the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Dec 2024
Nitrate reduction in groundwater as an overlooked source of agricultural CO2 emissions
Abstract. Nitrate pollution from agriculture poses a global environmental and public health threat. Nitrate levels in water can be reduced through denitrification, which increases dissolved inorganic carbon (DIC) via organic carbon mineralization and/or carbonate dissolution. This DIC potentially acts as a net anthropogenic source to atmospheric CO2; however, its overall impact remains unclear. This study quantified CO2 production from denitrification in Denmark, utilizing extensive observational datasets and national-scale modelling tools. We identified dominant denitrification processes in groundwater and predicted a national process map. Our results indicate that hydrogeology plays a central role in determining the dominant processes. CO2 production from denitrification in groundwater varied spatially, depending on nitrogen leaching and the denitrification processes. We estimated that denitrification in groundwater produces about 204 kt of CO2-eq. yr-1 as DIC, and ~50 % would be emitted to atmosphere. The Intergovernmental Panel on Climate Change (IPCC) guidelines account CO2 emissions related to agriculture from liming, urea, and other C-containing fertilizers, and these were 250, 1 and 4 kt of CO2-eq. yr-1, respectively, for Denmark in 2020. Although CO2 is a minor agricultural GHG emission (2 % of the total), our findings suggest that the agricultural GHG inventories should include denitrification-related CO2 emissions.
- Preprint
(3605 KB) - Metadata XML
-
Supplement
(1265 KB) - BibTeX
- EndNote
Status: open (until 20 Jan 2025)
-
RC1: 'Comment on egusphere-2024-3706', Anonymous Referee #1, 16 Dec 2024
reply
General comments
- This manuscript represents in my opinion a very useful contribution to the multi-faceted research area of denitrification in groundwater systems and will be of substantial interest to a wide audience.
- While denitrification-related CO2 emissions are featured in the title, the most valuable contributions to the state of the art might possibly lie elsewhere. The (redox) clustering done based on groundwater data from more than 6,000 wells, the cluster interpretation with regard to likely electron donors, and the linkage between clusters and landscape elements, might prove more valuable overall than the estimation of CO2
- Complete denitrification to N2 is assumed in all calculations presented in this manuscript. The topic of indirect N2O emissions that could result from incomplete denitrification in groundwater systems (additionally to nitrification) is not mentioned at all. While admittedly not the focus of this study, given the potency of N2O as GHG, I would like to suggest inserting a brief justification why complete denitrification was assumed, and references to a few studies on indirect N2O emissions (e.g. by Clough, Weymann, Jahangir, Jurado).
- 2a suggests that Cluster 3 is dominant in most of DK, followed by Cluster 7 in the areas not covered by ice sheets during the last glaciation. Pyrite has been identified as the key electron donor in both of these clusters, while organic carbon only appears to serve this role in clusters 1 and 6 (with minor spatial extent), while no clear dominance was evident in clusters 4 and 8. Oxic conditions (Clusters 2 and 5) seem to have insignificant spatial coverage. Given the importance of these findings, I would suggest to explicitly provide information on the spatial extent (km2) of each cluster, the area of pyrite-driven vs. organic carbon-driven denitrification, and references to any field research on electron donors that may underpin these results.
- The results suggest that substantial nitrate reduction occurs in most groundwater systems in DK (Fig. 3a). Nevertheless, CO2 emissions attributable to denitrification were estimated to add a maximum of 0.9% to the total emitted CO2 equivalents (see below). While DK has excellent availability of relevant data and scientific expertise, most other countries utilising the IPPC scheme will be less well equipped (and often will have smaller fractions of reduced groundwater). Accordingly, I suggest that most countries are not in a position to credibly estimate what might be a very small contribution relative to all other processes contributing GHG emissions in agricultural landscapes (please see below for detail). I would like to suggest that resources might be more usefully employed in combatting GHG emissions, rather than in adding small new components to the IPCC accounting system. Please consider these points when revising your Conclusions.
- The Specific comments listed below are largely of a minor or technical nature, but addressing them should improve the clarity of the manuscript.
Specific comments
15: what is meant by ‘dominant denitrification processes’? Different electron donors driving denitrification?
Table 1: Equations 1 and 2 are both assuming complete denitrification to N2. Could you please add a sentence on the effect incomplete denitrification would have.
65 ff: The calculations marked by * and ** in Table 1 are valid for situations where calcite saturation occurs. Could you please provide the reader with information on how common such conditions are within the groundwater system and where groundwater discharges into surface water bodies? Could it be argued that the CO2 emissions estimates represent an upper limit?
68: ‘triggered by anthropogenic nitrate input’. Not all N in groundwater originates from fertiliser application. Is the fraction of the denitrified N that might have come from natural sources considered negligible?
92: ‘map of denitrification processes’ seems a misnomer. The map is showing the distribution of six clusters with reduced groundwater redox chemistry.
93: Please make sure you clearly define in Section 3.4 what exactly you mean by ‘agriculture GHG inventory’.
106: Given that at least five measurements were required over the entire period (1890-2022), can you please provide the reader with a summary statement from which period most of the used data originate (e.g. 80% of the data were collected between 2001 and 2022)? Can we assume that the analysis is not affected by concentration trends during this period?
Sections 2.2 and 2.3: I would like to disclose that my understanding of ML techniques is very limited, and therefore cannot evaluate the choice of methods. Another reviewer may be able to fill this gap.
107: Numbers are reported for ‘screens’ rather than bore/well sites. Does this account for multiple screens possibly being located at different depths at one site?
