sensor had obvious offsets from the lab measurements for short periods. These were adjusted to the lab measurements.
Line 213-215: Which land surface recharge model did you use? Is it a 1-D model? What is its principle for calculating land surface recharge? How did you define the values of parameters? From your reference, I learnt that the model is named the soil-water balance model (GDA-LSR). It’s better to introduce a bit of the model and justify model performance because rainfall recharge is one of your main studied variables.
Agree, more description on the model and how the parameters were defined is needed.
It is a 1-D model. Recharge is the rainfall volume above the soil water storage capacity minus evaporation, uptake and runoff.
Parameters were determined from the nearby lysimeter, weather station (PET) and mapped soil properties.
Line 225: At what depth was the UV nitrate sensor installed in the Hurunui River?
The sensor is deployed at different depths between warm and cool seasons to account for changes in base flow conditions.
Line 227: What is the difference between “rainfall recharge” here and that calculated from the land surface model in Line 212?
Groundwater levels are the main focus in section 3.4. These give some indication of regional annual recharge. This is to show the differences in conditions over the study period but the recharge has not been calculated. Rather at the site, the recharge has been calculated using the LSR model. We will rewrite the sentence to clarify this distinction.
Line 241: What’s the HARP (full name?) algorithm, and what do values represent in Line 348, e.g., 0.65, 0.35 and 50% and 55%?
Agree, can add the full name here before the acronym.
Hysteresis, Area, Residual and Peak.
The algorithm is an R script, so it is probably better to link the github address where it can be accessed.
In the results section (Line 348) the decimal values indicate the area of the hysteresis curve.
The percentages indicate the time to reach peak groundwater level or peak nitrate-nitrogen concentration along the hysteresis curve during the event. We will add a few additional clarifying phrases to the text.
Line 259: Does “the volume of water moving in a horizontal direction from the aquifer to the Hurunui Rive” refer to the same as groundwater discharge (qi) in Eq. (1)? Can you introduce the equation used for calculating qi?
Agree, can introduce Darcy’s Law here.
Line 275: Add units to each variable used in the equations and provide clear definitions.
Agree, can add variables from the equation to lines 263-265 to show what we used in this study.
Line 291: Why did the groundwater level remain low and stable from 2020 to June 2021?
Agree, can add detail.
There were low groundwater levels due to low rainfall during 2020 and the first half of 2021.
Line 306: Include a statement clarifying the two distinct climate types and how they influenced the rates of nitrate-nitrogen leaching.
Agree, can add detail.
Low rainfall that resulted in declining groundwater levels, followed by a high rainfall storm event with more consistent rainfall following.
Line 355: What is the residual analysis for and what is meant by “not changed system state” in Fig. 4?
Can add (low nitrate-nitrogen concentrations) after pre-event conditions.
The residual analysis is part of the HARP analysis suite to determine if the environment shifts to higher concentrations, or a new system state, such as eutrophication or another biological condition. It was not included in the analysis as it did not apply to the results due to concentrations returning to pre-event conditions.
Line 383: How did you end up with the conclusion of “The annual and quarterly measurements over or underestimated the load depending on whether they captured the winter pulses of nitrate-nitrogen” from the data in Table 3?
Can add figure 2 to this statement as it is a useful visualisation.
If the discrete annual or quarterly measurement captures an elevated nitrate-nitrogen concentration then it is above the integration load estimation, if not then it underestimates the load by not accounting for the export during the pulses of nitrate-nitrogen.
Line 385: “As a sensitivity analysis” for groundwater discharge calculation?
Agree that more detail is needed here.
There are a range of potential conductivity and hydraulic gradient values that could be used to calculate the discharge of groundwater.
We used a range of potential literature values to give some indication of the exported load range. This should be cited and we will amend the text to reflect this.
Line 402: “We infer that the reduced nutrient demand, increased rainfall and favourable soil water balance conditions induced higher nutrient losses from the forest soils during winter”, where the “favourable soil water balance conditions” is a general indication, could you specify what kind of soil water balance conditions based on your findings in this study?
Agree, can be more specific here.
I used this term because studies have shown that higher soil water content induces increased mineralisation of organic nitrogen after rainfall.
Also, that recharge is occurring more consistently in winter, so there is more transport of nitrate-nitrogen to groundwater.
I will split these biological and hydrological components of soil moisture and soil water balance.
Technical comment
- Line 128: Add a punctuation mark before “Conversely”.
- Line 168: Add a comma before “sections”.
