the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A dynamical process-based model AMmonia–CLIMate v1.0 (AMCLIM v1.0) for quantifying global agricultural ammonia emissions – Part 1: Land module for simulating emissions from synthetic fertilizer use
Abstract. Ammonia (NH3) emissions mainly originate from agricultural practices and can have multiple adverse impacts on the environment. With the substantial increase of synthetic fertilizer use over the past decades, volatilization of NH3 has become a major loss of N applied to land. Since NH3 can be strongly influenced by both environmental conditions and local management practices, a better estimate of NH3 emissions from fertilizer use requires improved understanding of the relevant processes. This study describes a new process-based model, AMmonia–CLIMate (AMCLIM), for quantifying agricultural NH3 emissions. More specifically, the present paper focuses on the development of a module (AMCLIM–Land) that is used for simulating NH3 emissions from synthetic fertilizer use. (Other modules, together termed as AMCLIM-Livestock, simulate NH3 emissions from agricultural livestock, are described in Part 2). AMCLIM–Land dynamically models the evolution of N species in soils by incorporating the effects of both environmental factors and management practices to determine the NH3 emissions released from the land to the atmosphere. Based on simulations for 2010, NH3 emissions resulting from the synthetic fertilizer use are estimated at 15.0 Tg N yr-1, accounting for around 17 % of applied fertilizer N. Strong spatial and seasonal variations are found. Higher emissions typically occur in agricultural intensive countries (such as China, India, Pakistan and US), and mostly reach the maximum in the summer season. Volatilization rates indicate that hotter environments can result in more N lost due to NH3 emissions, and show how other factors including soil moisture and pH can greatly affect volatilization of NH3. The AMCLIM model also allows estimation of how application techniques and fertilizer type have impacts on the NH3 emissions, pointing to the importance of improving management practice to tackle nutrient loss and of appropriate data-gathering to record management practices internationally.
- Preprint
(9786 KB) - Metadata XML
-
Supplement
(1390 KB) - BibTeX
- EndNote
Status: open (until 04 Jun 2024)
-
RC1: 'Comment on egusphere-2024-962', Anonymous Referee #1, 25 Apr 2024
reply
The manuscript reports a new process-based model for the emission of ammonia following fertiliser application, and then applies this on a global scale. The topic of agricultural ammonia emissions is an important one, and there is clearly a need for well-calibrated, process-based models as an aid to understanding the basic processes involved in ammonia volatilisation and as a means of upscaling to regional/global scales.
My understanding is that the AMCLIM model is a lightweight model that focuses on ammonia emissions in a period of 1-2 weeks following fertilisation, when the vast majority of ammonia is emitted. It uses a relatively simple Ohm's law like structure, which leads to most of the equations being linear. Non-linear contributions to N cycling, for example due to microbial activity, coupling between C and N and mineralisation of organic N compounds are ignored as being unimportant for predicting ammonia emissions shortly after fertilisation. Soil moisture and soil temperature are not explicitly modelled, but rely on measurement data. The relative simplicity makes the model a potentially valuable tool for experimental groups carrying out ammonia emission measurements. As compared to more detailed process based models (e.g. Daycent or DNDC type models) I assume the model is considerably easier to set-up, faster to run and doesn't require long spin up periods. Furthermore, the comparison to the GRAMINAE site shows that the model does a good job of capturing the variation in ammonia emissions in the days following fertilisation.
On the other hand, I think the application of the AMCLIM model to global ammonia emissions from croplands is premature. Insufficient evidence is provided to show that the model is well calibrated. The comparison to a single grassland site in Germany suggests that the model shows promise in capturing the diurnal cycle of ammonia emissions following fertilisation. However, before applying the model to global croplands I would like to see (see also calibration comments below):
1. Improved evidence that the pH dependence of ammonia emissions is well represented. Figure 13a,b provides some information in this direction, but it is hard to conclude from this that the model is well calibrated (for example a factor 2 difference in the y axis scale is needed to show the modelled Pv values as compared to the measured values).
2. Evidence that the pH change and the impact this has on ammonia emissions following urea application is well calibrated.
3. Evidence that the total ammonia emissions are well represented across multiple crop, soil and climate conditions. For example, that the model can capture ammonia emissions from paddy rice fields in South East Asia. Figure 12 goes some way in this direction, but I find it hard to conclude from this figure that the model is performing well across multiple conditions.
