the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Investigating Relationships Between Nitrogen Inputs and In-Stream Nitrogen Concentrations and Exports Across Catchments in Victoria, Australia
Abstract. Accurate information on nitrogen (N) inputs to the landscape is crucial for understanding and predicting in-stream N concentrations and downstream N export. However, obtaining detailed catchment-scale data remains challenging due to their spatial and temporal variability. This study developed a statistical model based on mean annual rainfall to estimate fertiliser N inputs for four agricultural land uses in Victoria, Australia. These estimates, along with contributions from biological fixation and atmospheric deposition, were used to (a) examine how N inputs relate to stream total nitrogen (TN) concentrations and export, and (b) assess the influence of rainfall, hydrology, and other catchment characteristics on TN export across 59 diverse catchments. The model revealed a strong positive correlation between average rainfall and fertiliser N input for each land use at the Catchment Management Authority (CMA) (i.e., regional) level, with R2 values ranging from 0.55 to 0.72. Stream TN concentrations were strongly correlated with total N inputs (R2 = 0.72) and fertiliser N inputs (R2 = 0.68). Stream TN export also showed significant relationships with total N inputs (R2 = 0.47) and fertiliser N inputs (R2 = 0.53). The proportion of total N inputs exported varied widely, ranging from 1.4 % to 26 %, with an average of 7 %. This variation was strongly influenced by agricultural activity and hydroclimatic factors. Moreover, the average export proportion was notably lower than values reported for other regions globally, which may reflect Australia’s generally lower N input levels. These findings provide a useful tool for water quality assessment and can guide targeted strategies to reduce nitrogen pollution in streams.
- Preprint
(1730 KB) - Metadata XML
-
Supplement
(611 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2456', Anonymous Referee #1, 13 Jul 2025
-
AC1: 'Reply on RC1', Olaleye John Babatunde, 08 Sep 2025
We thank Referee 1 for their constructive review and thoughtful comments. Their feedback has been very helpful in improving the clarity and rigor of our work. Below we respond to each point in turn.
RC1: General Comment.
This paper is very clear and nicely written. The analysis is straightforward. The authors discuss the relationship between N loads, referring to the well-known NANI methodology, and riverine N flows across several catchments in Victoria, Australia. The catchments are predominantly forested with some grazing and dairy farming activity. A critical problem is to estimate the components of NANI to compare with statistical estimates of riverine fluxes (using the WRTDS model). NANI includes fertilizer, N fixation, atmospheric deposition and net food/feed inputs. The authors estimate fertilizer by developing regressions against rainfall for several land use categories, which explain about 55-72% of the variability. Atmospheric deposition is considered a small, spatially invariant value, which essentially has no explanatory power. N fixation estimates are based on crop specific estimates made in a previous study. Net food/feed imports were not included because the authors judged that the data were insufficient to estimate livestock feed, and human food imports were ignored, presumably because the human waste stream was considered insignificant (human wastewater sources were also dismissed as insignificant). Thus, the authors are assuming that the only sources of N in this region are due to fertilizer, as estimated by a rainfall proxy in different land use categories, and crop-dependent N fixation rates developed in an early study. The authors are aware of the limitations of the analysis and explicitly state them in section 4.5. I think it would be useful to either develop proxies for net food/feed inputs (say, based on even rough estimates of livestock numbers and population density as they vary by catchment) or, if this is not possibly, simply frame the analysis as a relationship between rainfall-based estimates of fertilizer N loads and streamflow N, which is effectively what it is.
Author Response:
Thank you for highlighting this important point. We agree that net food/feed inputs are a core component of the NANI framework, and their absence is a limitation of our current analysis. In the revised manuscript we will incorporate estimates of net food/feed inputs using proxies informed by established farm-gate nutrient balance approaches (Gourley et al. 2012; Stott & Gourley 2016; Sargent et al. 2025). Specifically, we will estimate supplementary feed inputs by combining livestock numbers and stocking rates with published values of feed nutrient content and regional survey data (e.g., ABARES Dairy Industry Survey, Dairy Farm Monitor). This addition will provide a defensible proxy for imported feed N at the catchment scale and strengthen the completeness of our NANI framework.
