the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Jul 2026
Land-use-specific soil NOx emissions in eastern China derived from TROPOMI observations
Abstract. Quantifying soil nitrogen oxide (NOx) emissions is essential for constraining the nitrogen cycle and understanding its impacts on atmospheric chemistry and regional air quality. This study presents the first land-use-specific estimate of soil NOx emissions across eastern China derived from TROPOMI satellite observations at a spatial resolution of 0.2° × 0.2°, covering forest, cropland, grassland/shrubland and bare land. We derive the emission intensity factor β for these land use types as functions of soil temperature (Ts) and soil volumetric water content (θv). Total soil NOx emissions from eastern China in 2019 are estimated to be 1305 ± 368 Gg N, with largest contributions from cropland (623 ± 267 Gg N, ~ 48 %) and forest (486 ± 124 Gg N, ~ 37 %). High-emission areas (> 4 kg N ha-1 yr-1) are concentrated in the North China Plain, and all land use types show summer-peak seasonal patterns. Soil NOx emissions exhibit land-use-specific responses to temperature and moisture. Most land use types show a clear optimum temperature and high-temperature inhibition effect, which are captured by a Gaussian response function. We find that existing global emission inventories may underestimate soil NOx emissions from forests in China. The comparison with other emission inventories and soil emission observations demonstrates that the proposed method provides a reliable basis for estimating soil NOx emissions by capturing land-use-specific responses to environmental drivers. This satellite-based approach can provide input to the formulation of emission reduction strategies and support ecosystem management under climate change.
- Preprint
(2884 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3440', Anonymous Referee #1, 10 Sep 2026
-
RC2: 'Comment on egusphere-2026-3440', Anonymous Referee #2, 12 Oct 2026
Review of “Land-use-specific soil NOx emissions in eastern China derived from TROPOMI observations”
General Comments
Zhao et al. present a satellite-based framework for estimating land-use-specific soil NOx emissions over eastern China by integrating top-down total NOx emissions (derived from DECSO algorithm), anthropogenic emission inventories (ABaCAS-EI and CAMS-GLOB-ANT), land-use information (CGLS-LC100m), and meteorological data (ERA5-Land). The authors further derive land-use-specific relationships between soil NOx emission intensity and soil temperature under different soil moisture conditions. Overall, the manuscript is well organized, and the proposed approach is interesting. However, several aspects of the methodology and interpretation require further clarification and evaluation: 1. the attribution of residual emissions to soil sources needs to be more carefully justified; 2. the identifiability and stability of the land-use-specific regression coefficients require further evaluation; 3. several methodological choices, thresholds, uncertainty assumptions, and mechanistic interpretations require stronger support from the literature or quantitative analysis.
Major Comments
1. Attribution of residual emissions to soil NOx
The interpretation of the residual emissions as soil NOx requires a more complete source-closure discussion. DECSO represents total surface NOx emissions from both anthropogenic and natural sources, whereas the attribution procedure appears to interpret the residual after subtracting anthropogenic emissions predominantly as soil emissions. Although the manuscript states that lightning and free-tropospheric NOx contributions are excluded from DECSO, it does not sufficiently explain how these contributions are removed. The authors should also clarify how other surface natural sources, such as biomass burning, are treated in DECSO and the anthropogenic emission inventory. This is particularly important because these sources may exhibit substantial spatial and seasonal variability and could therefore be incorrectly attributed to soil emissions from specific land-use types.
In addition, the manuscript itself suggests that the anomalously high β values in winter may result from heating-related anthropogenic emissions that are not fully captured by the anthropogenic inventory. This indicates that systematic biases in the anthropogenic inventory may remain in the residual and subsequently be attributed to soil emissions. The authors should therefore provide a more complete source-closure description of the attribution framework and discuss the potential influence of insufficiently constrained natural sources and residual anthropogenic emissions on the inferred soil NOx emissions and β values.
2. Insufficient justification and supporting evidence for methodological choices and interpretations
Several methodological choices and interpretations throughout the manuscript appear to lack sufficient support from either the literature or quantitative analysis. These include the selection of key thresholds used for classifying soil moisture conditions and screening grid cells (e.g., LUR > 0.05 to remove water coverage), the uncertainty values assigned to different input variables, and some mechanistic interpretations of the derived spatial patterns and temperature-response relationships. The authors are encouraged to provide stronger literature support, quantitative analyses, or sensitivity tests where appropriate. Specific issues are detailed in the comments below.
