lable at ScienceDirect

Environmental Pollution 159 (2011) 3183e3192

Contents lists avai

Environmental Pollution

journal homepage: www.elsevier .com/locate/envpol

Effects of farm heterogeneity and methods for upscaling on modelled nitrogenlosses in agricultural landscapes

T. Dalgaard a,*, N. Hutchings a, U. Dragosits b, J.E. Olesen a, C. Kjeldsen a, J.L. Drouet c, P. Cellier c

aAarhus University, Department of Agroecology, Blichers Allé 20, P.O. Box 50, DK-8830 Tjele, DenmarkbCEH Edinburgh, Bush Estate, Penicuik, Midlothian EH26 0QB, Scotland, UKc INRA, UMR Environnement et Grandes Cultures, BP 01, 78850 Thiverval-Grignon, France

a r t i c l e i n f o

Article history:Received 22 February 2011Accepted 26 February 2011

Keywords:AgricultureGreenhouse gas emissionsHeterogeneityLandscapeNitrogen lossesUpscaling

* Corresponding author.E-mail address: tommy.dalgaard@agrsci.dk (T. Dal

0269-7491/$ e see front matter � 2011 Elsevier Ltd.doi:10.1016/j.envpol.2011.02.043

a b s t r a c t

The aim of this study is to illustrate the importance of farm scale heterogeneity on nitrogen (N) losses inagricultural landscapes. Results are exemplified with a chain of N models calculating farm-N balancesand distributing the N-surplus to N-losses (volatilisation, denitrification, leaching) and soil-N accumu-lation/release in a Danish landscape. Possible non-linearities in upscaling are assessed by comparingaverage model results based on (i) individual farm level calculations and (ii) averaged inputs at landscapelevel. Effects of the non-linearities that appear when scaling up from farm to landscape are demon-strated. Especially in relation to ammonia losses the non-linearity between livestock density and N-lossis significant (p > 0.999), with around 20e30% difference compared to a scaling procedure not taking thisnon-linearity into account. A significant effect of farm type on soil N accumulation (p > 0.95) was alsoidentified and needs to be included when modelling landscape level N-fluxes and greenhouse gasemissions.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Excess of nitrogen (N) is one of the main characteristics ofmodern, intensive agriculture, leading to severe problems with Npollution (Erisman et al., 2007) and related emissions of green-house gases (Johnson et al., 2007). In this context, The NitroEuropeResearch Project (Sutton et al., 2007; NitroEurope, 2010) studies theinteractions between sources and sinks of N in order to understandits effects cascading through ecosystems in its multiple forms,represented by the more than nine different oxidation stepspossible (Galloway, 1998).

In relation to agriculture, the three main types of N pollution arefrom ammonia (NH3), nitrate (NO3

�) or nitrous oxide (N2O).Emissions of NH3 gas are mainly related to intensive livestockproduction and manure management, and the environmentalproblems caused relate to its contribution to the formation ofsecondary particulate matter in the atmosphere which has anadverse impact on biodiversity when deposited in vulnerableterrestrial or aquatic ecosystems (Bobbink et al., 2010). Nitrateleaching leads to eutrophication of surface waters or pollution of

gaard).

All rights reserved.

groundwater (Sonneveld et al., 2010; Kronvang et al., 2008). Finally,the greenhouse gas N2O can be emitted when ammonium isnitrified to nitrate and when nitrate is partially denitrified,processes that can occur in manure management systems(Dämmgen and Hutchings, 2008) and soils, or in groundwater orwetland areas receiving nitrates in runoff or leaching from agri-cultural land (Chirinda et al., 2010; Johnson et al., 2007).

Landscapes consist of a range of elements such as farmland,forestry and wetlands, that each can receive and lose a range of Ncompounds. The losses of N to the atmospheric or aquatic envi-ronments from these landscape elements are usually accountedseparately (e.g. Hansen et al., 2000 for farmland; Gundersen et al.,1998 for forestry; Maltais-Landry et al., 2009 for wetlands), usingecosystem models that are spatially one-dimensional (vertical).This contrasts with models of the atmospheric and aquatic envi-ronments, where three-dimensional spatial models are common(e.g. Hellsten et al., 2008; Hutchings et al., 2001). In these models,interactions in the horizontal plane are acknowledged to beimportant. For example, it is acknowledged that the distancebetween an N-sensitive ecosystem and a source of NH3 is importantin determining deposition of N from the atmosphere to thatecosystem (Dragosits et al., 2006).

Agricultural systems are commonly major sources of Ncompounds at the landscape scale, the different types and

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e31923184

quantities of N pollution being dependent on management prac-tices (Kros et al., in this issue; Hansen et al., 2011). The fate of theseN compounds and the consequences for the environment dependson the proximity of agricultural and non-agricultural landscapeelements and their degree of connectivity via transport mecha-nisms (e.g. atmospheric or aquatic transport pathways). Gergel(2005) showed that spatial heterogeneity at the landscape levelaffects N pollution and greenhouse gas emissions, and that land-scape level design of farming practices consequently may be aneffective measure to mitigate such losses.

The importance of the geographic location of landscapeelements and their interconnectivity have been investigated in theNitroEurope EU research project, where a unified landscape levelNitroScape model is being constructed (Sutton et al., 2007; Duretzet al, in this issue).

In this context, the present paper is a pilot study for the furtherlandscape scale work, in which we focus on the farming systemcomponent, and the effect of heterogeneity within this system,when scaling up from farm to landscape level. In our paper, wetherefore aim to test the effects of farm heterogeneity on themodelling and upscaling of nitrogen losses and greenhouse gasemissions, exemplified in a specific Danish landscape. We espe-cially focus on the effects of spatial heterogeneity in livestockdensity, the type of farming (ruminants, granivores, arablefarming), and the use of N in the arable and the livestock produc-tion system. Farm N-surpluses are calculated and partitionedbetween the three main types of N-losses: ammonia emissions,nitrate leaching and denitrification, and the corresponding changesin N in the soil. Based on these results, we compare examples ofdifferent upscaling approaches, and identify and discuss theimportance of non-linearities to be included in landscape levelfarm management models.

2. Materials and methods

In this chapter we describe the models for farm N-fluxes and the relatedgreenhouse gas emissions (Section 2.1), the approaches for the upscaling of N-lossesand greenhouse gas emissions from the farm to the landscape scale (Section 2.2),

Fig. 1. Schematic presentation of the farm nitrogen balance, and the nitrogen flows included isum of N in inputs minus N in outputs (both marked in red) and is distributed into N-lossesnitrous oxide (N2O), or changes in the soil-N pools. This modelling is based on emission factoDEN (Vinther and Hansen, 2004) component models. (For interpretation of the references to

and the agricultural landscape dataset for which these models are set up and onwhich the upscaling approaches are tested (Section 2.3).