112: ‘The cleaned dataset was analyzed to categorize redox conditions and to identify dominant processes by combining two machine learning techniques’. 1) Does ‘dominant processes’ refer to nitrate reduction processes (e.g. driven by pyrite vs organic carbon)?; 2) Before embarking on ML techniques, have you tried to characterise the redox conditions using the ‘classical’ framework by McMahon & Chapelle (2008)?
134: The oxic clusters 2 and 5 are shown in Fig. s2, not Fig. s1 as stated.
158: ‘the redox interface’ is defined as ‘the bottom of the nitrate-reducing zone’. Maybe specify ‘the first redox interface’, as Koch et al. (2024) makes it clear that more complex vertical stratification occurs widespread in DK.
169/170: Could you please provide the absolute number or percentage of screens excluded?
172: ‘depths of groundwater screens shallower than the depths of redox interface minus 5 (D5), 10 (D10), and 15 (D15) meters’ is unclear; please reformulate this explanation. The caption to Fig. s1 suggests that e.g. D5 stands for wells with ‘screen tops deeper than 5 (…) meters below the redox interface’. However, the corresponding well numbers given for D5 (235), D10 (566), and D15 (1019) seem to contradict this information. Does D5 stand for all wells where the screen is a maximum of 5m below the redox interface?
174: First time ‘wells’ is used rather than ‘screens’. Maybe consider using one term throughout the manuscript or clarify why different terms are used if there is a reason for it.
187: The 1990-2010 period was used for nitrate reduction estimates. Were the measurements from the 6,273 screens (line 108) also predominantly from this period?
208 ff: Could you please provide the number of wells in each of the eight identified clusters.
Would it be useful to apply the USGS redox classification scheme to the wells in these clusters? Also, could the clusters interpreted as reflecting heterotrophic denitrification be grouped (and presented) according to the redox sequence (weakly to strongly reduced: 2,5<4<8<1<6)?
Please also consider if the key cluster information provided in Sections 3.1 and 3.2 could usefully be presented in a Table? This would facilitate direct comparison between clusters, both concerning their chemistry and spatial distribution.
Given the variability in the data within a cluster (e.g. Fig. 1c), could some variability be interpreted as indicating that nitrate may have been reduced along its flowpath to the well screen by a combination of heterotrophic and autotrophic denitrification?
270 ff: It would seem useful to start here with info on the spatial extent (km2 or % of DK area) of the clusters, as Cluster 3 appears to be dominant, followed by Cluster 7, and all others well behind. Accordingly, pyrite would appear to provide much more widespread denitrification potential in DK than organic carbon.
273: Final ‘maps of denitrification processes’ (Fig. 2a). I find the use of the term ‘process’ somewhat misleading. As I understand it, Fig. 2a represents a spatial prediction of groundwater chemistry clusters. As outlined in Section 3.1, these clusters are thought to reflect the prevalence of one or more of the reactions listed in Table 1. Accordingly, I would suggest replacing ‘denitrification processes’ with ‘denitrification clusters’ or even wider ‘redox clusters’ (as denitrification reactions are only a subset of the reactions defining the clusters).
Fig. 2a: Maybe move the label ‘Main stationary line’ out of the black square that indicates the enlarged area, to make it clear that it refers to the somewhat inconspicuous dotted line, not the more prominent square. I also wonder, how to better present the less prominent clusters? Maybe colours could be swapped between Clusters 4 and 6, so that Cluster 4 areas in the still fairly small enlargement can be more easily recognised? Making Cluster 4 more prominent would also help with the discussion of Fig. 3 (highest DIC production in northern Jutland).
280/81: Maybe replace ‘outside’ with ‘west and south’ and ‘behind’ with ‘east and north’?
315ff: Please either add ‘Jutland’ and ‘Zealand’ labels on the map or provide more location info in the text (e.g. in the west of DK).
324 ff: I would suggest emphasising more that the spatial patterns of nitrate reduction and DIC production differ substantially, as the electron donors fuelling denitrification differ spatially.
Section 3.4: Notwithstanding the GHG contributions by LULUCF, the ‘agricultural contributions’ in the narrow sense comprise CH4, N2O, and CO2 from liming, urea, and other fertilisers. It would seem to me that the 90-104 kt estimated below almost pale into insignificance relative to the total GHG emissions attributed to ‘agriculture’ (amounting to 11,268 kt CO2-eq. yr-1, see Fig. 4).
338: I’m unsure if ‘excluding’ is the right word here? Would ‘after’ be more suited?
375: If I understand the numbers correctly, the upper limit of 104 kt (Fig. 4) would result in an increase of CO2 equivalents of 0.9%; the CO2 contribution to GHG emissions rising from 2.3 to 3.1% (358 out of 11,372 kt). While acknowledging that substantially smaller contributions are accounted for in the IPCC guidelines, these are more easily quantifiable (e.g. from fertiliser sales statistics). I am unconvinced that estimating CO2 resulting from denitrification could be added to the IPPC procedure in a credible manner. DK may be in the enviable position of being a virtual laboratory, but even under the favourable Danish conditions the estimates rely on a number of assumptions which introduce uncertainty. Estimates for most other countries around the world would inevitably be markedly less certain than the results presented here.
405: I find the 38% number for ‘agricultural emissions’ misleading, as the 254 kt calculation basis refers in the IPPC system only to the minor contributions made by liming, urea and other fertilisers (254 kt), rather than the total of 11,268 kt CO2-eq. attributed to agriculture (incl. 5132 kt arising from N2O and the 5881 kt from CH4, Fig. 4).
Citation: https://doi.org/10.5194/egusphere-2024-3706-RC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 58 | 14 | 4 | 76 | 13 | 1 | 0 |
- HTML: 58
- PDF: 14
- XML: 4
- Total: 76
- Supplement: 13
- BibTeX: 1
- EndNote: 0
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1