- Line 192: Add a punctuation mark after “(Fig. 2)”.
- Line 215: Add the unit to “soil water capacity (86.0 mm)”.
- Line 252: Add a comma after the “bore”.
- Line 262: Add a comma after the “…bgl)”.
- Line 301: Just some suggestions for better visualisation of Figure 4, (1) Between two consecutive years on x-axis, add scale bars with monthly timestep will help readers to identify the winter and summer months easier; (2) Label each plot with (a)-(g) for future reference; (3) Combine the last two graphs, e.g., plotting the point series data as the secondary y-axis with different colour scheme on the continuous dataset, to compare lab measurements and UV sensor data while saving space.
- Line 311: Figure 5, the well count for red groups seems to be in the reserve order, then it should range from [0%, 100%] from the top to the bottom at the y-axis on the right hand side.
I agree with these changes.
For change 7(3), I would prefer keeping the two time series separate, so that they can be easily referred to but I agree that they could be combined to save space if needed.
For change 8, the graph indicates the percentage of wells that are in each category. I think including a more detailed scale bar would be beneficial for readability.
References
Pellerin, B. A., Bergamaschi, B. A., Downing, B. D., Saraceno, J. F., Garrett, J. D., and Olsen, L. D.: Optical techniques for the determination of nitrate in environmental waters: Guidelines for instrument selection, operation, deployment, maintenance, quality assurance, and data reporting, US Geological Survey, 2013.
Citation: https://doi.org/10.5194/egusphere-2024-964-AC1
-
RC2: 'Comment on egusphere-2024-964', Anonymous Referee #2, 03 Sep 2024
I have read the manuscript by Wilkins et al. with great pleasure. I think they make great use of an excellent and unique dataset on nitrate in groundwater. I have a couple of more general comments:
(1) Some details are missing in the methods section (e.g. climate of study area, information on post-processing, more details on the modelling). Details are provided in the detailed comments attached. I also think that the load calculation explanation could use some clarification and it should be better explained how the Hurunui River data is being used.
(2) From only reading the manuscript, it is not clear how many hysteresis events were finally analysed. This could be clarified in the text, as well as in the box plots with the hysteresis metrics.
(3) The manuscript could benefit from an integration of the results and discussion, as the discussion section is relatively short and some of the results are hard to interpret based on the current formulation. It might also make the flow more logical, as section 4.2 currently breaks the flow from the description of the time series nitrogen data to the hysteresis analysis which uses the same data.
(4) I am slightly sceptical about the use of a change in nitrate concentration to identify events, especially since the threshold of 1 mg/L is not necessarily justified. Similar to other event-based analyses, I would rather let it be guided by the occurrence of rainfall recharge. I think the authors need to better justify their choice of event identification.
The points above are mostly described in more detail in the minor comments in the pdf file attached and, at this stage, do not require a point-by-point response, but should ideally be taken into account when revising the manuscript.
-
AC2: 'Reply on RC2', Ben Wilkins, 16 Sep 2024
RC2 comments
Author comments
Thank you very much for your comments. The comments really highlight where the manuscript needs more detail or clarity and the suggested changes are very helpful. Thank you for your time spent reviewing this manuscript and I’m glad to hear it was interesting.
RC2 main points
(1) Some details are missing in the methods section (e.g. climate of study area, information on post-processing, more details on the modelling). Details are provided in the detailed comments attached. I also think that the load calculation explanation could use some clarification and it should be better explained how the Hurunui River data is being used.
(2) From only reading the manuscript, it is not clear how many hysteresis events were finally analysed. This could be clarified in the text, as well as in the box plots with the hysteresis metrics.
(3) The manuscript could benefit from an integration of the results and discussion, as the discussion section is relatively short and some of the results are hard to interpret based on the current formulation. It might also make the flow more logical, as section 4.2 currently breaks the flow from the description of the time series nitrogen data to the hysteresis analysis which uses the same data.
(4) I am slightly sceptical about the use of a change in nitrate concentration to identify events, especially since the threshold of 1 mg/L is not necessarily justified. Similar to other event-based analyses, I would rather let it be guided by the occurrence of rainfall recharge. I think the authors need to better justify their choice of event identification.
The points above are mostly described in more detail in the minor comments in the pdf file attached and, at this stage, do not require a point-by-point response, but should ideally be taken into account when revising the manuscript.
I agree that the above points are important to consider when revising the manuscript and have provided a detailed response in the points below.
