the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Jun 2026
Eddy covariance measurements of nitrogen dioxide exchange at a grazed savanna grassland in South Africa
Abstract. South Africa is a global hotspot for anthropogenic atmospheric NO2, where emissions from the industrialised Mpumalanga Highveld influence air quality across the southern African region. In the atmosphere, NO2 is a key player in oxidative chemistry and contributes to particulate nitrate formation and the biogeochemical nitrogen cycle through deposition. At the same time, various ecosystem processes act as sources and sinks of NO and NO2. The net result of these factors implies differences in the ecosystem scale flux of NO2. Here we perform the first-ever high-resolution NO2 measurements with a quantum cascade laser (QCL) instrument in a grazed African savannah landscape from 2015 to 2020. Micrometeorological eddy covariance measurements were used to quantify the NO2 flux and explore temporal trends at diurnal, monthly, seasonal and interannual scales. Our findings highlight the variability of NO2 flux within this system, with notable interannual change observed at both monthly and hourly scales. Seasonal differences in NO2 flux and deposition velocity were strongly linked to the rainfall season, with negligible differences between dry season months. Diurnal flux trends peaked during daylight hours, with consistently low NO2 flux during nighttime. These findings contribute to our understanding of near-surface atmospheric NO2 dynamics in arid landscapes and, for the first time, can be used to estimate ecosystem-scale compensation point of NO2.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Biogeosciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(3388 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 15 Aug 2026)
-
RC1: 'Comment on egusphere-2026-3066', Anonymous Referee #1, 18 Jul 2026
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-3066/egusphere-2026-3066-RC1-supplement.pdfReplyCitation: https://doi.org/
10.5194/egusphere-2026-3066-RC1 -
RC2: 'Comment on egusphere-2026-3066', Anonymous Referee #2, 30 Jul 2026
reply
General comments
This manuscript presents results from five years of high‑time‑resolution EC measurements of NO₂ fluxes over a grazed savanna grassland. These experiments, and the substantial maintenance they required, represent a major investment of time and effort and have yielded an extensive and valuable dataset. A key finding is the temporal variability in NO₂ fluxes across seasons and years. The authors report predominantly positive (emission) fluxes of NO₂ at this site, particularly during the wet summer period. The flux measurements are further used to derive an ecosystem‑scale compensation point for the grazed savanna grassland. In addition, for periods of net deposition, deposition velocities are quantified and compared with values obtained from modelling.
Overall, the measurements of NO₂ fluxes and the associated supporting data (e.g. temperature, radiation, soil moisture/temperature) appear comprehensive and robust. However, to further strengthen the manuscript, I would recommend clarifying certain aspects of the methodological description and expanding the interpretation and discussion of the results. While the dataset is strong, the scientific significance would benefit from more comprehensive analyses and a better integration of the various data components.
Specific comments
Major comments
- An uncertainty analysis is missing. Please discuss the random and systematic uncertainties of your measurements, fluxes, etc.
- Throughout the results and discussion section (Sec. 3), many interpretations are rather speculative. For example statements referring to correlations with soil moisture, rainfall, temperature, solar radiation or turbulence to explain observed fluxes, but without plotting any correlations at all (like in line 276-278, 281-282, 346-347, 349-351, 364, 438-439 etc.). Apparently you have the data (Fig. 3 and 4), so it would be interesting and more convincing to show the correlations.
- Soils emit NO, which is then converted into NO2 in the presence of O3. Therefore, usually fluxes of all three compounds are measured. Please explain why it is sufficient in your case to only measure NO2. How large is the influence of the chemistry between NO, O3 and NO2 on the measured NO2 fluxes? And how different is the NO2 flux you measure at the measurement height from the actual soil/vegetation NO2 emissions? Which part of the variability you observe could be explained by chemical reactions?
- Your results show predominantly emission fluxes, please explain better the implications of the analysis of deposition velocities you do. How relevant are conclusions based on deposition data for a site which acts as a source of NO2?
Smaller comments
- Line 33-34: the definition of dry deposition is not “the gravitational settling of particles”!
- 1: what is the figure on the right? Satellite data? Modelled data? Please note the source?
- Line 103: how do we see in Fig. 1 that WAMS is located at border grassland and savanna?
- Line 140: describe shortly your method for stationarity filtering.
- Line 242: these are quite extreme flux values. Are they physically possible or should they be filtered out? And why do we not see them in Fig. 3h?
- Line 296-298: how do you know these are not just outliers? For example instrument artefacts? These values are so extreme compared to the average fluxes.
- Line 337: do you mean venting NO or NO2 from the soil?
- Line 360: which patterns do you mean with “these patterns”?