Major Comments RC1
- Line 160: Given the seemingly strong relationship between elevation and rainfall shown in figure 1, it is surprising that elevation is nowhere mentioned as an explanatory variable, nor included in any tables. Was it investigated?
Author Response:
We appreciate this observation. We did not include elevation as an explanatory variable because rainfall was considered the primary driver of spatial variation in our study, given that moisture availability is a limiting resource for productivity across Victoria. Elevation itself is unlikely to control productivity and hence we feel that it is not an appropriate proxy. Furthermore, including both rainfall and elevation would have introduced issues related to collinearity. In addition, the elevation difference across Victoria is relatively small, further limiting its explanatory power. We will clarify this point in the revised manuscript.
- Line 183: Need to cite a reference for QGIS
Author Response:
Thank you for this suggestion. In the revised manuscript we will include the recommended citation for QGIS.
- Line 239: Adams et al 2014 considered only wet deposition of N and did not distinguish between reduced and oxidized forms. Reduced N (including NH3, NH4+, etc) are well known to be associated with volatile losses of NH3 from agricultural sources including manure from dairy herds and fertilizer applications, which are likely relevant here. (The authors note the significance of manure in the context of the discussion of NOx vs TN in agricultural vs forested land covers around line 488.) While the overall term is likely relatively small compared to other sources, using a constant value across the landscape cannot possibly be meaningful to estimate spatially varying loads in the models developed here. Either do a better job of estimating the true atmospheric deposition component and its spatial variation or drop it as insignificant.
Author Response:
We appreciate this thoughtful comment. In our analysis, atmospheric deposition was included as a constant background value across the study area. We acknowledge that this simplification does not capture the spatial variability in reduced versus oxidized forms of deposited N, particularly given agricultural contributions from manure and fertiliser. As the contribution of atmospheric deposition is relatively small compared to other inputs, and because we are unable to resolve its spatial heterogeneity with the available data, we will remove this term from the revised analysis. This will avoid over-interpretation of an uncertain input and allow us to focus on the more significant and spatially variable N sources.
- Line 507: What is “fertiliser additive land use”?
Author Response:
Thank you for pointing out this ambiguity. By “fertiliser additive land use,” we meant the sum of agricultural land-use categories likely to receive fertiliser applications (cropping, dairy, horticulture, mixed farming and grazing, and non-dairy livestock). This terminology was adapted from Mitchell et al. (2009), who used the concept of “fertiliser-additive land use” (FALU) to represent fertiliser-dependent production land in the Tully River catchment, Queensland. To avoid confusion, we will replace this with the simpler term “agricultural land use” in the revised manuscript.
- Line 588: Schaefer et al (2009) actually show lower export as a percent of N inputs in some western US watersheds than Schaefer and Alber (2007) show in southeastern US watersheds, presumably because of the relatively dry environments in much of the west which affect N delivery (as you note in section 4.3).
Author Response:
We thank the reviewer for this useful clarification. We agree that Schaefer et al. (2009) showed lower export percentages (generally <20%, with an average of ~12%) in western U.S. watersheds compared to the higher values reported by Schaefer and Alber (2007) in southeastern U.S. systems. This contrast reflects the drier environments and hydrological variability in the west, which constrain N delivery. We will revise Section 4.2 to explicitly acknowledge this contrast and ensure consistency with our discussion in Section 4.3.
- Section 4.3: A potential difficulty with the discussion of TN export as a fraction in inputs is that the estimated total catchment N inputs may be biased because of incomplete estimates. Some discussion of this specific issue is warranted.
Author Response:
We appreciate this important observation. We agree that uncertainties in estimating total catchment N inputs (e.g., omission or simplification of terms such as net food/feed imports and spatially varying deposition) may bias estimates of TN export as a fraction of inputs. In the revised manuscript, we will add a statement acknowledging that incomplete or spatially uniform input estimates could lead to either underestimation or overestimation of export percentages. We will also note that, despite these uncertainties, the relative patterns across catchments remain robust because the same methodological approach was applied consistently.
