AM0101 Final Report
Appendix IIIB. Fertilizer/crop emission sub-model
T.H. Misselbrook (IGER), M.A. Sutton (CEH) & D. Scholefield (IGER)
1. Background
In the inventory of NH3 emissions from UK agriculture, emissions from fertilizer applications were calculated using emission factors given by van der Weerden & Jarvis (1997). They gave different emission factors for urea and all other N fertilizer types and assumed that emission factors for applications to arable land would be 50% of their respective values for applications to grassland. This appendix describes the development of the N fertilizer module for the National Ammonia Reduction Strategy Evaluation System (NARSES), in which the emission factors for different N fertilizer types were derived from more process-based models, enabling the variation in climate, soil type and management practice across the UK to be accounted for when modelling the emissions at a local or national scale. Mechanistic models describing NH3 emissions from urea applications to soils exist (Sherlock and Goh 1985; Rachhpal-Singh and Nye 1986; Fleisher et al. 1987; Roelcke et al. 1996), but very few, if any, for other fertilizer types, reflecting the much greater potential for NH3 emissions from urea. These mechanistic models require a varying number of input parameters, some of which are very site- or soil-specific and not easily obtainable. This precludes the use of such detailed models in a more generalised model such as NARSES, which requires easily-obtainable parameters, averaged over fairly coarse spatial (e.g. 10 km grid square) and temporal (monthly) units. In a recent review of the topic, Harrison and Webb (2001) recommended that the emission factors derived by van der Weerden and Jarvis (1997) should continue to be used to calculate UK fertilizer-N emissions, with the exception of ammonium sulphate (AS) which should have a much greater emission factor on calcareous soils. However, they concluded that development of more process-based models was required for better prediction of NH3 emissions from N fertilizers and the foliage of fertilized crops. Bi-directional exchange (i.e. emission and deposition) of NH3 occurs over vegetated surfaces and the net exchange is a combination of leaf surface, stomata and soil exchange processes. Application of N fertilizer to the vegetation will lead to a net increase in NH3 emission (Sutton et al. 1993; Sutton et al. 2001). In practice, it is difficult to separate the NH3 emissions due directly from the fertilizer and those via the plant, and the model developed gives the net emission from the fertilized soil-crop system. It should also be made clear that only short-term emissions (i.e. within 2-3 weeks after fertilizer application) are being modelled and other periods of net emission (e.g. when grass is cut), which may also be a function of the nutrient status of the crop, and therefore indirectly due to fertilizer N application, are not considered.
2. Model description
Important influencing variables which are included in the model are type of N fertilizer, soil pH, land use, application rate, rainfall and temperature. Each fertilizer type is associated with a maximum potential emission (EFmax), which is modified by functions relating to the other variables (soil pH, land use, etc.,) to give an emission factor (EF) for a given scenario:
(1)
where RF are the reduction factors, expressed as a proportion, associated with each variable.
2.1. Fertilizer type
The maximum potential emissions for each fertilizer type refer to the maximum which might be expected under realistic field conditions. These were derived from the maximum reported emissions in published field experiments. From a number of reported studies, maximum cumulative emission for urea fertilizer applications was in the range 30 – 40% of applied N (Black et al. 1985a; Black et al. 1985b; Black et al. 1987; Ryden et al. 1987; Sommer and Jensen 1994; Fox et al. 1996) although a greater emission, of 46% of applied N, was reported by van der Weerden and Jarvis (1997). For the model, therefore, a maximum potential emission of 45% was used for urea. There are fewer reported studies for other fertilizer types, but those that exist suggest a maximum emission from ammonium nitrate (AN) of 4% of applied N (Black et al. 1985a; Ryden et al. 1987; van der Weerden and Jarvis 1997). Emissions from AS and di-ammonium phosphate (DAP) may be as large as those from urea, from applications to calcareous soils (Gezgin and Bayrakli 1995, Fenn and Kissel 1974), so maximum potential emission for these fertilizer types was also set at 45%. For urea ammonium nitrate (UAN), which is increasingly being used in the UK, the maximum potential emission was set at 25%, mid-way between those for urea and AN.
