the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Global atmospheric inversion of the NH3 emissions over 2019–2022 using the LMDZ-INCA chemistry-transport model and the IASI NH3 observations
Abstract. Ammonia (NH3) emissions have continuously increased due to extensive fertilizer usage in agriculture and increasing production of manure and livestock. However, the current NH3 emission inventories exhibit large uncertainties at all the spatiotemporal scales. We provide atmospheric inversion estimates of the global NH3 emissions over 2019–2022 at 1.27°×2.5° horizontal and daily (at 10-day scale) resolution. We use IASI-ANNI-NH3-v4 satellite observations, simulations of NH3 concentrations with chemistry-transport model LMDZ-INCA, and finite difference mass-balance approach for inversions of global NH3 emissions. We take advantage of the averaging kernels provided in IASI-ANNI-NH3-v4 dataset, by applying them consistently to LMDZ-INCA NH3 simulations for comparison to the observations and then to invert emissions. The average global anthropogenic NH3 emissions over 2019–2022 is estimated as ~98 (95–101) Tg/yr, which is ~63 % (~57 %–68 %) higher than the prior CEDS inventory’s anthropogenic NH3 emissions and significantly higher than two other global inventories: CAMS’s anthropogenic NH3 emissions (by a factor of ~1.9) and CAMEO’s agricultural and natural soil NH3 emissions (by ~1.4 times). The global and regional budgets are mostly within the range of other inversion estimates. The analysis provides confidence in their seasonal variability and continental to regional scale budgets. Our analysis shows a ~4 % to ~33 % rise in NH3 emissions during COVID-19 lockdowns in 2020 across regions. However, this rise is probably due to a decrease in atmospheric NH3 sinks due to decline in NOx and SO2 emissions during the lockdowns.
- Preprint
(3933 KB) - Metadata XML
-
Supplement
(14782 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-162', Anonymous Referee #1, 22 Mar 2025
General comments
The paper describes the application of global ammonia (NH3) emission inversion estimates over 2019-2022. As the current top-down emissions of NH3 are rather inconsistent across spatiotemporal scales, this approach provides a new insight into the NH3 emission budget, at relatively high resolution and daily scales. The inversion uses an IASI averaging kernel (AK) to constrain the profile of NH3 concentrations; results are compared with two global inventories and two top-down estimates. The average estimate shows a higher value, compared to previous budgets, either globally or regionally. The emission results are used to analyze the impact of COVID-19 lockdowns in 2020, as compared with that in 2019. However, the rise of emissions in 2020 seems to be likely due to the decrease in atmospheric NH3 sinks (e.g., NOx and SO2) and induces large uncertainty to these emission estimates.
The inversion now uses IASI observations to constrain emissions, and other bottom-up inventories and top-down inversions for validation. Would it be possible to use simulated NH3 concentration based on updated emissions to compare with the IASI (or CrIS) concentration or in-situ observations? The current emission validation shows quite large differences between emission products, would the simulated concentration based on this emission estimate also give very different results with NH3 concentration observations? The consistency of this inversion method would be very necessary to check first.
Although the finite difference mass-balance (FDMB) inversion approach has been applied to update the anthropogenic NOx emission inventories, it has been rarely used in NH3 emission. Actually, the emission perturbation of 10-20 % is sometimes applied to get the scaling factor (beta). Could the authors further explain why a larger 40 % is applied in NH3 emission and distribution of the beta could be shown to clarify why it should be within the range of 0 to 10?
The spatial resolution of the model is 1.27° × 2.5°, which challenges the assumption that there is no transport in a grid, considering the normal wind speed of more than 100 km in a day and the lifetime of NH3 around a day. And the prior inventory (in 0.5°) may not be able to capture the NH3 concentration dynamics at a finer scale, if not overridden by some regional inventories. Moreover, the NH3 is actively reacted with NH4+, so it is worth discussing whether the sensitivity of NH3 and NH4+ together would be better to capture the sensitivity of the NHx (NH3 + NH4+) family to the emission.
The longer period has been used for spin-up (2010-2018), by using the CEDS global bottom-up gridded inventories as a prior. However, post-2019 was set with the carbon emission growth rate, which I think is inappropriate for the NH3 since 1) NH3 does not have an intense relationship with fossil fuel emissions, as an agricultural-based emission, and 2) they have different trends in anthropogenic sources, but may have a similar reflection on the biomass burning. Instead, the post-2019 prior could be set as invariant and adjust the simulated NH3 columns with the IASI observations, and the derived NH3 emission could be corrected by SO2/NOx change during the COVID lockdown. Or authors could just update it after the release of new CEDS emissions.
Although the paper focuses on the application of the inversion system in a high spatiotemporal resolution, the setup and suitability of the system are not sufficient enough to publish, before more tests and discussions on its sensitivity and consistency. For some parts a more detailed and cleared description could be useful, as described below in the Specific Comments. Overall, the paper is easy to read with a good structure, but could still not be published in the ACP in terms of the above scientific concerns.
