The land cover/use distribution data for areas of grassland, rice and other crops came from Zuidema et al. (1994). This global database derives from maps of natural vegetation and climate in combination with statistical information from FAO (2000). As the resolution is 0.5°, there are errors in the gridded data for small countries. For example, the total land area of a small country may not be represented by the 0.5° grid cells. Such problems occur often with islands or coastal areas.
The data on grassland management are extremely scant. However, in order to extrapolate NH3 volatilization from fertilizer and manure application, it is necessary to know the location of more or less intensively used grasslands. This process involved some simple assumptions with three types of grassland being defined: (i) intensively used grasslands, receiving only inputs from animal manure, and defined as grassland located within arable areas; (ii) fertilized grasslands, receiving nutrient inputs from mineral fertilizers as reported by IFA/IFDC/FAO (1999) in addition to animal manure (these grasslands are located within intensively used grasslands); (iii) extensive grasslands consisting of the remaining grasslands given by Zuidema et al. (1994) were not considered as they are grazed and deliberate application of fertilizers or manure was assumed to be negligible.
For most countries, the intensively used grasslands are in those cells where arable land makes up at least one third of the area. It was necessary to make many exceptions in order to match the areas with those of IFA/IFDC/FAO (1999).
For wetland rice fields, the N fertilizer use per hectare values are from IFA/IFDC/FAO (1999), with some exceptions. The N fertilizer use for the other, dryland crops was calculated as the difference between total N fertilizer use from IFA (1999) and the sum of the N used in wetland rice and grasslands from IFA/IFDC/FAO (1999), with exceptions for some countries.
The quantity of animal manure N available for fertilizing crops and grasslands was calculated from total excretion by excluding the excretion during grazing, use of animal manure as fuel and NH3 losses during storage of animal manure. Half the animal excreta available as fertilizer were assumed to be used in croplands and the other half in grasslands. For some countries, it was necessary to adjust the percentage of the manure applied on grasslands in order to avoid excessive application rates. Thus, the assumed maximum N application rate was 200 kg/ha for industrialized countries, and 50 kg/ha for developing countries; this affects the availability for crops. Manure application rates for wetland rice and upland crops were assumed to be equal.
Ammonia volatilisation rates are linear with respect to the N application rate. However, the models for N2O and NO emissions show a non-linear response of emissions to N application rates. Therefore, for N2O and NO the average country application rates used for ammonia would not give a correct estimation of emissions. To account for the fact that not all agricultural fields receive fertilizer, IFA/IFDC/FAO (1999) data on mineral fertilizer N application rates and the percentage of each crop actually receiving fertilizer N were used to calculate the mean application rate for upland crops for the area actually fertilized. These data are available for 49 countries, including some with 100 percent of the fields fertilized. The analysis included data for 38 developing countries and made the assumption that a fraction of the fields receive fertilizer in order to determine apparent total N use per country. For other developing countries, N application rates were assumed equal to calculated rates for Central America, South America, Asia and Africa. There are no statistics on the use of animal manure in croplands and grasslands. Estimates for regions within countries may be available, but sometimes they do not correspond to official statistics or are outdated. The paucity of data makes it necessary to generalize. Therefore, for animal manure applied on croplands the same minimum N rates were used. For manure applied to grasslands N application rates of 50 percent of those for upland crops were assumed to reflect lower application rates in grasslands relative to crops.
In addition to the land use data, the 0.5° resolution information on soil pH, CEC, organic C content, soil texture and soil drainage came from Batjes (1997), and a map of climate types with 0.5° resolution from Fresco et al. (1998). For NH3 this climate database was not used as it is difficult to estimate the temperature at the time of fertilizer application (generally the beginning of the growing season). Therefore, for extrapolating the NH3 volatilization rates, the assumption was that the temperature at fertilizer application is >20°C between 40°N and 40°S, and <20°C at other latitudes.
A number of additional assumptions had to be made. Measurements of N2O and NO emissions during the crop season when the rice fields are inundated dominate the data set. Therefore, in the extrapolation of N2O emissions from rice, for rice the model parameters for 'other crops' were used to account for the emissions during the post-harvest period.
Fertilizer management, which had a significant effect on NH3 volatilization rates, is the least known aspect in the extrapolation. Although much is known at local and national level, it is difficult to generalize such information in a global inventory. Much of the fertilizer applied to rice in Southeast Asia is either broadcast directly onto flooded soil 14-21 days after transplanting or broadcast directly onto flooded soil after transplanting. Broadcasting is also common in grasslands.