Citation: https://doi.org/10.5194/egusphere-2025-2456-AC1
-
AC1: 'Reply on RC1', Olaleye John Babatunde, 08 Sep 2025
-
RC2: 'Comment on egusphere-2025-2456', Anonymous Referee #2, 19 Aug 2025
The study addresses an important knowledge gap by developing a rainfall-based model to estimate fertilizer N inputs and linking them to stream N dynamics across diverse catchments. Its major strengths—based on long-term multi-site monitoring data from a large-scale empirical approach (59 catchments), pragmatic modeling of fertilizer inputs, and demonstration of scalable relationships between N inputs and stream exports—offer actionable insights for regional water quality policy. While some methodological simplifications exist, the work achieves its core goal of establishing baselines for N budgeting in semi-arid landscapes.
Some comments:
Firstly, the fertilizer model relies solely on mean annual rainfall as a predictor for four distinct agricultural land uses (presumably dairy, cropping, etc.). However, the irrigation is also very important in this arid zone. How much the N input by irrigation? And did irrigation rates could affect this model?
Line 53-54 The phrase "complex to manage and modify" is redundant and awkwardly implies that modifying the models (rather than their operation) is the focus.
Line 57, The comparison is syntactically incomplete. "Unlike process-based models" lacks a clear verb to contrast what those models do differently.
Line 58 Absence of "that" before "N undergoes" creates a grammatical error.
Line 126, The phrase "geographic representation across..." is slightly awkward. The concepts of "data availability" and "representation" are not perfectly parallel. The intent is clear but can be expressed more forcefully and directly.
Line 325 The pronoun "That" is slightly vague. While it logically refers to the data collection period of Gourley et al. (2008-2009), the reference can be made more explicit for immediate clarity.
Line 347, "natural soil N" This term is vague. In soil science, the preferred terms are typically "native soil nitrogen" or "soil nitrogen mineralization."
Line 364, The phrase "cropping land use, including horticulture" is problematic. In standard agricultural classification, "horticulture" (intensive, high-value crops like vegetables and fruits) is often considered distinct from broadacre "cropping" (extensive field crops). Grouping them under one "cropping land use" umbrella is confusing and requires justification, as their nutrient management practices differ vastly.
Line 380, While "data" is technically plural, it is often treated as a singular mass noun in scientific writing, especially when referring to a dataset as a whole. Using the singular verb is more modern and common.
Line 451, The space between the numeral and the percent symbol ("43.4 %") is a typesetting style often required by publishers. However, many style guides (e.g., APA) recommend no space ("43.4%"). The key is to be consistent throughout the manuscript.
Line 597, The verb "indicate" is weak and overused in scientific writing. A more assertive and precise verb would strengthen the opening statement.
Citation: https://doi.org/10.5194/egusphere-2025-2456-RC2 -
AC2: 'Reply on RC2', Olaleye John Babatunde, 08 Sep 2025
We thank Referee 2 for their constructive comments, which have been very helpful in improving the clarity and rigor of our work. Below we respond to each point in turn.
RC2: General Comment:
The study addresses an important knowledge gap by developing a rainfall-based model to estimate fertilizer N inputs and linking them to stream N dynamics across diverse catchments. Its major strengths—based on long-term multi-site monitoring data from a large-scale empirical approach (59 catchments), pragmatic modelling of fertilizer inputs, and demonstration of scalable relationships between N inputs and stream exports—offer actionable insights for regional water quality policy. While some methodological simplifications exist, the work achieves its core goal of establishing baselines for N budgeting in semi-arid landscapes.
Author Response:
We thank the reviewer for the positive assessment and constructive feedback.
Major Comment RC2:
Firstly, the fertilizer model relies solely on mean annual rainfall as a predictor for four distinct agricultural land uses (presumably dairy, cropping, etc.). However, irrigation is also very important in this arid zone. How much the N input by irrigation? And did irrigation rates could affect this model?