My opinion is that the model description and application to the GRAMINAE site would already make a valuable paper. I would suggest to leave out the application to global croplands and publish this at a later date, once more extensive calibration can be performed using site-scale data.
More detailed questions/comments follow below:
** Model details **
The AMCLIM model includes N uptake by plants, but ignores N uptake by microbes (immobilisation). This may be an important process, especially for fertilisation events at or before planting, when plant N uptake is low. Please discuss why/under what circumstances it is reasonable to ignore microbial N uptake.
To what extent has the sensitivity of the model to changes in temporal and spatial resolution been tested? Ideally the spatial and temporal resolution is reduced until the ammonia emissions become relatively independent of further decreases (and I see no reason why the resolution cannot be made finer than the meteorological and soil inputs, e.g. by splitting each soil layer into sub-layers). Figure 5 suggests that changing the spatial resolution of the top soil layer leads to large changes in the model behaviour (comparing circles for z1=1,2,3 cm), and thus that the model behaviour has not yet converged.
I would expect the time-step to be important as the underlying processes have very different response times. In particular the chemical equilibrium reaction between NH3 and NH4+ is much faster than plant N uptake or nitrification. As such I think it is important to have some time-step control, especially in the minutes following broadcast fertilisation (is 15mins / 1hour short enough?). Similarly, the spatial scale will control the interaction between the concentration of NH3 and NH4+ in the top soil layer following broadcast fertilisation and the transport processes (is 4 soil layers enough?).
As a related point, it would be useful to briefly mention how the coupled differential equations are solved. Is this by a Euler method or is a higher order method used?
I think the section 'volatilisation of NH3' could be improved, in particular the description of how the surface NH3 concentration is calculated.
It would also be useful to provide the recovery function for soil pH following urea fertilisation.
For the plant N uptake, what values are used for W_r,i (SM14 and SM17)? I couldn't find this in the supplementary information. Also it would be useful to mention how perennial crops such as grass are treated, especially since this is relevant for the GRAMINAE site (the stages in Table S1 seem to be for annual crops).
** Calibration **
As far as I understand, model parameters are taken from the literature, and are mostly not calibrated by comparing the AMCLIM model to measurements (a small number of model variations are shown in Figure 5, but are only compared to a single 10-day measurement at 1 site). Compare, for example, to Gurung et al. Nutr Cycl Agroecosyst (2021) 119:259–273, where Bayesian techniques are used to perform a joint calibration of the 18 parameters relevant to ammonia volatilisation, comparing different levels of model complexity and by taking into account 8 different experimental sites with 42 site-year treatments. I understand that the authors cannot do everything in one paper, and am not expecting them to perform a full Bayesian calibration in this manuscript. However, I think that a comparable level of calibration is necessary before applying the model at a regional/global scale.
I would find it useful to have a table of all model parameters and their values (e.g. in the supplementary material).
** Global simulations **
I mentioned above that I believe the application to ammonia emissions from global croplands is premature, and requires additional calibration of the model. However, if the authors choose to retain the global simulations in the manuscript, it would be useful to address the following points:
As discussed in the manuscript, correct fertiliser timing is important, due to the sensitivity of ammonia emissions to meteorological conditions (especially temperature). As such, the assumption that 50% is applied at planting and 50% midway through the growing season on a global scale seems a very crude approximation. Is no better data available? If not, how much do emissions change when these assumptions are varied?
The lack of model adaption for paddy rice systems means it is likely unreliable for these systems. I would suggest to either adapt the model to paddy rice, or to leave rice out of the global simulation.
An uncertainty estimate is given in the discussion section, but no details are provided as to how this was calculated. Please provide details so that the reader can judge how seriously to take this estimate.
** Discussion **
I would find it useful to discuss:
How the model differs at a process level from other process-based models such as FAN, DLEM, Daycent or DNDC type models. What are the advantages and disadvantages of the AMCLIM model with respect to these established models and what do the authors see as the future role of the AMCLIM model?
How does the calibration procedure compare to these other models, and what are the consequences for the level of confidence we should have in the AMCLIM model results as compared to established models?
Why has urease inhibition not been considered? This is required by many countries when broadcast spreading urea fertiliser and has important consequences for ammonia emissions. To what extent does this limit the usefulness of the model?
Citation: https://doi.org/10.5194/egusphere-2024-962-RC1
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
138 | 24 | 10 | 172 | 9 | 7 | 7 |
- HTML: 138
- PDF: 24
- XML: 10
- Total: 172
- Supplement: 9
- BibTeX: 7
- EndNote: 7
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1