- Line 375: Butterbach-Bahl is not about NO2 but N2O as far as I can see?
- Line 377: where do we see those abrupt daytime reversals and significant pre-dawn minima? Refer to figure and explain how we see that in the figure.
- Line 412: this is not partitioning!! This is just separating positive and negative net fluxes, so use a different name/description than partitioning.
- Line 440-448: why do you think these data exceeding the 99th percentile are extreme flux events and not just outliers due to for example measurement artefacts?
- Line 440-448: what was the meteo like in 2015? And did you check other correlations than ambient NO2 concentration?
- Line 453: you write that both positive and negative NO2 fluxes need to be considered independently, but next you only discuss the deposition fluxes and deposition velocities, even though the site is a net source of NO2. Why not more analysis of the emissions?
- Line 460: which recurring monthly patterns do you mean?
- Line 463-464: aren’t these extreme values outliers (as you call them in line 462)? Why calculate vd and compare with modelled values for data including outliers?
- Line 465: if May consists of so few measurements and is “not an accurate representation of the month”, shouldn’t you then exclude it from the analysis?
- Line 466-467: please substantiate this with data.
- Line 478: is this including or excluding the high-magnitude outliers? (see line 461-462)
- Line 494: where do we see the means? In Fig. 10 (and Fig. 9 as well) I only see medians.
- Line 497-498: please show that the May spike coincides with boundary layer height if you make that statement.
- Line 502: in which figure do we see that? Is this figure missing?
- Line 515-522: this is quite a long description of what one can also see in the figure, superfluous. And I miss an interpretation of what it means what we see.
- Line 535-537: please explain why you exclude certain conditions from the analysis.
- Line 539: does it make sense that cp for wet and dry season are so similar, considering the difference between them in Fig. 11?
- Line 546: explain why you compare with trees and not grass.
- Line 550-551: what is the difference between this vd and the one discussed before (like in Fig. 9 and 10)? Is that because of the filtering? Would we expect them to be so different and what do both vd mean then?
Technical issues
- 2a: flux is not in ppb, check units
- 2b: list the relevant results from the linear regression (like slope, offset, R2)
- 3: also plot NO2 concentration?
- Line 251: reference to Fig. 7 should be Fig. 5?
- 5: solid line indicating y=0 helps to guide the eye what are positive and negative fluxes. Also, the boxes seem rather narrow, for visibility it would be nice if they are broader.
- 6: confusing that flux axis label sometimes 0.03 and sometimes 20 x 10-3. Again, plot the y=0 line for fluxes. Also, it would help to indicate which months are summer/winter period (for the northern hemisphere readers). And why is the scale so different? (I can imagine that a diurnal trend would be visible in April as well if the flux scale was similar to that of September)
- 7: plot y=0 line.
- Line 375: Ganzeveld et al. 2002 is not in your list of references
- 8: plot y=0 line, indicate which months are which season (summer/winter)
- Line 460: Figure 9b instead of Figure 9?
- Figure 9: for visibility please make the boxes bigger. It would help to indicate which months are summer/winter.
- Line 471: these are not “partitioned NO2 fluxes”.
- Line 494: Figure 12 must be Figure 10 or Figure 9?
- Figure 11: explain the shaded region around the fit.
- Figure 12: legend is missing. And in Fig. 12a the title is “… 2015 & 2016” and in Fig. 12b it is “… 2015-2016”
Citation: https://doi.org/10.5194/egusphere-2026-3066-RC2 -
RC3: 'Comment on egusphere-2026-3066', Anonymous Referee #3, 31 Jul 2026
reply
1. Overall assessment
This manuscript presents a rare and potentially important multi-year dataset of ecosystem-scale nitrogen dioxide (NO₂) exchange over an African savanna-grassland system. The geographical setting, the use of a fast-response quantum cascade laser combined with eddy covariance, and the comparison with an inferential dry-deposition model are relevant to the scope of Biogeosciences. Direct flux measurements from African semi-arid ecosystems remain scarce, and the study could make a valuable contribution to regional nitrogen-cycle research and to the evaluation of deposition models.
The manuscript nevertheless has major methodological and interpretative weaknesses in its present form. The eddy-covariance processing chain is insufficiently documented to verify the validity and reproducibility of the NO₂ flux product. Important information is missing on time-lag determination, spectral attenuation, tube losses, coordinate rotation, detrending, dilution or density corrections, random uncertainty, detection limits, storage, advection, and chemical flux divergence. The central analyses are also affected by large and non-random data gaps, inconsistent record counts, uncertain treatment of extreme values, and conceptual problems in the calculation and interpretation of deposition velocity.