2.1. Nitrogen modelling

2.1.1. Farm nitrogen balances and N-surplusThe farm gate N-surplus is defined as the difference between net input (i) to the

farm in the form of net feed import, net fertiliser import and N from the atmosphere;and net N output (o) from the farm in the form of milk (o1) and meat. The net meatexport is calculated as N in the meat exported (o2) minus N in imported livestock(i5). The net feed import is calculated as the sum of N in imported feed (i1) and seeds(i2), minus N in cash crops and straw materials sold (o4). The net fertiliser import iscalculated as the sum of N in imported synthetic fertiliser (i3) and manure (i4),minus N in manure exported (o3). Finally, the sum i6 þ i7 covers N from theatmosphere in the form of N-deposition, and N fixed by legumes (equation (1),Dalgaard et al., 1998). The N-surplus is calculated as:

N� surplus ¼ i1þ i2þ i3þ i4þ i5þ i6þ i7� o1� o2� o3� o4 (1)

This nitrogen balance (N-surplus) is calculated on an annual basis for each farmbythe Farm-N model (Jørgensen et al., 2005; Hutchings et al., 2006; Zander et al., 2009,http://www.fasset.dk/Upload/Fasset/Document/FARM-N_scientific_description.pdf). Within this model, the manure and synthetic fertiliser are distributed to theregistered crops on each farm in accordance with the official Danish fertilisationmanual, where N-fertilisation rates are determined as 10% below the economicoptimum in relation to the actual fertiliser prices (The Danish Plant Directorate,2007). This manual is also used as a reference book to determine crop yields basedon the soil types registered and assumptions about crop rotations estimated by thelinear programming probabilitymodel of Detlefsen and Jensen (2007). For each crop,the N-fixation from the atmosphere is calculated via the procedures of Høgh-Jensenet al. (2003, 2004), and the local N-deposition is set to 15 kg N ha�1 yr�1 (Dalgaardet al., 2002a,b). The feeding for each livestock category is based on the demand forenergy and protein in terms of the production data defined (Poulsen et al., 2001), andthe need for straw is derived from the housing system defined for each livestockcategory and each farm. Cropproduction surpluses are exported from the farm,whilefeedstuffs needed to implement the defined feed plan are imported to the farm.Finally, the Farm-Nmodel distributes the N-surplus into farmnitrogen loss pathwaysin the form of ammonia emission (Nemission), nitrate leaching (Nleaching) or denitrifi-cation (Ndenitrification), and estimates the soil-N accumulation (Naccumulation), definedas the accumulation of nitrogen in the soils farmed by each individual farm (equation(2), Fig. 1). This modelling is briefly described in the following sections.

N-surplus ¼ Nemission þ Nleaching þ Ndenitrification þ Naccumulation (2)

2.1.2. Ammonia emissionThe ammonia emission from manure and synthetic fertilisers spread in the

fields, livestock houses, andmanure stores (Fig.1) is calculated via the official Danish

n the Farm-Nmodel (Hutchings et al., 2006). The farm gate N-surplus is calculated as thein the form of ammonia (NH3), nitrate (NO3

�) or denitrification to free nitrogen (N2) orrs (EF; Hutchings et al., 2001) in combination with the C-tool (Petersen, 2007) and SIM-colour in this figure legend, the reader is referred to the web version of this article.)

Data Model

ResultModel

Steps in the scaling procedure

1

0

Scale (log km2)

Aggregation &

averaging

Aggregation &

averaging

Landscape level

Farm level

Fig. 2. Two approaches for upscaling of farm information and model results from farmto landscape level: 1) modelling before aggregation and averaging (the dotted line),and 2) averaging before modelling and landscape level aggregation of model results(the solid line). Based on Marshall et al. (1998) and Kjeldsen et al. (2006).

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e3192 3185

emission factor model (Geels et al., 2006), and according to the equations docu-mented by Hutchings et al. (2001).

2.1.3. DenitrificationThe denitrification in the form of N2 and N2O is modelled with the empirical

model SIM-DEN (Vinther and Hansen, 2004). In this model the N2O emission isderived from the input of N in mineral fertiliser and animal manure and emissionfactors recommended by the IPCC (1997) methodology. Tabulated values of the N2/N2O ratios are in the SIM-DEN model based on literature values, which were foundto be related to hydraulic properties of the soil and the clay content. The inputs tothis model are therefore the amount of N added in synthetic fertiliser or manure

Fig. 3. The Danish NitroEurope Study landscape is located in Midwest Jutland on the borde(European Environmental Agency, 2008). The left hand map shows the exact location of tsurrounding area, where data for the 56 farms studied were collected.

(applied N minus the ammonia emission), the N fixed in N-fixing crops, the claycontent, and the soil fertility level. E.g. ruminant farms have a high soil fertility, andcash-crop farms, with no manure added, have a low soil fertility and a relativelylower soil organic matter content (Heidmann et al., 2001).

2.1.4. Soil nitrogen storageIn accordance with the results by Petersen (2007), the changes in soil-N-storage

was estimated using the dynamic C-TOOL model (Petersen et al., 2002). The timescale for the calculations with C-TOOL was set to 10 years. With an assumption ofconstant C:N ratios for the different C-pools (see Petersen, 2007), changes in soilorganic N were calculated as the average annual decay minus addition.

2.1.5. Nitrate leachingThe N-leaching for the defined combination of cropping and fertilisation prac-

tices was modelled with the NLESmodel (Simmelsgaard and Djurhuus,1998), wherethe percolation in mm is calculated with the WATCROS model (Aslyng and Hansen,1982). Often the sum of modelled field N-losses (emission, leaching, denitrificationand soil-N mining) is not equal to the calculated field N-surplus (inputeoutput), andin order to yield the same result of equations (1) and (2), the N unaccounted for bythe models is partitioned between the harvest yield, the denitrification, the soil-Nstorage and the N-leaching using an algorithm based on expert judgement(Hutchings et al., 2006).

2.2. Upscaling framework

Two different approaches for upscaling farm level model results to the land-scape level are compared (Fig. 2). In the first approach, the models are executedseparately for each farm in the landscape before aggregating and averaging results tolandscape level for ammonia emission, nitrate leaching or denitrification. In thesecond approach, the input data are aggregated and averaged before modelling, anda comparison of the results from the two approaches is carried out to test whethersignificant non-linearities, in relation to each of the types of N-losses modelled, areaveraged out by the second approach compared to the first approach.

A key requirement for testing the effects of heterogeneity when scaling up fromfarm to landscape, is the availability of detailed, farm level landscape datasets, suchas the one from the Danish landscape described below. With such datasets, and thefarm models implemented, we illustrate and test effects of farm (Section 3.1), andlandscape heterogeneity (Section 3.2). In this context, the effects of spatial variationare illustrated via the linkage of farm data and model results to the national Danishscale 1:10,000 field block dataset (Kampsax, 1994). In this spatial dataset, all farm

r between the Atlantic and the Continental biogeographical zones of Western Europehe NitroEurope landscape (indicated with a rectangle) and the farm structure for the

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e31923186

fields are part of a unique field-block polygon, i.e. geographical units includingbetween 1 and 10 fields of about 2e5 ha each, demarcated by permanent landscapefeatures such as roads, streams or hedgerows (this is the basis for Fig. 3).

2.3. Study landscape

The data needed for the present exercise were collected from the areasurrounding the Danish NitroEurope study landscape in Midwest Jutland (Fig. 3).The landscape is situated in the River Gudenå Valley, is dominated by non-irrigatedsandy and loamy moraine soils, and has an average annual crop yield of around60 h kg ha�1 for winter wheat, which is the dominant cereal crop. This yield level istypical for this part of Denmark. Around 70% of the total area is farmed, while theremainder is put to other land uses, mainly forestry, urban areas, moorland andwater bodies. The farms in the area include both intensive pig and dairy farms,which account for about half of the agricultural land, and more extensive cash-cropfarms dominated by part-time farmers (Höll et al., 2002).