Therefore, the general assumption about fertilizer application was that of a basal application by broadcasting. Exceptions included anhydrous ammonia (estimates based on incorporation), and fertilizer solutions (liquid). Animal manure used in rice cultivation was assumed to be incorporated. If incorporation of fertilizers is a more common practice, the estimates of NH3 loss would require adjustment downwards.
It is difficult to quantify uncertainties associated with scaling. Uncertainties in fertilizer use stem primarily from the grouping of different N fertilizer types into one category. The data on fertilizer use by crops have a varying degree of reliability; data are more complete and probably more certain in industrialized countries than in developing countries. However, there are no statistics on fertilizer management. The extrapolation included assumptions on the mode of application of fertilizers and manure. Although application rates vary between crops and farmers within countries, the spatial distribution of fertilizer application rates is not well known. Moreover, it is not known if certain fertilizer types are preferentially used for specific crops or grasslands.
The distribution of arable lands is well known, but data on the distribution and the management of grasslands are uncertain. Although information is available on the use of mineral fertilizers in grasslands, the application of animal manure in grasslands is based on a global estimate. For different world regions and individual countries, the application of animal manure is highly uncertain.
Animal populations are well known, although the season in which censuses are made does cause some uncertainty. The uncertainty in the data on animal populations is probably <10 percent. Most of the uncertainty in the NH3 volatilization rate for animal manure stems from the assumptions on N excretion and waste management. For developing countries, in particular, there are no reliable data on waste management practices and on the use of animal manure in grasslands and arable lands. Therefore, uncertainty is probably highest for tropical countries and lowest for western Europe.
The simple models for N2O and NO discussed in Chapter 4 were used in a GIS with the geographic information and the assumptions on fertilizer management discussed above.
The combined resulting figures from Tables 10 and 11 yield a global annual N2O-N emission (about 3.5 million t) and NO-N emissions (about 2.0 million t). Some 34 percent of the global annual N2O-N emission from cropped fields of 3.2 million t stems from developed countries and 66 percent from developing countries. The global emission from cropped fields is 3.3 percent of the N fertilizer input (3.3 percent in developed countries and 3.4 percent in developing countries). The annual NO-N emission from fertilized crops amounts to about 1.5 million t, 44 percent from developed and 56 percent from developed countries. This is 2.0 percent of the N input for developed countries and 1.4 percent for developing countries.
The annual N2O-N emission from fertilized grasslands amounts to about 0.3 million t (58 percent from developed countries and 42 percent from developing countries), and that of NO-N to about 0.5 million t (68 percent stemming from developed and 32 percent from developing countries) (Table 11).
The fertilizer induced N2O emissions were calculated as the total emission minus that of unfertilized fields (all other conditions equal to the fertilized plot), expressed as a percentage of the N input (Table 12). The results indicate important differences between fertilizer types. The global annual fertilizer induced N2O-N emission is about 0.9 million t, or 0.8 percent of the N input, lower than the 1.25 percent estimated by Bouwman (1996). The highest fertilizer induced emission rates are for urea and other straight N fertilizers (primarily ABC used in China), and the lowest for organic fertilizers.
The global annual fertilizer induced emission for NO-N is about 0.6 million t, or 0.5 percent of the N input (Table 12), which is in agreement with the estimate of 0.5 percent by Veldkamp and Keller (1997a). The highest fertilizer induced emission rates of 0.8 percent are for compound NK-N and other straight nitrogen fertilizers (primarily ABC used in China). The lowest fertilizer induced emission rates are calculated for organic fertilizers (0.4%).
The calculation of a median estimate involved the use of log-transformed emissions and balancing. The result is a `best' estimate of the emission for a specific combination of factor classes. With the residual maximum likelihood procedure it is not possible to produce an estimate of the standard error of the model estimate of emissions. Therefore, the standard errors were calculated for a range of combinations of the factors considered for a weighted multiple regression model for N2O. The multiple regression used only those records in the data set that have values for all factors (353 measurements). The variance explained by this model is 50 percent, and the overall uncertainty ranges from about -40 percent to +70 percent based on twice the standard error accounting for 95 percent of the data. The uncertainty of the residual maximum likelihood model is probably less than that of the weighted multiple regression because the residual maximum likelihood procedure considered more measurements (846 for N2O). For NO, the number of degrees of freedom after weighting is too small to derive a model that resembles the residual maximum likelihood model, and it is therefore difficult to estimate the uncertainty of the model.
The models for N2O and NO are non-linear with respect to the N application rate. Although estimates for the areas actually fertilized were used, the heterogeneity in N application rates may still influence the model results. The associated error in the extrapolated emission estimates is acceptable when compared to other uncertainties in the global extrapolation. The models are not suitable for predicting emissions from measurements in individual research papers for specific sites. However, the mean emission rates for factor class combinations are of more interest for extrapolation to `landscape' conditions than individual measurements.