Minor comments
L10: ‘discrete’ – I would suggest to use something like ‘low frequency’, because even in situ
sensors provide discrete measurements.
We agree that the terms used to describe the measurement types needs some thought and adjustment. We will have to consider the terminology used in other similar studies to try and be consistent. Having a description of the measurement type and a comment early in the manuscript that it will be referred to, for example, as ‘discrete’ or ‘high frequency’ henceforth will provide more clarity.
L12: ‘continuous’ – would not use this as adjective, as the sensor is not continuous. You could
emphasise the high frequency data collection with ‘measuring at xxx minute interval’. Or call
them ‘high-frequency sensors’, as you do in L99 (and please use the same term consistently).
See the response to the previous comment. We agree that the terminology for measurement type should be clearer and more consistent.
L16 and whole document: are concentration in mg NO3 per L or mg N per L? Please clarify in the manuscript.
Concentrations are reported as nitrate-nitrogen (NO3-N) in mg/L throughout the document. We can include a statement early in the manuscript (line 134) to make sure this is clear, especially when we refer to the ‘UV nitrate sensor’, which may cause confusion.
L19-20: ‘larger average area’ – What are the implications of a larger or smaller area (or more or less hysteresis)? What can we learn from this information? Might be too complicated to include in the abstract and needs to be explained properly in the main text.
The area represents the degree of hysteresis, which in this context is the lag (or difference in rate of change) between increasing/decreasing nitrate-nitrogen and groundwater levels.
The key aspect, which needs further detail in the manuscript (lines 115 to 125), is linking the observed shape of the hysteresis curve to processes or physical properties within the catchment.
For example, the hysteresis curves of the first flush events with larger areas indicate the initial recharge has a higher nitrate-nitrogen concentration than subsequent recharge events. Linking this finding back to physical properties and processes in the catchment, the results indicate that the source of nitrate-nitrogen is limited, the source has a short travel path to the UV nitrate sensor and is readily mobilised.
L44: ‘forestry ecosystems’ – Does the description of processes in the following sentences refer
to forestry ecosystems only (which, I presume, are plantation or intensively managed forests
only?) or all forest ecosystems? If it applies to all forests, I would use the term ‘forest
ecosystems’ instead.
We think that forestry is the more accurate term as the site does not have the full multi-canopy ecology of a forest. The references used to describe the processes are focused on forestry disturbance.
However, we can specify that the processes in line 45 applies to both forestry and forest ecosystems for clarification.
Table 1: Include the abbreviation ‘NO3-N’ in the caption behind ‘nitrate-nitrogen’. Interesting
additional information for the table would be an indication of the time period over which each
study was conducted and potentially even the method used to obtain the nitrate data.
Agree, we are happy to add that detail to the table.
L100-103: I presume that you mean that high frequency monitoring could be done in multiple
places to understand the high variability in inorganic N movement? Or do you mean that one
high frequency sensor can help to better understand the spatial variability as well? It might be
good to clarify this in the sentence.
Lines 100-103 refers to the value of using high frequency monitoring in a range of settings given the variability of inorganic N movement. This sentence can be improved to make this point clearer.
L133-140: While the objective makes sense for a case study perspective, the ‘bigger picture’ is
missing. Why is this study relevant for others than those working in the case study location?
What can be learned from it, applied elsewhere?
Agree, that we can expand on wider impact of these results and methodology. The main ideas of the bullet points below would be included in the discussion and the summary in the abstract and conclusion. For example:
- The study highlights the value of high frequency monitoring where there are rapid changes in nitrate-nitrogen concentrations.
- High frequency data helps with identifying the drivers of nitrate-nitrogen pulses.
- Hysteresis analysis can be applied outside of traditional CQ relationships to groundwater levels and concentrations. This allows further investigation of nitrate-nitrogen dynamics within the vadose zone and aquifer.
- High frequency data can improve groundwater load estimates where nitrate-nitrogen concentrations are variable.
Summary: The study shows that high frequency measurements of nitrate-nitrogen and groundwater levels along with the analysis techniques that it enables can help elucidate nitrate-nitrogen dynamics from source to discharge point.
Section 2: You could include some more background information on the climate in the study area, particularly rainfall amounts and seasonality (if applicable) and evapotranspiration.
Agree, we can add this information to this section.
L177: ‘and telemetry’ – Is this really described in this section? I could not identify it.
Correct, it is not included. We can delete telemetry from the heading.