3. Robustness of the land-use-specific emission factors
The robustness of the land-use-specific emission factors (β) derived from Eq. (1) needs further evaluation. Because multiple land-use fractions are simultaneously included as predictors, correlations among these fractions may affect the stability of the individual β estimates. Please provide the distributions and appropriate regression diagnostics to demonstrate that the emission contributions of different land-use types can be reliably separated.
In addition, the manuscript removes a land-use type from Eq. (1) when fewer than 20 grid cells have LUR > 0.2 and considers that land-use type not to be a major emission source. However, an insufficient number of samples for estimating β does not necessarily indicate a negligible emission contribution. Removing a land-use term may also influence the β values estimated for the remaining land-use types. Please justify this treatment and evaluate its influence on the results. The observed LUR ranges should also be reported because β is defined for LUR = 1.
Minor Comments
L16: The term “first” should be used with caution.
L21 and Sect. 4.2: The criteria used to classify emission intensity as “high” (> 4 kg N ha-1 yr-1) and “relatively low” (< 1 kg N ha-1 yr-1) are unclear. Please clarify the basis for selecting these thresholds.
L25: The term “soil emission observations” may be ambiguous. These should instead be described as observation-based emission estimates.
L27: “This satellite-based approach” could be more explicitly described as “This satellite-based inversion approach.”
L31–32: This sentence could be removed, as the health impacts of NOx have already been stated and are not the focus of this study.
L44-45: for the definition of soil NOx emission, does it also include sources from the agricultural field?
L104–105: Please further clarify how the influence of lightning NOx emissions is excluded.
L109–110: Please further clarify how the 25% uncertainty is determined, and the contributions of the individual error sources to this uncertainty.
L114–115: Please clarify which version of ABaCAS-EI was used and provide the corresponding reference. The manuscript states that the anthropogenic emissions are for 2019, whereas the cited earlier ABaCAS-EI publications describe older inventory years. The updated ABaCAS-EI v2.0 (Li et al., 2023) provides annual emissions from 2005 to 2021, including 2019. Please ensure that the dataset version and reference are consistent with the inventory used in this study. Also provide the spatial resolution for this dataset.
Reference
Li, S., Wang, S., Wu, Q., Zhang, Y., Ouyang, D., Zheng, H., Han, L., Qiu, X., Wen, Y., Liu, M., Jiang, Y., Yin, D., Liu, K., Zhao, B., Zhang, S., Wu, Y., and Hao, J., Emission trends of air pollutants and CO2 in China from 2005 to 2021. Earth Syst Sci Data 2023, 15, 2279–2294.
L120: Since CAMS-GLOB-ANT is a composite emission inventory rather than an independently constructed dataset, please briefly introduce its underlying data sources.
L131: Please clarify that the 0–7 cm soil depth corresponds to the first soil layer defined in ERA5-Land.
L157-158: Figure 1: The flowchart could be further improved to more accurately reflect the methodology described in the text:
- the subtraction of anthropogenic NOx emissions from the DECSO total emissions before applying Eq. (1) is not explicitly shown. Therefore, the statement “Separate the contribution of each land use type from total emissions” may be misleading and should be revised accordingly.
- the meaning of “Urban LUR (Dynamic)” is unclear, and the additional DECSO-based criterion applied during the cold season is not shown in the flowchart. Please revise the figure to more completely reflect the season-dependent grid-cell selection criteria described in the text.
- The left and right part in this figure looks separate, I wonder how they are connected to each other, and what output you derived from each step (1-4)
- It is shown the “improved method”, could you illustrate the part where you improved relative to Lin et al. (2024)?
L165–167: The choice of 30% volumetric soil moisture as the threshold for separating dry and wet conditions is insufficiently justified. Please provide relevant literature support or further analysis, including the variation in grid-cell proportions under different thresholds, to justify this choice.
L180–183: The selection of 7.5 kg N ha-1 yr-1 as the DECSO emission threshold is insufficiently supported. Please provide relevant references or data analysis.
Sect. 3.3: Please provide more details on how the performance of the multiple linear regression was evaluated.