Specific comments
- line 20 'all the spatiotemporal scales': provide concrete ranges of spatial scales (e.g., regional to global, 0.1° to 2° resolution) and temporal scales (e.g., daily, seasonal, interannual variations from 2010 to 2018)
- line 25 'prior CEDS inventory’s anthropogenic NH3 emissions': if only update the global anthropogenic NH3 emissions, consider modifying the title correspondingly to accurately reflect the focus
- line 31-32 Post-2019 emission trends: this conclusion highlights a limitation in the post-2019 prior emission, particularly for NOx and SO2, which may propagate unrealistic trends in NH3. I recommend explicitly addressing their impact to correct this unrealistic NH3 emission trend.
- line 74: NH3 emission estimates or NH3 concentrations? The majority of the paragraph is talking about NH3 observations, but latter you also mention the NOx emissions, a bit unclear.
- line 212: which kind of pre-/post-retrieval filters you applied, except for the cloud coverage?
- line 267: any cases for NH3?
- line 383-386: As shown in Figure 2, IASI NH3 columns are much higher than the model simulation. Assuming you still keep those negative values in IASI retrievals, does this bias arise from underestimated agricultural emissions in the prior inventory or systematic biases in IASI retrievals?
Technical corrections
- Acronym consistency: consistently use the acronym 'AK' for 'averaging kernels' from top to bottom, the same applies to others
- line 97-99: add some references to support
- line 111 'EDGAR': full name
- line 239 and 276 'interpolated onto the model horizontal grid': which interpolation method?
- Figure 1 clarification: what is the difference between 'LMDZ-INCA original' (orange) and 'LMDZ-INCA without AK' (red). Besides, the AK is higher with the lower pressure (higher elevation), but why the largest discrepancy happens at around 600-800 hPa
- line 312 'given hourly': 9-10 AM?
- 319-320: what is the definition of the 'high-quality IASI pixels' it looks rational if you exclude negative columns as long as the negative IASI NH3 total column has been kept after the filtering process.
- line 453-465: the description of the gap-filling method is important but consider moving it into the 'Material and methods' part
- line 658-660: it is inequivalent to compare your anthropogenic emissions with Luo 2022 and Dammers 2022, since they also included the natural sources (e.g., biomass burning). But your emission estimates (98 Tg yr-1) are still higher than Luo's (78 Tg yr-1), which is quite interesting and worth discussing by comparing with your spin-up stage (prior to 2019). I would like to see such a comparison in a Table or Figure.
- line 704: for India comparison, there is a new study you could check: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3938/#discussion
- line 785: is it possible to quantify the uncertainty for your emission estimates, via satellite retrieval errors/number and model transport biases?
Citation: https://doi.org/10.5194/egusphere-2025-162-RC1 -
RC2: 'Comment on egusphere-2025-162', Anonymous Referee #2, 31 Mar 2025
In this study, the authors investigate global ammonia (NH3) emissions from 2019-2022 by using satellite observations from IASI and a chemistry-transport model called LMDZ-INCA. The study updates nh3 emissions use the finite difference mass-balance through an atmospheric inversion technique. They use averaging kernels from the latest IASI data to improve accuracy when comparing model simulations to the satellite measurements. The research finds that existing emission inventories may significantly underestimate global anthropogenic NH3. Furthermore, the paper examines regional variations in NH3 emissions and their seasonality, noting discrepancies with current inventories and potential influences from COVID-19. The manuscript is well-structured and is well-written. However, there are certain things to be clarified before the MS can be accepted.
Specific comments:
L240-242: You mentioned that you use NOx and NH3 from CEDS for eleven sectors including the agricultural sector, and you also mentioned that CEDS emissions of NO and NH3 from agricultural soils with both synthetic and manure fertilizers. Are the NO and NH3 from agricultural soil emissions not included in the agricultural sector and provided separately?
L247-248, you use the CO2 data from the Carbon Monitor dataset to calculate emission growth rates of other species. This leads to noticeable variation in emissions of SO2 and NOx. Did you compare the changes with other inventories (such as global cams) to check if the changes are realistic? It would be nice if you could provide a figure in the supplement.
L267, finish the sentence.
L335-340, to select the grid cells with dominant NH3 emissions, do you use monthly emissions or yearly emissions?
Figure 3, please provide the figure with a higher resolution. The legends in the sub-figures are not easy to read.
Section 4.1 you compared your results to other emission datasets including emissions derived from CrIS. The overpass times of IASI and CrIS are different. The emission rates are different at the two overpass times. How accurate is the diurnal cycle of NH3 emissions in the model? I guess this could also be another reason for the difference in emissions deriving from IASI and CrIS.
Section 4.3. The uncertainties in emissions and limitations are discussed without quantifying the uncertainties of the estimated emissions. It would be nice to provide a simple estimate of errors/bias caused by uncertainties/ bias from satellite data. Furthermore, the gap-filling for the emissions can also introduce bias and errors.
Citation: https://doi.org/10.5194/egusphere-2025-162-RC2
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
290 | 51 | 9 | 350 | 24 | 7 | 5 |
- HTML: 290
- PDF: 51
- XML: 9
- Total: 350
- Supplement: 24
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|---|---|---|
United States of America | 1 | 126 | 36 |
France | 2 | 73 | 20 |
China | 3 | 41 | 11 |
India | 4 | 12 | 3 |
Germany | 5 | 9 | 2 |
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
- 126