L207-208: ‘were adjusted to concurrent validation measurement’ – This sound to me like you
calibrated the sensor using the validation sample measurements (and the annual/quarterly
measurements?). Perhaps use the word ‘calibration’ instead of your current formulation?
Agree, calibrated is a better word. Validation samples were used for calibration but an annual or quarterly measurement that was different from the nitrate sensor might initiate taking another validation sample.
Section 3.1: Did you do any data post-processing or quality control (other than comparison to
grab samples) to the time series data? E.g. outlier detection, gap-filling, etc. If so, please
describe.
Agree, more detail is needed on post-processing in this section.
- We reviewed the data for outliers by comparing high frequency and discrete samples and adjusting the time series to discrete measurements when the difference was more than the accuracy of the UV nitrate sensor.
- We also removed obvious outliers, typically single measurements.
- We identified step changes in the UV nitrate sensor, which indicated interference effects. The step changes were obvious and were removed by using linear interpolation between two points that did not exhibit interference effects.
- We identified time points where there was no measurement due to a loss of signal or error. These were typically a singular 15-minute interval where there was no data recorded. We used linear interpolation to fill these gaps.
Section 3.2: It is not clear on which information the choice of parameter values for soil water
capacity, evaporation reduction function, crop factor and drainage threshold were based, nor
where the data for precipitation (from the rain gauge next to the well?) and Penman PET were
obtained. Please add this information.
Agree, we can add this information.
Section 3.3 and 3.4: It is not entirely clear to me how these data will be used (i.e. related to the
groundwater N data) based on the description. It would be helpful if this could be highlighted in
a sentence in each of the sections.
Fair point. We can add more context on how this data will be used.
L225-226: Was the sensor also adjusted using these monthly grab samples, like with the sensor
installed in the well?
Yes, we can add this detail to the manuscript
L235: This is probably the question you do not want to get, but how was the 1 mg/L threshold
determined?
Good question. Currently the 1 mg/L does seem arbitrary and needs more explanation in the manuscript. We can add further details to the manuscript at line 233.
We considered the UV nitrate sensor and groundwater level accuracy. We found that we could not reliably identify or analyse hysteresis curves below 1 mg/L based on the accuracy of the measurements.
We also considered the baseline conditions, which are generally below 1 mg/L but do vary within that range. We wanted the events to be clear and obvious and 1 mg/L seemed to be a reasonable cutoff.
We did not define an event with a requirement of rainfall recharge previously occurring as we identified at least one event induced by rising groundwater levels intercepting vadose zone nitrate-nitrogen storage.
L235-237: What was the point of splitting up the time period? Was the rate of change determined for each of the five sections?
The hysteresis curve was split into five equal sections to visually show the rate of change. We can improve the explanation on this in the manuscript here. We think it helps visualise the rate change in concentrations and groundwater levels over equal time periods. For example, a fast initial rate of change (recharge) and slower return (discharge) to baseline conditions.
The rate of change was not calculated as we thought the proportion of time that peak groundwater/nitrate-nitrogen was reached into the event in the HARP analysis was a similar analysis.
L241-243: Could you explain the relevance of the different metrics? What do they indicate?
We used the HARP analysis from Roberts et al. (2023) and the R script they developed. We can improve the description of metrics here, so the results are easier to interpret.
The hysteresis curves can be linked to processes and physical properties in the catchment that influence the shape of the hysteresis curves during recharge events.
The metrics are relevant because they allow comparisons between recharge events, which we have used to understand how nitrate-nitrogen dynamics evolve throughout successive recharge events, or between conditions in different years.
The metrics in the study’s setting can be used to determine if the catchment is source limited or transport limited. At a smaller scale hysteresis curves can indicate flushing or dilution. For example, at Balmoral it can be linked to how easily the nitrate-nitrogen is mobilised, the distance from the source to the UV nitrate sensor and how permeable the gravel aquifer is.
Hysteresis Area: The area within the hysteresis curve is the hysteresis area, which is calculated using intergration or piecewise integration for complex (figure eight) hysteresis curves. It represents the lag between increases/decreases in nitrate-nitrogen and groundwater level. A large hysteresis area represents
Residual: This metric looks at the difference between initial and end concentrations of the hysteresis curve. It can be used to analyse if the initial conditions will persist or if the system will enter a new state. In Roberts et al. (2023) they use the example of eutrophication in a surface water body as a system state change.
We determined that this metric was not relevant to this case study because after the pulse of nitrate-nitrogen, the concentration returns to baseline conditions.