2.2. Soil pH
While solution pH is a major influencing variable in NH3 emission rate, there is an interaction between fertilizer type and the influence of soil pH. A number of studies on the influence of soil pH have been reported. Harrison and Webb (2001) concluded that for urea, hydrolysis greatly increases pH around the fertilizer granule leading to a large NH3 potential, which is relatively unaffected by soil pH (and the same might be assumed for UAN). For fertilizers, such as AS and DAP, that form sparingly soluble salts with calcium, thereby increasing dissolution of calcium carbonate and solution pH, emission rates will be greater from calcareous soils. For other fertilizers, such as AN, which form readily soluble salts with calcium, emission is largely unaffected by soil pH. In the model, therefore, a soil pH function is applied only to AS/DAP, with the maximum potential emission remaining at 45% for applications to calcareous soils (pH > 7) or reducing to 4% (as for AN) for applications to other soils. A value of 0.0889 was therefore used for RFtype for applications of AS/DAP to non-calcareous soils and 1.0 for applications to calcareous soils and all other fertilizer types.
2.3. Land use
There are few data relating to the influence of land use on NH3 emissions from fertilizers. There is no reason to expect a significant difference between emissions from applications to grassland or short arable crops and these are used as the default option in the model for which there is no modification of the emission factor. Applications to a taller crop (e.g. cereals of growth stage 30+) may be associated with reduced emissions, with the crop canopy reducing wind speed and temperature at the soil surface and also absorbing some NH3 emitted at the soil surface. Such phenomena have been noted for slurry applications placed in bands beneath a growing cereal crop (Sommer et al. 1997). Incorporation of the fertilizer into the soil either by direct placement or by cultivation soon after application will significantly reduce emissions (Black et al. 1989; Roelcke et al. 2002). In the model therefore, values of 0.7 and 0.2 were used for RFlanduse for applications to a taller crop and for incorporation or direct soil placement, respectively.
2.4. Application rate
Application rate was used as a modifier for the emission factor for urea, UAN, AS/DAP only. Black et al. (1985a) reported an increase in NH3 emission with increasing urea application rate, showing there to be an associated increase in soil surface pH. Emissions increased from 13% to 33% of applied N between application rates of 30 to 200 kg N ha-1. Van der Weerden and Jarvis (1997) reported similar increases in emission rates, although applications were made at different times and were therefore not directly comparable. For urea, RFrate was therefore determined by the function:
(2)
where rate is the application rate (kg N ha-1). Laboratory experiments have shown similar effects of rate for applications of AS/DAP to calcareous soils, but little effect on other soil types (Harrison and Webb 2001). With no field experimental data from which to derive a relationship, Equation (2) was used for UAN, AS/DAP applications to calcareous soils while a value of 1.0 for RFrate was used for other soil types and fertilizers.
2.5. Rainfall
Rainfall was only considered to have a significant effect where emissions are large (i.e. for urea and UAN to all soils and AS/DAP to calcareous soils). The intensity and timing of a rainfall event in relation to application timing are important factors. A number of authors report a decrease in emissions following rainfall (Black et al. 1987; Velthof et al. 1990; van der Weerden and Jarvis 1997). However, for urea, repeated small rainfall events may increase emissions by stimulating hydrolysis (Black et al. 1987). The impact of rainfall on the emission factor was therefore modelled according to rainfall intensity and timing for urea and UAN to all soils and AS/DAP to calcareous soils, with RFrainfall of 0.75, 0.80, 0.85, 0.90, 0.95 and 1.0 for significant rainfall events (>5 mm in 24 h) within 24 h, 24-48 h, 48-72 h, 72-96 h, 96-120 h and >120 h of application, respectively. For other fertilizer types, RFrainfall had a value of 1.0.
2.6. Temperature
In theory, we would expect a strong relationship between temperature and NH3 emission, as the partial pressure of NH3 in solution increases exponentially with temperature (Freney et al. 1983). However, while Ryden et al. (1987) noted an increase in emission from urea with increasing temperature, other authors noted very little influence of season (Black et al. 1985a; Sommer and Jensen 1994; van der Weerden and Jarvis 1997), although obviously factors other than temperature will be different for applications at different times of year.
For AN (and assumed for other N types where emission is largely through the soil/plant system), Sutton et al. (2001) have showed that, for a given month, emission rates did not vary significantly with geographical differences in temperature across the UK. However, there were significant seasonal differences, with emission rate increasing by a factor of approximately 3 for every 5oC increase in temperature. For AN and other N to all soils and AS/DAP to non-calcareous soils, RFtemperature is determined from:
(3)
where Tmonth and Tannual are the local mean values of monthly and annual temperature.
