Increased nitrogen enrichment and shifted patterns in the world’s grassland: 1860–2016

. Production and application to soils of manure excreta from livestock farming signiﬁcantly perturb the global nutrient balance and result in signiﬁcant greenhouse gas emissions that warm the earth’s climate. De-spite much attention paid to synthetic nitrogen (N) fertilizer and manure N applications to croplands, spatially explicit, continuous time-series datasets


55
While the availability of national-level statistics is a fundamental component of our knowledge 56 base, environmental problems related to nitrogen pollution or emissions are best tackled at local 57 scale and often require finer, geo-spatial information, for example to assess proximity to water 58 bodies and thus pollution risks. In particular, a number of studies have focused on downscaling 59 existing national information to develop geospatially explicit regional and global datasets of referred to in the literature as 'pastures and rangelands'. We note that 'grassland' is in fact a land 78 cover definition. In order to avoid the confusion often made in the literature between land cover 79 and land use terminology, we will adopt FAO land use terminology of 'permanent meadows and 80 pastures,' to which the various national regional and global land use statistics cited in this work 81 refer. Furthermore, using results from the HYDE 3.2 dataset (Klein Goldewijk, 2017), we may 82 split the FAO land use category into 'pastures' and 'rangelands', to highlight differences 83 between managed intensive and unmanaged extensive systems, as needed. To enhance our 84 understanding of the role of livestock on the global GHG balance and nutrient budgets (e.g., 85 ammonia emissions, nitrate leaching), global biogeochemistry models require spatially explicit 86 estimates of N inputs. In this study, we developed datasets for major sources of N inputs in 87 agriculture (i.e., manure and fertilizer application and manure deposition on permanent meadows 88 and pastures), using the recently published FAOSTAT statistics on manure N use in agriculture 89 (FAOSTAT, 2018). The latter are estimates based on IPCC Tier 1 methodology, i.e., they rely on 90 default coefficients prescribing, among other variables, N excretion rates by animal type and 91 region, as well as regional compositions of manure management systems (FAOSTAT, 2018).

92
Through combining the land-use dataset HYDE 3.2, FAOSTAT fertilizers N statistics, and 93 gridded manure production data in Zhang et al. (2017), we developed three annual global 94 datasets at a spatial resolution of 0.5° × 0.5°, as follows: 1) manure N application rates to The concepts of grassland, pastures and meadows span several international land cover and land 105 use statistical definitions, specifically those used by FAO (FAOSTAT, 2018). In this paper, we 106 follow the relevant FAO land use definition of 'permanent meadows and pastures,' considering 107 our focus on livestock production. Importantly, complete country, regional and global statistics 108 available from FAO refer to this land use category. This land use definition is roughly equivalent 109 to the one adopted by the academic community engaged in global biogeochemical modeling, for  For mineral and chemical fertilizers, we further split the FAO definition using HYDE 3.2, into 115 'pastures' and 'rangelands,' the former representing land use areas managed to support high 116 stocking densities of grass production for hay/silage, whereas the latter represents unmanaged   We finally spatialized the pasture N data using HYDE 3.2, obtaining gridded maps of synthetic 134 fertilizers N application rates on pastures in each grid cell area, over the period 1961−2016 ( Fig.   135 1). We assumed even application rates within each country. Although gridded livestock density 136 maps were available from FAO, these are currently fixed for specific time periods, mainly 2010, 137 so that we deemed their use not particularly relevant to improve estimates for the 1961-2016 138 time series considered herein. Improved live density map products from FAO will considerably 139 improve our work and reduce uncertainty, and will be used when available.  Furthermore, the FAOSTAT data do not separate manure application to cropland and pastures 145 and data of manure N application rates to pastures are currently not available. We therefore 146 assumed that manure N application rates in pastures and croplands were the same, considering 147 that the overall uncertainty in the input manure N data would not justify further assumptions at 148 this stage of knowledge. Improved FAO statistics on both use and application rates will be used 149 when available to improve this current work. Through combining land-use data HYDE 3.2, we 150 calculated the total cropland and pasture areas within each country where manure application 151 amount was larger than zero. We then computed mean manure N application rates on pastures, 152 annually over the period 1961−2016 (Fig. 2). 153 We calculated the national-level ratio of manure application to production ( 2 , ) by