Author Response:
We thank the reviewer for raising this important point about the role of irrigation in our fertiliser model, particularly in the arid context of Victoria. To clarify, irrigation was included in our analysis (Section 2.3, lines 207–212). Each land parcel was classified as irrigated or non-irrigated based on regional land-use data. For irrigated parcels, we adjusted water inputs by adding mean irrigation water depths (sourced from the Australian Bureau of Statistics for 2016–2017 and 2018–2019 at the Catchment Management Authority scale) to mean annual rainfall, and this adjustment influenced nitrogen input estimates. This step improved the accuracy of fertiliser nitrogen input estimates for irrigated areas, and we will make this aspect of the methodology more explicit in the revised manuscript. Regarding the specific questions raised, we did not calculate a separate nitrogen input from irrigation water itself. Rather, irrigation water volumes were used to scale fertiliser nitrogen inputs for irrigated land uses. In this way, irrigation rates indirectly influenced the model outputs through their effect on fertiliser nitrogen estimates.
Major Comment RC2:
Line 53-54 The phrase "complex to manage and modify" is redundant and awkwardly implies that modifying the models (rather than their operation) is the focus.
Author Response:
We thank the reviewer for this helpful clarification. Our intention was to highlight the challenges of applying these models due to their extensive data requirements, rather than implying difficulties in modifying the models themselves. We would revise the sentence to: “However, these process-based models can be challenging to apply because of their extensive input requirements, including climate variables, soil properties, and agricultural management practices such as crop rotations, fertiliser application, and irrigation”.
Major Comment RC2:
Line 57: The comparison is syntactically incomplete. "Unlike process-based models" lacks a clear verb to contrast what those models do differently.
Author Response:
We agree that the sentence was syntactically incomplete. In the revised manuscript, we will rephrase the sentence to: “Unlike process-based models, which simulate the internal cycling and transformation of nitrogen within the catchment, the NANI approach compares nitrogen inputs to exports without accounting for these internal processes”.
Major Comment RC2:
Line 58: Absence of ‘that’ before ‘N undergoes’ creates a grammatical error.
Author Response:
We agree and will correct this by inserting “that.” The revised sentence will read: “…the NANI approach compares N inputs to exports without accounting for the complex transformations that N undergoes as it moves through the catchment.”
Major Comment RC2:
Line 126: The phrase "geographic representation across..." is slightly awkward. The concepts of "data availability" and "representation" are not perfectly parallel. The intent is clear but can be expressed more forcefully and directly.
Author Response:
We agree and will revise the sentence for clarity and parallel structure. The revised text will read: “These sites were selected to ensure long-term data availability and broad geographic coverage of diverse land uses and climatic conditions across the state.”
Major Comment RC2:
Line 325: The pronoun "That" is slightly vague. While it logically refers to the data collection period of Gourley et al. (2008-2009), the reference can be made more explicit for immediate clarity.
Author Response:
We agree and will revise the sentence to make the reference explicit. The revised text will read: “The 2008–2009 data collection period overlapped with the Millennium Drought, which likely reduced fertiliser application rates and may explain why the median reported by Gourley et al. (2012) is lower than the range observed in this study.”
Major Comment RC2:
Line 347: ‘natural soil N’ is vague. In soil science, the preferred terms are typically ‘native soil nitrogen’ or ‘soil nitrogen mineralization.’”.
Author Response:
We agree and will revise the phrase for clarity. The revised text will read: “…These farms often rely on alternative N sources, such as native soil nitrogen and biological N fixation by leguminous crops, to meet their N requirements (Angus, 2001).”
Major Comment RC2:
Line 364: The phrase "cropping land use, including horticulture" is problematic. In standard agricultural classification, "horticulture" (intensive, high-value crops like vegetables and fruits) is often considered distinct from broadacre "cropping" (extensive field crops). Grouping them under one "cropping land use" umbrella is confusing and requires justification, as their nutrient management practices differ vastly.
Author Response:
We agree and thank the reviewer for this observation. In our dataset, fertiliser N inputs were only available for cropping as a single category, with no specific values reported for horticulture. For this reason, horticulture was grouped with cropping-related land uses in our analysis. We will clarify this in the revised text and add a note acknowledging that this aggregation may mask some differences, as horticultural systems typically involve more intensive nutrient management than broadacre cropping.
Major Comment RC2:
Line 380: While "data" is technically plural, it is often treated as a singular mass noun in scientific writing, especially when referring to a dataset as a whole. Using the singular verb is more modern and common.