For urea and UAN to all soils and AS/DAP to calcareous soils, where emissions are largely directly from the fertilizer, geographical differences in temperature are considered to be important and emission rate is considered to increase by a factor of 2 for every 5oC increase in temperature. For these situations, RFtemperature is determined from:
(4)
where TUKannual is the mean annual temperature for the UK. The denominators in Equations 3 and 4 are included as scaling factors such that RFtemperature for the warmest month is close to a value of 1.
3. Comparison with national inventory methodologies
The model described has been used to estimate total emissions from fertilizer use in UK agriculture based on mean national values for the input parameters. Table 1 shows the emissions from fertilizer applications to agricultural land (excluding grazing) as estimated using the model and the current national agricultural NH3 emissions inventory model (Misselbrook et al. 2000). The model presented here gives a marginally smaller estimate of emission from fertilizers than the UK inventory (28.7 and 30.4 kt NH3-N, respectively). For emissions from applications to grassland, the two methods gave similar results, with the UK inventory giving a somewhat greater estimate of emission from urea. However, for emissions from arable land the model presented here gave greater estimates of emission from AN, AS/DAP and other N fertilizers.
In order to explain these differences, we calculated the implied emission factors as the estimate of emission expressed as a percentage of the quantity of fertilizer N of each type applied (to either arable or grassland) (Table 2). Implied emission factors derived from the model depend on the values for the different reduction factors in Equation (1) and the values in Table 2 are for a ‘standard’ UK scenario, for the most part as used in the Inventory. The main differences between the two sets of implied emission factors are in the differentiation for applications to arable land or grassland and in the values for urea and AS/DAP. The UK inventory calculates emission factors for applications to arable land as 50% of their respective values for applications to grassland, except for UAN. The model presented here does take land use into account, but the resulting implied emission factors for applications to arable land are only reduced by 20% in comparison with their respective values for applications to grassland. The EFs derived from the model are smaller than those used in the UK inventory for urea applications to both arable and grassland. The implied emission factors for applications of AS/DAP are greater than those used in the UK inventory in which AS/DAP are differentiated from AN or other N.
When fully implemented within the NARSES model (Webb et al., 2002), emissions can be calculated at a 10 km grid level and aggregated to derive the national estimate. This methodology is likely to give a different estimate of total emission as the interactions between soil type, climate and management practice will be better reflected. A limitation to this approach may be the availability of management practice data at that level of disaggregation. However, implicit in the development of NARSES is the view that national inventories should be constructed from estimates for disaggregated grid cells or regions.
4. Sensitivity analyses and scenario testing
Model runs were conducted in which the values of the variables included in the model were varied in order to assess their relative influence on NH3 emission. The ‘standard’ UK scenario using the fertilizer N inputs from Table 3, was taken as the baseline. In turn, and keeping all other variables at their ‘standard’ values, the value of each variable was varied over a range and the total NH3 emission calculated at each value. The minimum and maximum values used for each variable are given in Table 4.
The results of this sensitivity analyses are given in Figures 1 and 2. As would be expected from Equations 3 and 4, emissions were very sensitive to temperature. Figure 1 shows the difference in total emission estimate for the extremes between the lowest and highest mean monthly temperatures, giving more than a 9-fold increase. Obviously, mean annual temperatures will not vary by this much, but this shows the potential impact of calculating emissions on a disaggregated basis if fertilizer-N applications are biased toward warmer or cooler parts of the UK. The proportion of N applied as urea fertilizer also had a large influence on NH3 emission (Figure 1). Increasing the proportion from the current usage (approximately 10%, Table 1) to 50% of N applied (excluding that applied to grazing) increased the estimate of emissions from fertilizers from 28.7 to 78.5 kt NH3-N. Changing the proportion of N applied as AS/DAP had a smaller effect; the effect of changing the proportion of N applied as UAN was mid-way between that of urea and AS/DAP. Other variables to which NH3 emission was sensitive, although to a lesser extent, were application rate and the proportion of applied N which was incorporated into arable soils. Varying each of these over the ranges given in Table 4 resulted in changes in emission of approximately 10 kt NH3-N. Of lesser significance were the influence of proportion of applied N receiving significant rainfall within 1d and the proportion of N applied to a taller crop (giving 3 – 4 kt N difference). Varying the proportion of N applied to soils of pH >7 had little effect on the emission estimate, as soil pH only influences the emission factor of AS/DAP and these fertilizers account for <3% of the N applied (Table 1). Obviously, there are interactions between the variables being changed and for scenarios other than the standard UK given here, the significance of the different influencing variables may be different. For example, if the proportion of N applied as AS/DAP in the UK was much greater, then varying the proportion applied to soils of pH > 7 would have a much greater influence.