2.3.1. Farm dataData from 56 representative farms within the study landscape of Fig. 3 were

collected for the year 2002. Based on methods by Kristensen and Kristensen (2004),and in accordance with the official European Statistics and Farm Accountancy DataNetwork methodology (McClintock, 1989; Dalgaard et al., 2002b), the farms weredivided into fourmain farm types (Table 1). The ruminant farms (mainly dairy cattle)had the highest livestock density, expressed by the highest amounts of bothmanureproduced, and manure spread per hectare. The granivore farms (mainly pigs) alsohad a relatively high manure production per ha, but a lower export of manure toother farms. The lowest manure production was on the cash-crop and the mixedfarms, but because of manure imports, some of these farms actually fertilise theirfields with a relatively large amount of livestock manure. Therefore, to distinguishthe farms with almost no manure fertilisation from those with major manure fer-tilisation, the cash-crop farm type was subdivided into farms with more than-, orless/equal to 10 kg N ha�1 yr�1 spread annually as manure (Table 1).

Farm data for the year 2002, in the form of livestock numbers, fertilisationpractices and areas of different crops, were taken from digital farm databases(Dalgaard et al., 2002b). More than 95% of the livestock in the area were cattle orpigs. However, for modelling purposes, ruminants other than cattle were convertedto beef cattle, and granivores other than pigs were converted to finishing pigs. Thisconversion is based on the number of livestock units, where 1 livestock unit (LU)equals 100 kg N yr�1 produced in the form of manure ex store (i.e. after N-lossesfrom housing and manure storages has been deducted the N excreted ex animal).Heifers are not included in the digital register, but were estimated as 1.1 times thenumbers of dairy cows and suckler cows present on each farm. The annual milk yieldper dairy cow was estimated as 8500 kg, corresponding to the average annual milkyield per dairy cow in Denmark. Finally, based on Happe et al. (2010), the numbers ofanimals on each farm were used to define the housing system for each animal type,and thus the distribution of manure types and the application techniques for eachfarmwas derived from the size of the buildings needed to house the herd. In generalwe assumed, that all slurry was spread via band-spreading or trail hoses, and that allmanure was spread in accordance with Danish environmental regulation (TheDanish Plant Directorate, 2007).

2.3.2. Weather and soil dataTemperature and rainfall information needed to estimate N-losses and atmo-

spheric deposition came from a nearby meteorological field station at 56�190 N

Table 1Annual livestock manure production and fertilization practices for the n ¼ 56 farmsstudied within the study landscape in 2002, divided into farm types in accordancewith the official European Statistics and Farm Accountancy Data Network meth-odology (Kristensen and Kristensen, 2004; McClintock, 1989). Standard 95% confi-dence intervals are added with an assumption about a normal distribution of thevariables, which in this case are not area weighted.

Farmtype

Ruminants Granivores Mixed Cash crops Average

Manureda Nomanureb

(n ¼ 15) (n ¼ 11) (n ¼ 10) (n ¼ 15) (n ¼ 5) (n ¼ 56)

kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1

Manure-Nex store

173 � 40 88 � 41 32 � 10 9 � 1 6 � 4 72 � 22

Manure-Nspread

141 � 20 87 � 40 32 � 10 69 � 36 6 � 4 80 � 18

Fertiliser-Nspread

34 � 16 71 � 31 115 � 6 71 � 28 126 � 9 74 � 13

a >10 kg manure-N spread per ha.b �10 kg manure-N spread per ha.

010�070 E, and the same data were used for all farms. Soil data were from theNational Danish scale 1:50,000 Soil Classification Map (Danish Ministry ofAgriculture, 1976). The soil types from this digital soil map were aggregated intotwo main soil classes: sandy soils (�10%(w) of particles �2 mm), and loamy soils(>10%(w) of particles <2 mm). Based on this map, the dominating soil type for eachfarmwas derived so that � according to the Danish Field block classification system(European Commission, 1992) � farms that have more than 50% of their agriculturalland on sandy soils are defined as “farms on sandy soils”, whereas the rest of thefarms are defined as “farms on loamy soils”. In summary, 44% of the cash-crop farms,45% of the ruminant farms and 49% of the granivore farms were on sandy soils. Ingeneral, sandy soils lead to relatively higher N-surplus and N-loss in the form ofNO3

�, whereas loamy soils lead to relatively higher denitrification and soil-N-accumulation compared to the average, but with such small differences in thepresent study, that the soil type is not expected to noteworthy affect the differencesbetween farm types in N-surplus and N-losses (Knudsen et al., 2006).

3. Results

3.1. Farm heterogeneity

3.1.1. Farm type differencesN-surpluses for all 56 farms were calculated for the year 2002

(Table 2), and these show significant differences (p> 0.95) in the N-surplus between livestock farms (ruminants and granivores) andcash-crop and mixed farms. Moreover, the N-surplus on farms withno manure-N application (<10 kg N ha�1 yr�1) was significantlylower (p> 0.95) than both the average farm N-surplus in the wholearea, and the separate N-surplus for all farm types where the fieldswere fertilised with livestock manure (>10 kg N ha�1 yr�1).

Whenwe looked into the different pathways for N-losses, it wasevident that nitrate leaching dominates, both on livestock and non-livestock farms. For ruminant and granivore farms, nitratescontributed between 54% and 62% of the losses excluding changesin the soil-N pool, and for cash-crop farms between 71% and 79% ofthe annual N-loss was in the form of nitrate. The second largest losswas in the form of ammonia volatilisation from livestock housesandmanure storages, even though the purely arable farmsmake nocontribution to this type of loss. Finally, ammonia volatilisation anddenitrification from fields also showed major contributions to thelosses, with cash-crop farms as the highest relative contributor.

It is also worth noting that cash-crop farms with no manureapplication suffer a net mining of the soil nitrogen pools. Incontrast, there is an accumulation of soil-N for all other farm types,especially ruminant farms. These havemost of the permanent cropsand grasslands, as well as the highest N-fixation from grass-cloverfields and the largest N-fertilisation in the form of organic manure.An average of 10 kg N ha�1 yr�1 accumulated in soil pools (Table 2).For an average soil C:N ratio of 20, this roughly corresponds to200 kg fixed C ha�1 yr�1, or 733 CO2-equivalents of greenhouse gasmitigation ha�1 yr�1. In comparison, a contribution of, for example,3 kg N2O ha�1 yr�1 from denitrification leads to an emission of894 CO2-equivalents of greenhouse gas emissions ha�1 yr�1, ata global warming potential of 298 CO2-equivalents per molecule ofN2O (IPCC, 2007). In addition, there is the immediate positive effectfrom N accumulated in the soils, because N accumulated in soilsdoes not readily cascade in the ecosystem, leading to pollutioneffects elsewhere. However, in the long run accumulation of N maylead to an increased risk of N-losses. From results in the presentpaper it is not possible to derive a full, long-term farmGHG balance,but further information on this issue can be found in the morecomprehensive study by Dalgaard et al. (in this issue), taking intoaccount the total C-balance, indirect GHG-emissions from Npollution, and effects of fossil energy use in agriculture.

In the model, both ammonia emission and nitrate leaching aresignificantly higher on intensive livestock farms than on cash-cropfarms with no or little manure application (p > 0.95). This does notcoincidewith a large difference in the area proportion of sandy soils

Table 2N-surplus modelled with Farm-N for the five farm types defined, and partitioned into soil-N accumulation, losses from livestock houses, manure storage facilities and fields.Standard 95% confidence intervals are added with an assumption about a normal distribution of the variables for the year 2002 (not area weighted).