The paper has high potential, but publication should be considered only after a substantial reprocessing and reanalysis. The authors should first demonstrate the technical validity and uncertainty of the flux measurements, then reconstruct the temporal analyses using explicit coverage criteria, reformulate the observed-model comparison on a like-for-like basis, and moderate causal interpretations that are not directly tested. If the required high-frequency diagnostics and processing information are unavailable, the main quantitative conclusions cannot be independently verified.
2. Major observations
- Eddy-covariance processing and flux validity. The complete NO₂ eddy-covariance processing chain must be described in a reproducible sequence. The authors should specify the software or code used and document despiking, detrending, coordinate rotation, time synchronisation, time-lag optimisation, covariance calculation, stationarity and turbulence tests, density or dilution corrections, and conversion to the reported mass flux. For the 20 m closed-path inlet, the manuscript must report residence time, Reynolds number, analyser response time, inlet heating, pressure control, filter maintenance, wall-loss tests, cospectra or ogives, high-frequency attenuation corrections, instrument calibration, precision, drift, water-vapour interference, detection limits, and random flux uncertainty.
- Reactive chemistry, storage, and advection. A positive NO₂ covariance at the measurement height represents apparent ecosystem-scale net NO₂ exchange; it does not directly prove biological NO₂ emission by soil or vegetation. The interpretation must account for NO-NO₂-O₃ chemistry, photolysis, turbulent transport, entrainment, storage below the measurement height, and horizontal or vertical advection. Available NO, O₃, radiation, turbulence, wind-sector, fire, plume, and boundary-layer information should be analysed. Biological mechanisms should be presented as hypotheses unless they are quantitatively demonstrated.
- Data accounting and temporal representativeness. The difference between 34,239 final valid half-hourly records and the 21,681 observations used in the analysis must be explained. The authors should provide a transparent data-flow diagram and observation counts by year, month, season, hour, flux sign, and quality-control step. Minimum coverage criteria are required for monthly, seasonal, annual, and diurnal statistics. Periods with inadequate coverage should be excluded from interpretation, and interannual 'trends' should be described as differences unless a formal trend analysis is performed.
- Flux-sign classification and extreme events. The classification of every positive flux as emission and every negative flux as deposition is not defensible without a measurement-uncertainty threshold. The authors should define a minimum detectable flux or signal-to-noise criterion and quantify how many observations are statistically distinguishable from zero. Extreme fluxes and deposition-velocity values must be checked against raw spectra, lag peaks, analyser diagnostics, stationarity, concentration, precipitation, wind direction, and neighbouring records. Unverified extremes should be removed from the scientific interpretation.
- Definition and calculation of deposition velocity. The exact equation, sign convention, concentration basis, concentration threshold, and uncertainty propagation used to calculate deposition velocity must be provided. A ratio calculated only for downward fluxes is a sign-censored conditional quantity and is not directly equivalent to the deposition velocity of a unidirectional resistance model. Because the manuscript also invokes a non-zero compensation point, the authors should reconsider whether half-hourly observed deposition velocities are physically meaningful for this bidirectional and chemically reactive exchange system.
- Observed-model comparison. The inferential model must be fully described, including code or version, equations, resistance parameterisations, meteorological forcing, temporal resolution, land-cover properties, canopy height, LAI or NDVI treatment, and surface wetness assumptions. Model estimates and measurements should be compared at identical timestamps. The post hoc exclusion of values above the 90th percentile should be avoided or rigorously justified. Robust metrics such as median bias, mean bias, MAE, RMSE, correlation, and normalised bias should be reported with confidence intervals that account for temporal dependence.
- Compensation-point analysis. The ecosystem compensation-point analysis requires stronger statistical support. The authors should report the number of observations per concentration bin, regression weighting, slope significance, confidence-interval method, and sensitivity to bin width and environmental filters. Temporally coherent block bootstrapping is recommended. The result should be called the x-intercept of the fitted relationship, not a zero-intercept. Given the overlapping wet- and dry-season confidence intervals, seasonal differentiation should not be claimed unless it remains robust after sensitivity testing and exclusion of extreme events.
- Rainfall analysis and causal interpretation. Rainfall totals calculated only during valid flux intervals are incomplete and must not be presented as annual or hydrological-year totals. The complete continuous precipitation record should be used independently of flux data. An event-based analysis should compare verified rainfall events with pre-event and post-event fluxes using explicit time windows and adequate sample sizes. Claims that rainfall directly caused NO₂ pulses, including a Birch-effect interpretation, should be replaced by cautious associations unless supported by this analysis.