Peak: Refers to the percentage of time taken during the event for the peak concentration or groundwater level to occur. It is useful for comparisons between events as the source of nitrate-nitrogen may be progressively depleted (longer time for peak nitrate-nitrogen) or there is increased hydrological connection after successive recharge events (e.g. increased soil moisture and a smaller vadose zone) resulting in peak groundwater level occurring earlier.
L245-246: Since you mention how many events were identified in 2022 and 2023, it would be
interesting to report in this paragraph as well, how many events were identified for 2020 and 2021.
Agree, we can add this detail.
L252: ‘bore’ – Should be ‘borehole’?
In New Zealand we refer to these as a bore, or a monitoring bore, if it is instrumented, and as such have used the local terminology; however, we will change the terminology as requested for the global audience.
L252-257: It is not entirely clear what you are describing here. I think you are not calculating the nitrate load yet, since you are only using the concentration data. The load calculation itself seem sto be described in L265-267, where you multiply the concentration with the volume ofgroundwater discharged from the aquifer. Or how does the area under the curve represent the load?
Agreed this section needs to use the right terminology. The section describes how the area under the UV nitrate sensor data was calculated with units of mg/L per unit of time, which is not the load as it does not have the volume component. We can refer to it as mg/L per unit of time and clarify that the load is the product of the discharge volume and concentration.
L294-295: ‘In November 2020, […] flow and rainfall’ – This comparison is a bit odd. What is the point of comparing the highest concentration in Nov. 2020 with the highest concentration in Feb. 2022? Unless you want to point something out (which you would have to specify), I would take it out or reformulate in a way that it does not become an odd comparison.
The comparison between the difference in observed concentrations outside of winter months is setting up the point covered in the discussion about recent forestry harvesting increasing nitrate-nitrogen availability. We also use it to compare dry season nitrate-nitrogen pulses compared to the more prevalent winter flushing. We can reformulate it so that there is more context.
L298-299: ‘Data was missing […] and interference effects.’ – This information could be included in the methods, including information on how much data (%) is missing. Were there no data gaps at all for the sensor deployed in the groundwater? Please specifically mention this (or the percentage of data missing) in the methods as well. I was wondering whether the sudden drops in nitrogen concentrations in the Hurunui River were also related to maintenance (e.g. cleaning),but since some of the increases are also reflected in the monthly nitrate data, I might be mistaken.
We cover the groundwater data in an earlier comment (under Section 3.1).
The Hurunui River UV nitrate sensor data is of a lower quality given it’s setting in a braided river with more interference effects.
Where there is no data on the Hurunui River UV nitrate sensor graph, the sensor was either out of the water for maintenance or the Sensor Quality Index and comparisons to discrete samples indicated that the data was not accurate.
We can add these details to the manuscript.
L306: I would refer to Fig. 5 in L307, following ‘[…] low rainfall recharge in Canterbury.’
Agree, we can make this change.
L307: Where is Balmoral? How does this site relate to the sites included in Fig. 5?
We can provide the percentiles of the groundwater level in the Balmoral monitoring bore to provide more context in this section.
Section 4.2: These sections seem a bit isolated from the rest of the manuscript. They could be
embedded better by directly linking it to the groundwater nitrogen data (although that might
require merging of result and discussion?) or they could be placed somewhere else than between
the initial description of the groundwater nitrogen data and the hysteresis analysis, which is
based on the same data.
We agree that this section needs more context to link the results shown here to the Balmoral data. We propose adding an explanation after heading 4.2 before presenting the data and adding the data from Balmoral to section 4.2.1, so that the groundwater level trends as an indication of rainfall recharge are clearer.
Another option is to present the analysis in 4.2.1 and 4.2.2 before 4.1 as a way to explain why we see the results in the timeseries.
Section 4.3/Fig. 7-8: Are these six the only hysteresis curves you analysed (i.e. the only identified
events)? This could be clarified by including in the text how many events were identified and
analysed.
Agree, we can specify the total number of events analysed. For clarification these are interesting hysteresis events that were selected and are referred to in the discussion.
L348-356: A lot of this information is difficult to interpret, because it is not clear what the relevance of the average area is and incomplete formulation. For example, in the methods you state ‘the proportion of time into the event that the peak groundwater level and nitrate-nitrogen concentration [occurs]’, which is a lot easier to understand than ‘time to reach peak groundwater level and peak nitrate-nitrogen concentration’ expressed in a percentage. Or ‘The residual analysis indicated a return to pre-event conditions…’, was this done for all events individually?