Farm type Ruminants Granivore Mixed Cash crops Average

(manured)a (no manure)b

kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1 kg N ha�1

N-surplus 128 � 13 123 � 14 91 � 7 77 � 11 59 � 1 100 � 8NH3 from house and storage 27 � 6 22 � 12 10 � 4 2 � 0 1 � 1 14 � 4N2/N2O from storagec 3 � 2 1 � 2 2 � 1 1 � 0 0 � 0 2 � 1NH3 from fields 8 � 1 9 � 3 6 � 1 6 � 3 4 � 0 8 � 1N2O from fieldsc 12 � 1 9 � 1 8 � 0 8 � 1 7 � 0 9 � 1NO3

� from fields 59 � 2 70 � 7 62 � 5 48 � 2 48 � 1 58 � 3Soil-N accumulation 19 � 5 10 � 2 3 � 1 10 � 7 �2 � 1 10 � 3

a >10 kg manure-N spread per ha.b <10 kg manure-N spread per ha.c Include all types of denitrification.

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e3192 3187

(on average 49% on granivore and mixed farms, 45% on ruminantfarms and 42e46% on cash-crop farms). However, it is interestingthat not only soil-N accumulation but also ammonia and nitrousoxide emissions from livestock housing andmanure storage tend tobe larger on ruminant than on granivore farms, even though thistrend is not statistically significant. On the other hand, nitrateleaching and ammonia emissions from fields are higher for grani-vore than ruminant farms.

3.1.2. Non-linearities between livestock density and nitrogen lossesTo test for non-linearities between livestock density (i.e. manure

spread) and the different types of N-losses, the modelled farm gateN-surplus, leaching, ammonia emission, denitrification and soilpooling were plotted against the use of livestock manure (Fig. 4).

A clear trend towards higher N-surplus with a higher livestockdensity was noted, and possible non-linearities were tested with

Fig. 4. Annual farm N-surplus, leaching, ammonia emission, denitrification and soil Nchange plotted against the livestock density (measured as the use of livestock manurein kg N ha�1 yr�1 ex store) for the 56 farms studied and modelled with the Farm-Nmodel for the year 2002. The significant exponential correlation between ammoniaemission and livestock density (p > 0.999) is marked with the dotted curve, comparedto the linear relationship between N-surplus and livestock density (full line).

the generalised linear modelling procedure (Dobson, 1990) in thestatistical computing package “R” (The R Foundation forStatistical Computing, 2009). All N-components, except fornitrate leaching, showed significant increases (p > 0.999) withincreased livestock density. Moreover, the same componentsshowed more or less clear tendencies towards non-linear corre-lations, and an accelerating increase in kg N ha�1 yr�1 lost withan increased livestock density. In particular, the relationshipbetween ammonia emission and livestock density showeda significant exponential relationship, y ¼ 7.4 � e0.010x (p > 0.999and R2 ¼ 0.72, compared with a linear regression with R2 ¼ 0.65,and a 2nd order polynomial regression resulting in an R2 ¼ 0.66.Furthermore, the effect of leaving out the “outlier” farm pointwith around 240 kg livestock manure-N ha�1 yr�1 was tested, butstill the exponential correlation showed a significantly higherR2 ¼ 0.71 than the linear regression R2 ¼ 0.61). The exponentialcorrelation was also significant for denitrification (p > 0.999),whereas for N-surplus and soil storage there was only a non-significant tendency for a non-linear relationship with livestockdensity (For N-surplus, the R2-values were 0.64, 0.63 and 0.64 forthe linear-, exponential- and the 2nd order polynomial regres-sions, respectively). Plots between N-surplus and the differenttypes of N-losses showed the same patterns as the livestockdensity plots discussed above.

3.2. Landscape heterogeneity

To explore the spatial variation and geographical pattern of NH3emissions in the landscape of Fig. 3, farm level ammonia emissionswere modelled and mapped for the whole landscape (Fig. 5). Themap highlights the variation, also illustrated in Fig. 4, and showsthe location of source areas with especially high NH3 emissionsfrom agriculture. Fig. 5 also shows a map of Europe indicating theregions with high and low NH3 emissions according to Klaasen(1992). From Fig. 5 we can thus see that the Danish Study land-scape is placed in a regionwith medium to high NH3 emissions, butwith large local variations.

3.3. Upscaling from farm to landscape

The landscape heterogeneity and the non-linear relationshipsbetween livestock density and N-losses may have importantimplications when modelling and upscaling results from the farmto the landscape scale. This can be illustrated by the followingsimple example, where ammonia emission results from the twoupscaling approaches of Fig. 2 are compared: Let us assume a smalllandscape with three equally sized farms of 100 ha each; one farm

Fig. 5. The map to the left shows farm level ammonia emissions modelled by the Farm-N model, and mapped for all farms within and around the Danish NitroEurope studylandscape (indicated with the rectangle). The two maps to the right show in the top the location of the study area in Denmark, and in the bottom the ammonia emission intensity inDenmark compared to the rest of Europe according to Klaasen (1992). Dark areas indicate high ammonia emissions and light areas low emissions.

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e31923188

with no manure applications, one farm with 100 kgmanure-N ha�1 yr�1, and one with an intensive livestock produc-tion and 200 kg manure-N ha�1 yr�1. If we follow approach 1 inFig. 2 and model the expected ammonia emission with the rela-tionship derived above (y ¼ 7.4 � e0.010x), the average ammoniaemission is, as summarised in Table 3, corresponding to8.2 tNH3eN yr�1 for thewhole landscape. However, if approach 2 isfollowed instead, and the ammonia emission is modelled for oneaverage farm with 100 kg manure-N ha�1 yr�1 applied, themodelled ammonia emission is only 20 kg NH3eN ha�1 yr�1, or 26%lower than with approach 1, corresponding to 6.0 t NH3eN yr�1. Incomparison, the summarised NH3 emission, modelled separatelyfor each of the 56 sample farms (approach 1), yields an average of21 kg NH3eN ha�1 yr�1, compared to a 21% lower emission of16.4 kg NH3eN ha�1 yr�1 estimated for the average manureapplication rate of 80 kg N ha�1 yr�1 (approach 2).

Table 3Example of ammonia emissions for a synthetic landscape with three farms,modelled with the derived exponential relation between farm emission and live-stock density (y ¼ 7.4 � e0.010x) and upscaled to the landscape level using the twodifferent approaches of Fig. 2: (1)model calculation for each farm before aggregationand landscape level averaging, and (2) averaging before modelling and landscapelevel aggregation.

Scalingapproach

Farmarea

Farm livestockdensity

Farmemission

Landscapeemission

(ha) (kg manure-N ha�1 yr�1)

(kg NH3eNha�1 yr�1)

(t NH3eN yr�1)

Approach 1Farm 1: 100 0 7.4Farm 2: 100 100 20.1 8.2Farm 3: 100 200 54.7Approach 2Average farm: 300 100 20.1 6.0

4. Discussion

4.1. Effects of farm heterogeneity on modelling N-fluxes and therelated greenhouse gas emissions

The comparison between farm types showed important differ-ences between livestock and non-livestock farms (Table 2). For bothfarm types, leaching is the dominant type of loss, having the largestrelative weight on cash-crop and granivore farms, and the lowestrelative weight on farms with ruminants. In this context it must benoted that Fig. 4 only shows amoderate and insignificant increase innitrate leaching as a function of increasing livestock density on the56 farms studied. This is caused by the fact that the ruminant farmswith the highest livestock densities have higher proportions ofnitrate-efficient grassland and roughage crop areas, compared withthe granivore livestock farms and the othermore extensive livestockand cash-crop farms, which are dominated by less nitrate-efficientcereal crops. Numerous empirical studies confirm such a relation-ship (Simmelsgaard and Djurhuus, 1998). Moreover, the granivorefarms and the extensive livestock farms have a lower rate of soil-N-accumulation as compared to the ruminant farms, or, as in the caseof some cash-crop farms, even significantly mine nitrogen from thesoil, leading to a relatively higher nitrogen leaching compared withthe fertilisation level. This difference is also confirmed by empiricalstudies (Heidmann et al., 2001). In contrast, compared to the live-stock density, the granivore farms tends to have a relatively higherammonia emission than the ruminant farms, mainly caused bydifferences in the housing and manure storage systems.