The summary model (Chapter 4) was used to calculate regional and global volatilization losses. This extrapolation exercise combined various sources of statistical data and geographical information. Tables 13 and 14 summarize the results of the extrapolation. The global ammonia loss from mineral fertilizers of 11 million t N/year (14 percent of mineral fertilizer N use) estimated in this report is in close agreement with Bouwman et al. (1997). The results indicate that NH3 volatilization rates in developing countries exceed those in developed countries by a factor of 4.3.
The global NH3 loss from the annual use of 11.8 million t of mineral fertilizer N in wetland rice cultivation amounts to 2.3 million t N/year, or 20 percent of the N application. Most of this loss occurs in developing countries (97 percent). In upland crops, 14 percent of the 61.7 million t of mineral fertilizer N is lost as NH3, with higher loss rates in developing countries (18 percent) than in developed countries (8 percent). In grasslands, the annual global use of mineral fertilizer N is 4.3 million t N, with estimated loss rates of 13 percent for developing countries and 6 percent for developed countries. Nearly 100 percent of mineral N fertilizer use in grasslands is in developed countries.
The global NH3 loss from the annual use of 12.4 million t N in animal manure in grasslands amounts to 2.7 million t N/year, of which about 60 percent stems from developed countries. The NH3 volatilization rate from animal manure is 22 percent of the N application (20 percent in developed countries and 25 percent in developing countries). In upland crops, 26 percent of the 17.4 million t N from animal manure is lost as NH3, with higher loss rates in developing countries (29 percent) than in developed countries (22 percent). This higher rate for developing countries is due to high temperatures and the dominance of the use of urea, ABC and ammonium sulphate in wetland rice cultivation. The volume of animal manure applied annually to upland crops is 8.6 million t N in developed countries and 8.8 million t N in developing countries. In wetland rice systems, the estimated annual use of N from animal manure is 3.3 million t N, mainly in developing countries. As incorporation is assumed to be prevalent in rice cultivation, the NH3 volatilization rates are lower than for upland crops (17 percent in developing countries and 16 percent in developed countries).
Table 15 compares the global mean NH3 volatilization rates for the fertilizer types (IFA, 1999) based on the summary regression model with other inventories from the literature. The results are in agreement with the ECETOC (1994) and Bouwman et al. (1997) estimates, except for ammonium sulphate and the compound fertilizers including MAP and DAP. The data collected in this report for ammonium sulphate are based on 176 measurements from a great number of different research papers and from different sites with different conditions. Hence, the results can be considered to be representative for actual conditions in the field.
The NH3 volatilization rates found for phosphorous containing fertilizers are higher than values presented by ECETOC (1994) and Bouwman et al. (1997). This may be due to the influence of phosphate which may change the environment to favour increased NH3 loss by precipitating Ca. Both the formula (pH) and form of the phosphate added can influence the reaction with Ca, and therefore NH3 losses. It is not certain what the composition is of the fertilizer categories NP-N and NPK-N. In this report the volatilization rate for these compound fertilizers was based on data for all NP fertilizers present in the data base; the NH3 volatilization rate for ammonium phosphate (AP) is based on the results for MAP and DAP (Tables 9 and 15).
Uncertainties in the results of the extrapolation stem from uncertainties in the summary model as such and uncertainties caused by scaling errors. The range of NH3 volatilization rates for the various fertilizers was calculated using the standard errors of the regression. The standard errors depended on the combination of factor classes selected, but appeared to be very similar for individual fertilizer types across crop types, climate and soil conditions. Therefore, the range of volatilization rates was calculated for a number of combinations of factor classes for each crop type and fertilizer type. This was done on the basis of twice the standard errors to include 95 percent of the observations. The average deviation from the model results in Table 15 was applied to the estimated global NH3 volatilization loss for each fertilizer category. The resulting range in the estimates for the global NH3 volatilization loss from all fertilizers is 10-19 percent for all mineral fertilizers and 19-29 percent for animal manure. The range for some individual fertilizers is much wider, depending on the number of representations in the data set (Table 15). The calculated ranges do not account for the omissions in the summary model or for scaling errors.
Because the fertilizer application rate was found not to influence the NH3 volatilization rate, NH3 losses could be scaled up based on average country fertilizer and manure application rates. Hence, it is recognized that there are errors in the spatial distribution of the mix of fertilizers and their application rates, and their combination with soil conditions. However, for the 0.5° resolution used, this error is acceptable.