Each hysteresis curve was analysed individually, so there is hysteresis area and time to peak groundwater level/nitrate-nitrogen concentration data for each hysteresis curve. This could go in the supplementary information.
We can ensure that the terminology used is consistent.
Further explanation on the meaning of the HARP metrics is available in response to other questions and this detail can be incorporated into the manuscript to make the results easier to interpret.
Fig. 9: Please include the unit (%) for the first two graphs and add ‘of hysteresis curve’ to the
label of the y-axis of the third graph. Also include on how many events these boxplots are based.
Agree, we can add these details.
L382-383, Table 3: If I understand correctly, these loads are based on the groundwater nitrate
data and groundwater validation samples. Did you also compare these loads with export values
calculated from the river nitrate data (at least for those periods for which there is data available
for both sensors)?
That is correct, they are based on the groundwater data.
We did not calculate the load for the Hurunui River. It could be an interesting addition but we are mindful of the word count and the accuracy of this data.
L388-395: It would be interesting to see a time series plot of the load data to better visualise the
pulses and their timing.
That is a good point. We have graphed the load data, although it is not in the manuscript, and the load data is very similar to the UV nitrate sensor nitrate-nitrogen time series, as expected. We could add a figure showing the load or refer to this finding specifically in the text.
Section 5: Since the discussion is relatively short, I would recommend integrating it with the
results. This avoids duplication of information and might make the interpretation of some of the
data also more straightforward (see previous comment about interpretation of hysteresis
metrics).
We would prefer to improve the methods and results section, based on the suggestions in the review comments, so that the discussion is easier to interpret. However, we would be open to a combined results and discussion section if all reviewers and the editor support this idea.
L404-407: Can you really fully attribute these differences to the harvesting? You point out in Section 4.2.1 that this period is also characterised by a change in weather (not climate!) from dry conditions to wetter conditions.
We agree that there is a measure of equal attribution to the differences as harvesting was concurrent with a transition in soil moisture conditions, and can change the manuscript to reflect that this cannot be fully resolved.
In this context we are referring to climate variability, that is the change in conditions that last longer than discrete weather events. It would not be appropriate to refer to these phases as “weather” which in the New Zealand context are short-lived, often frontal or cyclonic changes in conditions – which induce specific event responses.
We agree that “climate” itself may not be appropriate given that it usually refers to mean normal decadally-averaged conditions; however, in the context here the conditions of sustained dry conditions are a type of internal climate variability, which are locally associated with atmospheric oscillations, some of these oscillations are relatively short lived (i.e. weeks to months like MJO and SAM) whereas others may persist for longer (i.e years to decades, like ENSO, IPO/PDO). These climatic variability measures may be associated with sustained dry periods as the polar jet stream migrates (or bifurcates) and are associated with producing weather phenomena (especially blocking highs) that lead to dry conditions; or lead to high rainfall producing phenomena by drawing down more tropical and subtropical weather systems into the South Pacific, which may lead to sustained periods of higher moisture transport.
Such an analysis of the underlying factors of sustained dry spells and droughts are outside of the scope of this paper, but we can change the language in text to refer to changes in mean conditions.
L426-428: ‘Over consecutive recharge events […] after successive recharge events.’ – either
remove the end or the beginning of the sentence to avoid duplication.
Agree, we can make this sentence more succinct by removing successive recharge events.
L437-438: I see how this approach could complement the more ‘traditional’ analysis of concentration-discharge relationships or hysteresis analysis, but I think the formulation that the approach you used can be applied to improve hysteresis analysis in streams and riverine environments in incorrect. It can only be applied in cases whereby groundwater data (concentrations and levels) are available.
We can change the initial sentence to: These findings can be a valuable addition to improve hysteresis analysis in riverine environments.
In some catchments groundwater is a significant contributor to baseflows, which indicates that hysteresis in groundwater is worth considering in traditional CQ relationships, ideally when a study is being conceptualised, so that groundwater concentrations and levels are available.
We can improve this paragraph to convey these ideas.
Citation: https://doi.org/10.5194/egusphere-2024-964-AC2
-
AC2: 'Reply on RC2', Ben Wilkins, 16 Sep 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 256 | 82 | 116 | 454 | 84 | 21 | 23 |
- HTML: 256
- PDF: 82
- XML: 116
- Total: 454
- Supplement: 84
- BibTeX: 21
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1