160
The above-mentioned processes are represented by following equations: (1) 162 where year is from 1961 to 2016, and country number is 165. 2 , is the ratio (unitless) of 163 manure application to production in the year y and country j. , is the national total manure 164 application amount (kg N yr -1 ) derived from the FAO database for each year. is the area of 165 each grid (km 2 ).
is the gridded manure application rate (kg N km -2 yr -1 ) in year y and country j. 168 As the national-level manure application amount was not available during 1860−1960, we 169 assumed that 2 , is the same as for 1961. Combining with the gridded spatial maps of

177
The adjustment procedure is represented in the following equations: where year is from 1860−2016. , (kg N yr -1 ) is the calculated national-level manure 180 application amounts in the year y and country j. If , is less or more than , , an 181 adjustment is needed to keep calculated national total amounts consistent with amounts from the 182 FAOSTAT database. In this case, , is less than , using Eq. 3, thus an adjustment 183 is needed, using the following equations: 185 where , is the regulation ratio (unitless) in the year y and country j.  (Fig. 2).

202
The above-mentioned processes are represented by the following equations: where year (y) is from 1961 to 2016 and country number (j) is 157. 2 , is the ratio (unitless) 205 of manure deposition to production in the year y and country j. , is national total manure 206 deposition amount (kg N yr -1 ) derived from the FAOSTAT database for each year.
, is 207 the gridded manure N production rate (kg N km -2 yr -1 ) in the year y and grid g.
209 where , is the gridded manure deposition rate (kg N km -2 yr -1 ) in the year y and country j. 210 Finally, we calculated the manure deposition amount for each country through combining 211 , and grid area to compare with the national-level deposition amounts from the 212 FAOSTAT database, using the following equation: In contrast, Europe's synthetic N fertilizer use and contribution to the global total decreased 236 since the 1980s (Fig. 5a). This is a well-known trend, linked to EU-wide policy directives aimed Our results showed that the annual manure N application rates on pastures increased from 1.4 to 259 8.6 Tg N yr -1 during 1860−2016 (Fig. 3a). Manure N application rates showed rapid increases 260 across the globe and exhibited large spatial variations, shifting the regional use from North

261
America and Europe to Asia, during the study period (Figs. 4d-f). The global average manure 262 application rate was 5.3 kg N ha -1 yr -1 in the 1860s and roughly doubled by 2016 (10.7 kg N ha -1 263 yr -1 ) ( Table 1).
264 From the regional perspective (Fig. 5b) Our data show that the total amounts of manure N deposited on pastures and rangelands 292 increased from 14 to 84 Tg N yr -1 during 1860−2016 (Fig. 3b). Manure N deposition rates 293 increased steeply across the globe, but exhibited large spatial variations during the study period 294 (Fig. 4g-i). The increase was much larger in the eastern world (typically China and India) and 295 South America compared to the western world. The global average manure deposition rate was 296 11 kg N ha -1 yr -1 in 1860 and reached 25 kg N ha -1 yr -1 in 2016 (Table 1).

313
The significant contrast of changes in manure N deposition rates in Oceania and South America 314 between the 1860s and the 1960s is due to the substantial and rapid increase of grassland areas   We compared our datasets with other existing data sources ( grasslands into mixed and pastoral systems, and estimated grasslands area based on the country-366 or regional-level grazing intensity (Table 2). In addition, synthetic fertilizers were applied to the 367 area of mixed agricultural systems (grassland and cropland) and manure N was assumed to be  (Table 1). This study obtained country-level N fertilizer amounts applied to pastures from the   hotspots of global N fertilizer application in 1961 (Fig. 4b). However, these hotspots have shifted 406 from Western Europe towards southern Asia at the end of the 20th century (Fig. 4c). Southern  Germany (Fig. 4d), but by 2016, the highest application was in the North China Plain (Fig. 4f).   In the context of increasing livestock production, manure and fertilizer N inputs to permanent transformation of environmental problems. In this study, we have obtained N data from various 505 sources to fill the data gap; however, large uncertainties still remain in our datasets (e.g., N 506 application rate within each country, annual manure application amounts). More information is 507 needed to improve these datasets in our further work.

512
We are grateful to FAO and its member countries for the collection, analysis and dissemination 513 of fertilizers and land use statistics. We thank Dr. Wilfried Winiwarter in International Institute 514 for Applied Systems Analysis for constructive comments that have helped improve this study.