- Footprint heterogeneity and external influences. A fixed footprint radius is insufficient for a heterogeneous site. Dynamic footprints should be calculated or reconstructed for retained observations and summarised by stability, season, and wind sector. The influence of cropland, the access road, livestock grazing, fire, local activities, and regional plumes should be screened or analysed separately. Grazing intensity and timing should be reported. The back-trajectory analysis also requires a complete methods description or should be simplified and moved to the Supplement.
- Statistical methods and uncertainty framework. A dedicated statistical-methods subsection is required. It should identify each test, null hypothesis, sample unit, effective sample size, treatment of autocorrelation, multiple-testing correction, effect size, and confidence interval. Because the distributions are strongly skewed and temporally dependent, robust or non-parametric methods and block bootstrapping are preferable. Uncertainty should be reported around all central estimates.
- Internal consistency and causal overstatement. The abstract, Results, Discussion, and Conclusions contain contradictions concerning seasonal differences, wet- and dry-season definitions, means versus medians, figure titles, cross-references, and numerical values. These inconsistencies must be systematically corrected. Causal expressions such as 'confirms', 'direct consequence', 'thermal venting', 'mechanically driven', and 'fundamental shift' should be used only for mechanisms that are explicitly tested.
- Data and code availability. The statement that data are available upon request is insufficient for a study based on a unique long-term dataset and a complex processing chain. A FAIR-aligned repository with a persistent identifier should contain the half-hourly fluxes before and after quality control, concentration and meteorological data, quality flags, lag and spectral diagnostics, uncertainty estimates, footprint calculations, model inputs and outputs at matching timestamps, and scripts reproducing all analyses, tables, and figures. Any restrictions on raw high-frequency data should be clearly explained.
- Manuscript organisation and presentation. After reanalysis, the manuscript should be shortened and reorganised. Repetition between monthly, seasonal, and diurnal descriptions should be reduced; observations should be clearly separated from interpretation; common axis scales should be used where comparisons are intended; sample sizes should be shown; and secondary full-period or monthly diagnostics should be transferred to the Supplement. The main text should focus on a limited number of robust results.
3. Minor observations
- Use the spelling “savanna” consistently, clarify whether the site is a savanna–grassland ecotone, and use “semi-arid” rather than alternating with “arid”.
- Correct the definition of gaseous dry deposition. It should refer to non-precipitation transfer to surfaces through turbulent transport, molecular diffusion, and surface uptake, rather than gravitational settling of particles.
- Distinguish NO₂ from the broader term NOₓ throughout the Introduction and Discussion because NO and NO₂ have different chemistry and surface-exchange behaviour.
- State the study objectives and hypotheses explicitly at the end of the Introduction and avoid over-emphasising the compensation point relative to the other analyses.
- List the supporting measurements actually used, improve the readability of Figure 1, describe the back-trajectory method, and consider adding a site photograph and land-cover map.
- Present the four-season and wet/dry-season classifications together and apply the same definitions consistently throughout the manuscript.
- Provide the exact QCL manufacturer and model, optical and sampling-cell specifications, pressure, path length, response time, detection limit, calibration gases, and measurement uncertainty. Complete the unfinished sentence describing the climate-controlled enclosure.
- Use consistent SI formatting, including m s⁻¹, and explain the choice of the fixed friction-velocity threshold of 0.2 m s⁻¹ with a sensitivity analysis.
- Define every symbol and unit in the deposition-velocity equation, distinguish ppb from mass concentration, and state whether the flux represents mass of NO₂ or mass of nitrogen.
- Include the Appendix referred to for ancillary measurements or remove the corresponding reference.
- Explain why the post-processing and final-valid-record rows in Table 1 are identical and report yearly recovery rates.
- Correct Figure 2: flux cannot be expressed in ppb; clarify the H₂O variable, negative values, axes, units, regression equation, intercept, and outlier treatment.
- Expand the inferential-model methods to identify meteorological datasets, temporal aggregation, MODIS product and version, spatial resolution, quality filtering, LAI processing, vegetation height, and land-use parameters.
- Do not call the flux-synchronised meteorological subset a climatology. Report the correlation method, sample size, and treatment of autocorrelation.
- Improve Figures 3 and 4 by using readable axes, the complete precipitation record, appropriate labels, and a representative event period in the main text while moving the full time series to the Supplement.
- Reconcile N = 21,681 with Table 1, report uncertainty on the mean and median net flux, avoid declaring a net source until sampling bias is addressed, and exclude months with extremely small sample sizes such as N = 12.
- Correct the dry-season deposition percentage and avoid annual means or strong interannual interpretation for years with very low or seasonally incomplete coverage.