Ammonia is the second largest type of N-loss, with the highestrelative weight and potential for reduction on granivore farms. Incontrast, denitrification from fields and livestock production hasthe relatively highest weight on ruminant farms (primarily fromcattle production), which also have the highest rate of N-accumu-lation in soils caused by the highest use of organic manures and the

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e3192 3189

highest proportion of permanent grasslands and other forage crops.On average for the whole landscape, this N-accumulation and thecorresponding CO2 carbon fixation partly counterbalances thegreenhouse gas emissions in the form of nitrous oxide, but withlarge variations between farms. However, for cash-crop farmswithout manure applied, the opposite is the case, with a significant(p > 0.95) mining of soil-N, corresponding to an additional emis-sion of greenhouse gases from this process. As mentioned, thisresult was confirmed by Heidmann et al. (2001), who in a largestudy of soil organic matter on Danish farms from 1987e1998found a statistical difference between N-accumulation on livestockfarms, and a depletion of soil stocks on cash-crop farms withoutmanure. However, the uncertainty of such modelling is large, andbecause of the lack of data it is difficult to model such relationshipsfor real landscapes.

For all types of N-losses, except nitrate leaching, there wasa tendency towards non-linear relationships, with relative increasesin losses with increased manure application levels. Within the rangeof livestock densities studied, this non-linearity was found especiallyinteresting in relation to ammonia emissions. In general, the non-linear relationships with relatively higher losses at higher input ratesare a consequence of Justus von Liebig’s classical law of the dimin-ishing return, combined with relations between feeding andmanuremanagement practices; for example via higher protein contents inintensively fertilised grassland crops, and differences in ammoniaemissions between small and large farm housing systems (Happeet al., 2010). Other studies, including farms with higher livestockdensities, show an exponential relation between livestock densityand N-surplus (Dalgaard et al., 2002a), and the accelerating N-leaching with increased manure fertilisation is also a classical rela-tionship (Simmelsgaard and Djurhuus, 1998). However, the newelement in the present study is the integrated modelling approachthat includes relations between both the overall N-surplus, and all itscomponents in the form of emission, leaching, denitrification andchanges in soil N storage. Therefore, the higher proportions of N-efficient roughage crops and grasslands on farms with a high live-stock density (mainly cattle farms) compared to farms with a lowerlivestock density (mainly cash-crop and pig farms) is the majorexplanation for the low increase in N-leachingwith livestock density,and this is an important point to bear in mind when comparing thepresent integrated farmmodelling results with other types of modelresults.

4.2. Effects of landscape heterogeneity

Ammonia emissions modelled for the Danish study landscapeshowed a large heterogeneity and spatial variation (Fig. 5). Thissignificantly affects the total ammonia emission in the landscape,where the non-linear effects of farm heterogeneity exemplified inSection 3.1.2. show that a more homogeneous distribution of thelivestock production activities and the farm ammonia emissionswithin the landscape could help to reduce the total ammoniaemission from farms in the landscape by approximately 20e30%(see Table 3). However, the potential effects also depend very muchon the management practices, and their correlation with farm size,farm structural development, manure and housing systems etc.(Happe et al., 2010).

Some heterogeneity in the distribution of farm ammoniaemissions may be desirable from an environmental protectionpoint of view, to decrease atmospheric concentrations and depo-sition for sensitive ecosystems (Dragosits et al., 2002, 2006). Forexample, high ammonia emissions are revealed in the northeastcorner of the rectangle marking the NitroEurope study landscape inFig. 5. This may be problematic for the nearby protected heathlandarea of Mount Busbjerg, and removal of ammonia emissions, for

example via export of manures from the area, may be an effectiveway of mitigating pollution effects. To investigate such mitigationoptions properly, the agricultural farm management and emissionmodels need to be linked to atmospheric dispersion models (Kroset al., in this issue), as planned in the further NitroScape develop-ment. Such a model may also account for effects of the nitrogencompounds cascading through the ecosystem, and may reveal howtemporal dynamics can lead to pollution swapping, where reduc-tions in one compound (for example ammonia emissions) may leadto increased pollution from other nitrogen compounds (forexample nitrate leaching) (De Vries et al., 2010). Another advantageof such bottom-up modelling, and the inclusion of heterogeneity indata from all farms, is that it enables the exploration of spatialexplicitness and temporal dynamics of nitrogen pollution (Utheset al., 2010). This may have implications when scaling up resultsfrom the farm to the landscape level, and when analysing mitiga-tion options for the reduction of N pollution.

4.3. Effects of upscaling procedures

As illustrated for the case of ammonia, it is important to carefullyevaluate the approaches for modelling and upscaling farm levelemissions for landscape level analyses. In this context it is importantto evaluate the different models implemented to estimate emis-sions, denitrification and soil-N storage, and the non-linearitiesbuilt into these models. Possibly, had we chosen different modelswe would have achieved a different result (for example via thelinear relationships derived in the study of ammonia emissions inThe Netherlands by De Vries et al., 2010). In each case this should betested via a comparison of results from the two different upscalingpathways of Fig. 2. Performing such tests for all possible non-line-arities integrating several different models can be complicated.However, following the general framework of Bierkens et al. (2000)and Dalgaard et al. (2003), it is possible to define schematicprocedures to perform such tests. In the present case, one of themajor reasons for the non-linearity identified between livestockmanure spread and ammonia emissions modelled may be attachedto the exchange ofmanure between farms; i.e. many of the livestockintensive farms export manure to the less livestock intensive farms,but the emissions from livestock houses andmanure storage are notexported, leading to a relatively higher ammonia emission perlivestock unit from the farms exporting manure, compared to thefarms importing manure.

Fig. 5 shows the landscape level variation in farm ammoniaemissions and is a result of farm-by-farm modelling and mappingof averaged farm values. Compared with the Europe-wide NH3emission map of Fig. 5, the landscape level NH3 emission map isvery detailed, but if we zoom in on the individual farms, we realisethat there is still a intra-farm spatial variation, which is not rep-resented on the map. For example the NH3 emissions from housingand storage in Table 2 are in reality point emissions, and the NH3from fields are not uniformly distributed over the different fields ofindividual farms. The investigation of such intra-farm spatialheterogeneity is beyond the scope of the present paper, but will bean interesting subject for future studies (Sutton et al., 2007). Herewe will just conclude, in line with Neumann et al. (2009), that thespatial location of livestock activities is very important for theassessment of N-losses and the related greenhouse gas emissions,and that explicit consideration of scale is important for the results,and the conclusions from the studies. For example, the presentstudy shows how heterogeneity in livestock types, housing andmanure storage systems and exchange of manure between farmsmay induce non-linear relationships between livestock density anddifferent types of N-losses, whereas at the regional or national scalethese non-linearities may be averaged out (De Vries et al., 2010).