Author Response:
We agree and will revise the sentence accordingly. The revised text will read: “The data from Fig. 4 was aggregated to calculate the average TN input across each catchment…”
Major Comment RC2:
Line 451: The space between the numeral and the percent symbol ("43.4 %") is a typesetting style often required by publishers. However, many style guides (e.g., APA) recommend no space ("43.4%"). The key is to be consistent throughout the manuscript.
Author Response:
We agree and will ensure consistency in formatting of percentages throughout the manuscript, following the journal’s preferred style.
Major Comment RC2:
Line 597: The verb "indicate" is weak and overused in scientific writing. A more assertive and precise verb would strengthen the opening statement.
Author Response:
We agree and will revise the sentence to use a stronger verb. The revised text will read: “Our results demonstrate that hydrological factors, specifically runoff, precipitation, and runoff perenniality, are the primary drivers of the percentage of TN export across the studied catchments.”
Citation: https://doi.org/10.5194/egusphere-2025-2456-AC2
-
AC2: 'Reply on RC2', Olaleye John Babatunde, 08 Sep 2025
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
454 | 57 | 21 | 532 | 24 | 11 | 15 |
- HTML: 454
- PDF: 57
- XML: 21
- Total: 532
- Supplement: 24
- BibTeX: 11
- EndNote: 15
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Review of Babatunde et al., “Investigating Relationships Between Nitrogen Inputs and InStream Nitrogen Concentrations and Exports Across Catchments in Victoria, Australia”
This paper is very clear and nicely written. The analysis is straightforward. The authors discuss the relationship between N loads, referring to the well-known NANI methodology, and riverine N flows across several catchments in Victoria, Australia. The catchments are predominantly forested with some grazing and dairy farming activity. A critical problem is to estimate the components of NANI to compare with statistical estimates of riverine fluxes (using the WRTDS model). NANI includes fertilizer, N fixation, atmospheric deposition and net food/feed inputs. The authors estimate fertilizer by developing regressions against rainfall for several land use categories, which explain about 55-72% of the variability. Atmospheric deposition is considered a small, spatially invariant value, which essentially has no explanatory power. N fixation estimates are based on crop specific estimates made in a previous study. Net food/feed imports were not included because the authors judged that the data were insufficient to estimate livestock feed, and human food imports were ignored, presumably because the human waste stream was considered insignificant (human wastewater sources were also dismissed as insignificant). Thus, the authors are assuming that the only sources of N in this region are due to fertilizer, as estimated by a rainfall proxy in different land use categories, and crop-dependent N fixation rates developed in an early study. The authors are aware of the limitations of the analysis, and explicitly state them in section 4.5. I think it would be useful to either develop proxies for net food/feed inputs (say, based on even rough estimates of livestock numbers and population density as they vary by catchment) or, if this is not possibly, simply frame the analysis as a relationship between rainfall-based estimates of fertilizer N loads and streamflow N, which is effectively what it is.
A few other comments follow below.
160: Given the seemingly strong relationship between elevation and rainfall shown in figure 1, it is surprising that elevation is nowhere mentioned as an explanatory variable, nor included in any tables. Was it investigated?
183: Need to cite a reference for QGIS
239: Adams et al 2014 considered only wet deposition of N and did not distinguish between reduced and oxidized forms. Reduced N (including NH3, NH4+, etc) are well known to be associated with volatile losses of NH3 from agricultural sources including manure from dairy herds and fertilizer applications, which are likely relevant here. (The authors note the significance of manure in the context of the discussion of NOx vs TN in agricultural vs forested land covers around line 488.) While the overall term is likely relatively small compared to other sources, using a constant value across the landscape cannot possibly be meaningful to estimate spatially varying loads in the models developed here. Either do a better job of estimating the true atmospheric deposition component and its spatial variation, or drop it as insignificant.
507: what is “fertiliser additive land use”?
588: Schaefer et al (2009) actually show lower export as a percent of N inputs in some western US watersheds than Schaefer and Alber (2007) show in southeastern US watersheds, presumably because of the relatively dry environments in much of the west which affect N delivery (as you note in section 4.3).
Section 4.3: A potential difficulty with the discussion of TN export as a fraction in inputs is that the estimated total catchment N inputs may be biased because of incomplete estimates. Some discussion of this specific issue is warranted.