L205–208: Having fewer than 20 grid cells with LUR > 0.2 indicates insufficient samples for reliably estimating β, but does not necessarily imply that the corresponding land-use type is not a major emission source. Please revise or further justify this interpretation.
Sect.3.1: Please use Table to define your classification strategy and reduce reductant in this section. Also, justify why you set 30 % as dry-wet threshold and 2 ℃ for intervals.
Sect. 3.4: Please clarify why the median β value across different months is used as the representative value for each temperature interval and provide the number of β values available in each interval. In addition, please specify the criterion used to select the best-fitting function between the exponential and Gaussian models.
Sect. 3.5: Please provide literature references or supporting data analyses and figures to justify the uncertainty values assigned to each input variable.
Sect. 4.1: The mechanistic interpretations of the fitted relationships lack sufficient support from the literature or data analysis. Please provide additional evidence to support these interpretations. Also, if the results are grid-based, why only these several grids are shown in the figure.
L324–330: The interpretations of the spatial hotspots and summer emission peak in terms of agricultural activity, hydrothermal conditions, microbial activity, and fertilization lack sufficient support from references or quantitative analysis. Please provide additional evidence to substantiate these interpretations.
L372–377: CAMS-GLOB-SOIL and HEMCO-soil are both model/parameterization-based soil NOx emission products. Please briefly describe how these two products are derived, including their underlying parameterizations and major driving variables. From the spatial maps’ comparisons, we see the soil emission of agricultural land by this study is lower than HEMCO, while it is higher in forest and grassland. From the Figure 8 filed observations, the emission intensity may be overestimated. Does that mean the total soil Nox emission may also be overestimated?
Citation: https://doi.org/10.5194/egusphere-2026-3440-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 250 | 146 | 34 | 430 | 27 | 34 |
- HTML: 250
- PDF: 146
- XML: 34
- Total: 430
- BibTeX: 27
- EndNote: 34
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors presented a satellite-based analysis of land-use-specific soil NOx emissions in eastern China using DECSO NOx emissions, anthropogenic emission inventories, land-use information, and meteorological data. This is an interesting study and the approach has the potential to provide new constraints on soil NOx emissions from different land-use types. However, I have several major concerns regarding the attribution of the emissions. These issues need to be addressed before the quantitative results and conclusions can be considered robust.
Major comments:
1. The anthropogenic emission inventory plays a critical role in the derived soil NOx emissions. It is directly subtracted from DECSO emissions to estimate soil emissions. Anthropogenic inventories have substantial uncertainties in both their magnitude and spatial and temporal distributions, which is 35% said by the paper. Any bias in the inventory could therefore be directly attributed to soil emissions. This issue may be particularly important when soil emissions are small relative to anthropogenic emissions. The treatment of negative residuals as zero may introduce an additional positive bias. The authors should better evaluate how uncertainties and biases in the anthropogenic inventory affect the derived soil NOx emissions.
2. I have concerns about using land-use fractions to separate emissions from different soil sources. Land-use fractions represent source area, while important drivers such as fertilizer N input and irrigation are not included. The land-use fractions may be correlated, making it difficult to independently estimate the emission factors in the regression. In addition, the choice of LUR > 0.2 and at least 20 grid cells appears arbitrary. The authors should justify these assumptions and provide sensitivity tests and regression diagnostics.
3. I am not fully convinced that the fitted β–temperature relationships represent the response of soil NOx emissions to temperature. The β values in different temperature bins are derived from different locations and months, and they would reflect regional and seasonal differences in fertilizer application, vegetation, soil properties, or other factors. The authors should better demonstrate that the effects from these confounding factors. The sensitivity to the soil moisture threshold should also be evaluated.
4. The performance of the attribution framework is not sufficiently evaluated. The authors should provide model performance and diagnostics for Eq. (1) and evaluate how well the reconstructed emissions from Eq. (3) reproduce the satellite-derived emissions.
5. The final soil NOx estimates are capped by the DECSO total emissions when the calculated soil emissions exceed DECSO. It is unclear how this adjustment affects the final estimates.
6. The Monte Carlo analysis needs more explanation. The authors state that each input is randomly sampled within its uncertainty range, but the probability distributions used for the sampling are not specified. Several thresholds used in the method are fixed and their effects are not included in the uncertainty analysis. The authors should clarify the sampling procedure and evaluate the sensitivity of the results to the major methodological thresholds.