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e31923190

In Table 3, the meta-model y¼ 7.4� e0.010xwas used to illustratethe effect of the identified non-linearity between ammonia emis-sions and farm livestock density. The same effect could have beenillustrated with concrete farm examples calculated with the Farm-N model. However, as indicated in Fig. 4, the variation inN-surpluses between farms with the same livestock density can belarge, because of differences in management practices, livestocktypes, housing systems, import/export of manure, crop rotations,soil types etc., and the choice of this farm configuration affect theN-surplus estimated. To avoid this potential bias, we chose to usethe meta-model to demonstrate and estimate the general non-linearity effect. It is important to note, that in both cases, therelatively large import and export of manure between farms in theDanish landscape is one of the major reasons for the non-linearitybetween livestock density and ammonia emissions (the emissionsfrom livestock housing is not exported!). In other landscapes, witha lower export of manure from farms with high livestock densities,a lower degree of non-linearity would therefore be expected.

4.4. Validity of results

The relationships found here between farm N-surpluses andlosses are confirmed by several independent studies (Knudsenet al., 2006; Kristensen and Kristensen, 2004; Dalgaard et al.,1998). This also includes an empirical study of N balances from21 farms within the Danish NitroEurope study landscape for theperiod 1994e1998 (Dalgaard et al., 2002a). At that time, livestockdensities of farms in the landscape varied between 0 and300 kg manure-N ha�1 yr�1, and with a significant (p > 0.999)exponential relationship between livestock density and N-surplus.This is in line with the non-linearity found here between livestockdensity and ammonia emission, and the tendency for a non-linearrelationship between livestock density and N-surplus for livestockdensities between 0 and 240 kg applied manure-N ha�1 yr�1. Onaverage, the modelled N-surpluses are significantly lower(p > 0.95) than the literature values from the 1990s, but this wasto be expected (Hansen et al., 2011) because of the decreasedlivestock density, the decreased variation in farm livestockdensities, and the improvement of the general nitrogen efficiencyin the agricultural production in Denmark (Kronvang et al., 2008),where the updated fertiliser practices and N emission coefficientshave been included in the modelling here.

In the NitroEurope research project, an intensive measure-ment campaign has been carried out in the Danish as well as theother European study landscapes from 2007e2010 (SeeNitroEurope (2009) for further information). This includes thecollection of updated farm nitrogen balances, based on farminterviews, as well as strategically placed measurements ofammonia, nitrous oxide, nitrate and other N- and C-compoundswithin the landscape. These data will soon become available forthe further verification of the results presented here, and thefurther development of the NitroScape model, and it will beinteresting to see how the updated empirical studies of farm N-surpluses corresponds to the comparatively small N-surplusvalues presented in the present study, and to earlier studiescarried out (for example Halberg et al., 1995; Kristensen andKristensen, 1992). In addition, it will be interesting to see thedifference between the agricultural landscapes studiedthroughout Europe, and to investigate observed non-linearitiesand the effects of heterogeneity.

5. Conclusions

In conclusion, the present study illustrates the importance ofaddressing scale issues when modelling nitrogen losses and

greenhouse gas emissions at the landscape scale. A comparison oftwo different approaches for the upscaling of results is needed toidentify significant non-linear effects. If bottom-up modelling of allfarms separately before averaging (approach 1) leads to a signifi-cantly different result than averaging input data before modelling(approach 2), non-linearities may occur and need to be taken intoaccount. In this paper, such non-linearities were exemplified for thecase of ammonia emissions, with significant effects for landscapelevel analyses (p > 0.999), and the induction of pollution swappingwas illustrated (i.e. the mass balance is conserved, and if ammonialosses are reduced, other types of N-losses or soil accumulationincrease). We recommend that future landscape level studies ofmatter fluxes take these findings into account, in order to incor-porate effects of farm and landscape heterogeneity in themodellingof nitrogen losses and greenhouse gas emissions in agriculturallandscapes. In addition, the effect of such non-linearities to beincluded in regional, national and international scale inventoriesfor ammonia emissions, GHG-emissions etc. should be investigatedfurther, including tests of the significant effects of landscape scaleheterogeneity. In this way, new and effective options for the miti-gation of nitrogen pollution can be explored, including effects of thespatial and temporal heterogeneity in real world landscapes, andfor the benefit of the further development of sustainable lowemission farming systems.

Acknowledgements

The authors would like to thank The European Commission andThe Faculty of Agricultural Sciences at Aarhus University for thefinancial support of the to EU-funded research projects of Nitro-Europe (www.NitroEurope.eu) and MEA-scope (www.MEA-scope.org), in relation to which the presented research has been carriedout. Moreover, we thank computer-technician Margit StyrbækJørgensen, Aarhus University, for valuable support in relation to thedevelopment and simulation of the Farm-N model and its interfaceto the landscape databases.

References

Aslyng, H.C., Hansen, S., 1982. Water Balance and Crop Production SimulationWATCROS for Local and Regional Application. Hydrotechnical Laboratory, TheRoyal Vet. and Agric. Univ., Copenhagen, 200 pp.

Bierkens, M.F.P., Finke, P.A., de Willigen, P., 2000. Upscaling and DownscalingMethods for Environmental Research. In: Developments in Plant and SoilSciences, vol. 88. Kluwer Academic Publishers, Dordrecht.

Bobbink, R., Hicks, K., Galloway, J., Spranger, T., Alkemade, R., Ashmore, M.,Bustamante, M., Cinderby, S., Davidson, E., Dentener, F., Emmett, B.,Erisman, J.W., Fenn, M., Gilliam, F., Nordin, A., Pardo, L., De Vries, W., 2010.Global assessment of nitrogen deposition effects on terrestrial plant diversity:a synthesis. Ecological Applications 20 (1), 30e59.

Chirinda, N., Carter, M.S., Alberg, K.R., Ambus, P., Olesen, J.E., Porter, J.R.,Petersen, S.O., 2010. Annual emissions of nitrous oxide from arable organic andconventional cropping systems on two soil types. Agriculture, Ecosystems &Environment 136 (3e4), 199e208.

Dalgaard, T., Olesen, J.E., Petersen, S.O., Petersen, B.M., Jørgensen, U., Kristensen, T.,Hutchings, N.J., Gyldenkærne, S. Hermansen, J.E. Developments in greenhousegas emissions and net energy use in Danish agriculture e how to achievesubstantial CO2 reductions? Environmental Pollution, in this issue.

Dalgaard, T., Hutchings, N.J., Porter, J.R., 2003. Agroecology, scaling and inter-disciplinarity. Agriculture, Ecosystems and Environment 100, 39e51. Reviewpaper.

Dalgaard, T., Heidman, T., Mogensen, L., 2002a. Potential N-losses in three scenariosfor conversion to organic farming in a local area of Denmark. European Journalof Agronomy 16, 207e221.

Dalgaard, T., Rygnestad, H., Jensen, J.D., Larsen, P.E., 2002b. Methods to map andsimulate agricultural activity at the landscape scale. Danish Journal of Geog-raphy 3, 29e39.

Dalgaard, T., Halberg, N., Kristensen, I.S., 1998. Can organic farming help to reduceN-losses? Experiences from Denmark. Nutrient Cycling in Agroecosystems 52,277e287.

Dämmgen, U., Hutchings, N.J., 2008. Emissions of gaseous nitrogen species frommanure management: a new approach. Environmental Pollution 154 (3),488e497.

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e3192 3191

Danish Ministry of Agriculture, 1976. The Danish Soil Classification. TechnicalReview. (in Danish: Den Danske Jordklassificering, teknisk redegørelse). Land-brugsministeriets kontor for jordbundsklassificering, Vejle, Denmark.