- State consistently whether diurnal curves show means or medians, use comparable y-axis ranges, identify local time or UTC, and remove or reframe speculative claims about thermal expansion or advective expulsion of soil gases.
- Revise the comparison with Brümmer et al. because that study concerned total reactive nitrogen exchange, define NEE at first use, and use “separated by sign” rather than “partitioned”.
- Verify all flux ranges and deposition-velocity values, correct the likely decimal error in the 90th-percentile threshold, and replace the contradictory statement “not statistically significant (p < 0.05)” with the correct test result and exact p-value.
- Correct Figure 10, use “x-intercept” for the compensation point, fix typographical errors, remove the reference to the non-existent Section 3.3.3, and correct the period shown in Figure 12 while adding sample sizes for each bin.
- Rewrite the Conclusions after reanalysis, explicitly acknowledge fragmented coverage and measurement uncertainty, replace “available upon request” with a repository statement, disambiguate repeated author initials, and standardise references, author names, DOIs, subscripts, symbols, dashes, and English usage.
Final recommendation. Major revisions. The manuscript addresses an important scientific data gap and retains high publication potential. Reconsideration should depend on a demonstrated validation of the eddy-covariance flux product and a complete reanalysis of the source/sink behaviour, deposition velocity, compensation point, rainfall relationships, and model comparison.
-
RC4: 'Comment on egusphere-2026-3066', Anonymous Referee #4, 05 Aug 2026
reply
This manuscript describes an incredibly impressive long term record of NO2 fluxes. There are many interesting aspects and many misconceptions in the paper. It is difficult to capture all of them in a review.
My first recommendation would be to separate the observations and analysis into two papers. This would allow a reader to focus appropriately on the two different complex threads the authors are discussing.
Observations
1) It is no longer adequate for data to be made available on request. The data should be posted to a public open repository.
2) The first map -- figure 1-- describes a scale appropriate to advection, but not to fluxes. The figure should show the region within the flux footprint of the measurements. I('m not sure the height of the inlet is mentioned in the paper. It should be noted in the first paragraph of section 2.2.
3) I would like to see cospectra of the vertical wind and NO2 to affirm that the system is working well.
4) much of the analysis is limited by not having simultaneous NO and O3 measurements. Whatever the authors could describe about those in the region would be helpful. For the most part, we expect that what the authors are observing is NO emissions followed by reaction with o3 and then NO2 is the indicator of NO flux. Is the NO/NO2/O3 system at steady state by the height of the inlet or could variations in the inferred fluxes be because of differences in the conversion efficiency of NO to NO2 at the inlet height at different sampling times. Even trickier, downward flux of O3 would result in a downward flux of NO2 even if there were no NO2 emissions.
Analysis
1) Everything about a compensation point should be removed. This data is not capable of identifying a compensation point. The idea of a compensation point is that at a low enough NO2 concentration there would be direct NO2 emissions from an ecosystem--usually thought of as direct stomatal emission from the plant mesophyll. The existence of a compensation point is controversial. Assessing the existence of a compensation point in the ambient environment without simultaneous NO and O3 flux measurements is, I think, impossible.
2) Deposition of NO2. Delaria and Cohen beginning with the paper referenced and in several subsequent papers (reviewed in doi.org/10.1021/acs.accounts.3c00090) show that NO2 is removed by stomatal uptake and that there are no direct emissions from a wide series of plants. Their work implies there is no compensation point. It also explains the ad hoc parameter widely used in global models--the canopy reduction factor--as a poor surrogate for stomatal uptake. That work would suggest fluxes are modified by uptake to stomata within a plant canopy before they emerge above. The extent of modification should be proportional to the extent of stomatal opening. Is this a useful framing for the deposition observed? If not why?
3) It'd be helpful if the authors made contact with the satellite remote sensing (e.g. OMI, TROPOMI) and modeling literature (e.g. GEOS-CHEM) describing NOx emissions in their analysis.
4) More generally, these data should be connected to some sort of a model for fluxes. Large scale models have temperature as a key parameter, do these fluxes vary with temperature? Is there a figure that could be made affirming that? The first rain effect is notable. Is there a good figure showing that effect?
5) The Butterbach-Bahl reference is an excellent one for N2O and generic N emission processes. It doesn't say much about NO/NO2. I'm not sure the places in the paper that reference that one are appropriate. It makes me wonder about the other references.
Figures 2b, 3, 5, 8 and 9 should be moved to a supplement or deleted.
Citation: https://doi.org/10.5194/egusphere-2026-3066-RC4
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 169 | 48 | 26 | 243 | 21 | 21 |
- HTML: 169
- PDF: 48
- XML: 26
- Total: 243
- BibTeX: 21
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1