Detlefsen, N., Jensen, A.L., 2007. Modelling optimal crop sequences using networkflows. Agricultural Systems 94, 566e572.

De Vries, W., Lesschen, J.P., Oudendag, D.A., Kros, J., Voogd, J.C., Stehfest, E.,Bouwman, A.F., 2010. Impacts of model structure and data aggregation onEuropean wide predictions of nitrogen and green house gas fluxes in responseto changes in livestock, land cover and land management. Journal of IntegrativeEnvironmental Sciences 7 (2), 145e157.

Dobson, A.J., 1990. An Introduction to Generalised Linear Models. Chapman & Hall,London, ISBN 0-412-31110-0. 174 p.

Dragosits, U., Theobald, M.R., Place, C.J., Lord, E., Webb, J., Hill, J., Simon, H.M.,Sutton, M.A., 2002. Ammonia emission, deposition and impact assessment atthe field scale: a case study of sub-grid spatial variability. EnvironmentalPollution 117, 147e158.

Dragosits, U., Theobald, M.R., Place, C.J., Simon, H.M., Sutton, M.A., 2006. Spatialplanning at the landscape level to mitigate the effects of atmospheric ammoniadeposition. Journal of Environmental Science and Policy 9, 626e638.

Duretz, S., Drouet, J.L., Durand, P., Hutchings, N.J., Theobald, M.R., Salmon-Monviola,J. , Dragosits, U., Maury, O., Sutton, M.A., Cellier, P., 2011. NitroScape - a model tointegrate nitrogen transfers and transformations in rural landscapes. Environ-mental Pollution, in this issue.

Erisman, J.W., Bleeker, A., Galloway, J., Sutton, M.S., 2007. Reduced nitrogen inecology and the environment. Review paper. Environmental Pollution 150(2007), 140e149.

European Commission, 1992. Council regulation (ECC) no. 3508/92 of 27 November1992 establishing an integrated administration and control system for certaincommunity aid schemes. Official Journal L 355 (05.12.1992), 1e5.

European Environmental Agency, 2008. Biogeographical Regions, Europe 2008.European Environmental Agency, Copenhagen. http://www.eea.europa.eu/data-and-maps/data/biogeographical-regions-europe-2008.

Galloway, J.N., 1998. The global nitrogen cycle: changes and consequences. Envi-ronmental Pollution 102 (1), 15e24.

Geels, C., Bak, J., Callesen, T., Frohn, L.M., Frydendahl, J., Gyldenkerne, S.,Hansen, A.G., Hutchings, N., Jacobsen, P.P., Schneekloth, M., Winther, S.,Hertel, O., Moseholm, L., 2006 (in Danish: Vejledning om godkendelse af hus-dyrbrug. Faglig rapport fra arbejdsgruppe om ammoniak). Report. Report fromthe Official Danish Working Group on Ammonia, vol. 568. Danish Environ-mental Research Institute, Roskilde. 89.

Gergel, S.E., 2005. Spatial and non-spatial factors: when do they affect landscapeindicators of watershed loading? Landscape Ecology 20, 177e189.

Gundersen, P., Callesen, I., de, Vries W., 1998. Nitrate leaching in forest ecosystems isrelated to forest floor C/N ratios. Environmental Pollution 102 (Suppl.1), 403e407.

Halberg, N., Kristensen, E.S., Kristensen, I.S., 1995. Nitrogen turnover on organic andconventional mixed farms. Biology, Agriculture, Horticulture 14, 25e41.

Hansen, B., Thorling, L., Dalgaard, T., Erlandsen, M., 2011. Trend reversal of nitrate inDanish groundwater e a reflection of agricultural practices and nitrogensurpluses since 1950. Environmental Science and Technology 45 (1), 228e234.

Hansen, B., Kristensen, E.S., Grant, R., Høgh-Jensen, H., Simmelsgaard, S.E., Olesen, J.E.,2000. Nitrogen leaching from conventional versus organic farming systems e

a systems modelling approach. European Journal of Agronomy 13 (1), 65e82.Happe, K., Hutchings, N., Dalgaard, T., Kellerman, K., 2010. Modelling the interac-

tions between regional farming structure, nitrogen losses and environmentalregulation. Agricultural Systems 104 (3), 281e291.

Heidmann, T., Nielsen, J., Olesen, S.E., Christensen, B.T., Østergaard, H.S., 2001.Danish Monitoring Program for N and C-Pool Changes in Arable Soils1987e1998. (in Danish: Ændringer i indhold af kulstof og kvælstof i dyrket jord:Resultater fra Kvadratnettet 1987e1998). DJF Report No. 54. Faculty of Agri-cultural Sciences, Foulum, Denmark.

Hellsten, S., Dragosits, U., Place, C.J., Vieno, M., Dore, A.J., Misselbrook, T.H.,Tang, Y.S., Sutton, M.A., 2008. Modelling the spatial distribution of ammoniaemissions in the UK. Environmental Pollution 154 (3), 370e379.

Höll, A., Andersen, E., Rygnestad, H., Dalgaard, T., Sørensen, E.M., Nilsson, K., 2002.Scenario analyses for cultural landscape development. Danish Journal ofGeography 3, 1e12.

Hutchings, N., Kristensen, I.S., Detlefsen, N., Petersen, B.M., Jørgensen, M.S., 2006.A farm-scale internet-based tool for assessing the effect of intensification onlosses of nitrogen to the environment. In: Technology for Recycling of Manureand Organic Residues in a Whole-farm Perspective, vol. I. Aarhus Universitet,Det Jordbrugsvidenskabelige Fakultet, pp. 187e190. DIAS Report Plant Produc-tion No. 122. For further documentation, see also http://www.fasset.dk/Upload/Fasset/Document/FARM-N_scientific_description.pdf.

Hutchings, N.J., Sommer, S.G., Andersen, J.M., Asman, W.A.H., 2001. A detailedammonia emission inventory for Denmark. Atmospheric Environment 35,1959e1968.

Høgh-Jensen, H., Loges, R., Jensen, E.S., Jørgensen, F.V., Vinther, F.P., 2003. EmpiricalModel for Quantification of Symbiotic Nitrogen Fixation in Leguminous Crops.http://orgprints.org/316/, pp. 1e31.

Høgh-Jensen, H., Loges, R., Jensen, E.S., Jørgensen, F.V., Vinther, F.P., 2004. Empiricalmodel for quantification of symbiotic nitrogen fixation in grass-clover mixtures.Agricultural Systems 82, 181e194.

IPCC, 1997. Greenhouse Gas Inventory Reporting Instructions. Revised 1996Guidelines for National Greenhouse Gas Inventories, vols. 1e3. The Interna-tional Panel on Climate Change (IPCC), United Kingdom.

IPCC, 2007. Climate Change 2007. Intergovernmental Panel on Climate ChangeFourth Assessment Report (AR4). Intergovernmental Panel on Climate Change,Geneva, Switzerland. http://www.ipcc.ch/ 104 p.

Johnson, J.M.F., Franzluebbers, A., Weyers, S.L., Reicosky, D.C., 2007. Agriculturalopportunities to mitigate greenhouse gas emissions. Environmental Pollution150 (2007), 107e124.

Jørgensen, M.S., Detlefsen, N., Hutchings, N., 2005. FarmN: a decision support toolfor managing nitrogen flow at the farm level. In: 5th Conference of the ‘Euro-pean Federation for Information Technology in Agriculture, Food and Envi-ronment. Vila Real, Portugal, pp. 67e73. http://130.226.173.223/farmn/dokumentation/FarmNTool.pdf.

Kampsax, 1994. Report from the Pilot Project DMK Block. (in Danish: Rapportom pilotprojekt DMK block). Ministry of Agriculture and Fisheries,Copenhagen.

Kjeldsen, C., Dalgaard, T., Bøcher, P.K., 2006. Methodological issues of modellingfarm and landscape indicators for sustainable land systems. Danish Journal ofGeography 106 (2), 35e43.

Klaasen, G., 1992. Ammonia Emissions in Europe. Report CP-92e01. InternationalInstitute for Applied Systems Analysis, Laxembourg, Austria.

Knudsen, M.T., Kristensen, I.S., Berntsen, J., Petersen, B.M., Kristensen, E.S., 2006.Estimated N leaching losses for organic and conventional farming in Denmark.Journal of Agricultural Science 144, 135e149.

Kristensen, I.T., Kristensen, I.S., 2004. Farm types as an alternative to detailedmodels in evaluation of agricultural practise in a certain area. In: Brebbia(r-ed.), C.A. (Ed.), Management Information Systems 2004: Incorporating GIS andRemote Sensing. WIT Press, Southampton, England, pp. 241e250.

Kristensen, E.S., Kristensen, I.S., 1992. Analysis of Nitrogen Surplus and Efficiency onOrganic and Conventional Dairy Farms. Report 710. Danish Institute of AnimalSciences, Foulum, pp. 1e54.

Kronvang, B., Andersen, H.E., Børgesen, C.D., Dalgaard, T., Larsen, S.E., Bøgestrand, J.,Blicher-Mathiasen, G., 2008. Effects of policy measures implemented inDenmark on nitrogen pollution of the aquatic environment. EnvironmentalScience and Policy 11 (2), 144e152.

Kros, J., Frumau, K.F.A., Hensen, A. de Vries, W. Integrated analysis of the effects ofagricultural management on environmental quality at landscape scale. Envi-ronmental Pollution, in this issue.

Marshall, B., Crawford, J.W., Porter, J.R., 1998. Variability and scaling: matchingmethods and phenomena. Seminar Series 63. In: Gardingen, P.R., Foody, P.M.,Curran, P.J. (Eds.), Scaling-up from Cell to Landscape. Cambridge UniversityPress, Cambridge, pp. 253e272.

Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009. Greenhouse gasproduction and efficiency of planted and artificially aerated constructedwetlands. Environmental Pollution 157 (3), 748e754.

McClintock, J., 1989. Farm Accountancy Data Network. An A to Z Methodology.Report, first ed.. Commission of The European Communities, Brussels.

Neumann, K., Elbersen, B.S., Verburg, P.H., Staritsky, I., Perez-Soba, M., de Vries, W.,Rienks, W.A., 2009. Modelling the spatial distribution of livestock in Europe.Landscape Ecology 24, 1207e1222.

NitroEurope, 2010. Official Homepage for the NitroEurope Integrated Project: theNitrogen Cycle and Its Influence on the European Greenhouse Gas Balance.Centre for Ecology and Hydrology, Edinburgh, Scotland. www.NitroEurope.eu.

NitroEurope, 2009. TV Reports from the NitroEurope Measurement Campaignwithin the Danish Study Landscape April 20, 2009. Live Report from the StudyArea: http://djfgeo.dk/?p¼67, and a Following Evening News Report: http://djfgeo.dk/?p¼70, Danish Television, Channel 2, TV Midt-Vest, Holstebro,Denmark.

Petersen, B.M., Olesen, J.E., Heidmann, T., 2002. A flexible tool for simulation of soilcarbon turnover. Ecological Modelling 151, 1e14.

Petersen, B.M., 2007. Modelling Organic Matter Turnover in Agricultural Soils, Ph.D.Report. Aarhus Universitet, Det Jordbrugsvidenskabelige Fakultet, AarhusUniversitet.

Poulsen, H.D., Børsting, C.F., Rom, H.B., Sommer, S.G., 2001. Nitrogen, Phosphateand Potassium in Livestock Manure. 2000 Standard Values for FarmManure. Livestock Report No. 36. Danish Institute of Agricultural Sciences,Denmark. 152.

Simmelsgaard, S.E., Djurhuus, J., 1998. An empirical model for estimating nitrateleaching as affected by crop type and the long-term N fertiliser rate. Soil Useand Management 14, 37e43.

Sonneveld, M.P.W., Brus, D.J., Roelsma, J., 2010. Validation of regression models fornitrate concentrations in the upper groundwater in sandy soils. EnvironmentalPollution 158 (1), 92e97.

Sutton, M.A., Nemitz, E., Erisman, J.W., Beier, C., Butterbach Bahl, K., Cellier, P., deVries, W., Cotrufo, F., Skiba, U., Di Marco, C., Jones, S., Laville, P., Soussana, J.F.,Loubet, B., Twigg, M., Famulari, D., Whitehead, J., Gallagher, M.W., Neftel, A.,Flechard, C.R., Herrmann, B., Calanca, P.L., Schjoerring, J.K., Daemmgen, U.,Horvath, L., Tang, Y.S., Emmett, B.A., Tietema, A., Peñuelas, J., Kesik, M.,Brueggemann, N., Pilegaard, K., Vesala, T., Campbell, C.L., Olesen, J.E.,Dragosits, U., Theobald, M.R., Levy, P., Mobbs, D.C., Milne, R., Viovy, N.,Vuichard, N., Smith, J.U., Smith, P., Bergamaschi, P., Fowler, D., Reis, S., 2007.Challenges in quantifying biosphere-atmosphere exchange of nitrogen species.Environmental Pollution 150, 125e139.

The Danish Plant Directorate, 2007. Fertilization Manual for Danish Agriculture.(In Danish: Vejledning om gødsknings- og harmoniregler). The DanishMinistry for Food, Agriculture and Fisheries, Copenhagen, ISBN 87-91429-21-8. 108 p.

T. Dalgaard et al. / Environmental Pollution 159 (2011) 3183e31923192

The R Foundation for Statistical Computing, 2009. R Version 2.10.0 (2009.10.26).Institute for Statistics and Mathematics, Wien, ISBN 3-900051-07-0. www.r-project.org.

Uthes, S., Piorr, A., Zander, P., Kedziorac, A., Ungaro, F., Dalgaard, T., Kjeldsen, C.,Stolze, M., Moschitz, H., Schader, C., Happe, K., Sahrbacher, A., Damgaard, M.,Sattler, C., Toussainta, V., Casini, L., Müller, C., 2010. Regional impacts of abol-ishing direct payments: an integrated analysis in four European regions. Agri-cultural Systems 104 (2), 110e121.

Vinther, F.P., Hansen, S., 2004. SimDen e a Simple Model to Quantify N2O EmissionDenitrification. (DJF markbrug (in Danish)). Report No. 204. Det Jord-brugsvidenskabelige Fakultet, Aarhus Universitet, Denmark.

Zander, P., Uthes, S., Sattler, C., Reinhardt, F.J., Happe, K., Damgaard, M.,Sahrbacher, A., Dalgaard, T., Hutchings, N., Kjeldsen, C., Detlefsen, N.,Iversen, B.V., Ndah, T.H., 2009. The MEA-scope modelling approach. In: Piorr, A.,Müller, K (Eds.), Rural Landscapes and Agricultural Policies in Europe. SpringerVerlag, pp. 101e122.