Global and country-level data of the biodiversity footprints of 175 crops and pasture

The destruction of natural habitat for cropland and pasture represents a major threat to global biodiversity. Despite widespread societal concern about biodiversity loss associated with food production, consumer access to quantitative estimates of the impact of crop production on the world's species has been very limited compared to assessments of other environmental variables such as greenhouse gas emissions or water use. Here, we present a consistent dataset of the biodiversity footprints of pasture and 175 crops at the global and national level. The data were generated by combining maps of the global distribution of agricultural areas in the year 2000 with spatially explicit estimates of the biodiversity loss associated with the conversion of natural habitat to farmland. Estimates were derived for three common alternative measures of biodiversity – species richness, threatened species richness, and range rarity – of the world's mammals, birds, and amphibians. Our dataset provides important quantitative information for food consumers and policy makers, allowing them to take evidence-based decisions to reduce the biodiversity footprint of global food production.


a b s t r a c t
The destruction of natural habitat for cropland and pasture represents a major threat to global biodiversity. Despite widespread societal concern about biodiversity loss associated with food production, consumer access to quantitative estimates of the impact of crop production on the world's species has been very limited compared to assessments of other environmental variables such as greenhouse gas emissions or water use. Here, we present a consistent dataset of the biodiversity footprints of pasture and 175 crops at the global and national level. The data were generated by combining maps of the global distribution of agricultural areas in the year 20 0 0 with spatially explicit estimates of the biodiversity loss associated with the conversion of natural habitat to farmland. Estimates were derived for three common alternative measures of biodiversity -species richness, threatened species richness, and range rarity -of the world's mammals, birds, and amphibians. Our dataset provides important quantitative information for food consumers and policy makers, allowing them to take evidence-based decisions to reduce the biodiversity footprint of global food production.  Table   Subject Nature and Landscape Conservation Specific subject area Biodiversity impacts of global food production Type of data Table  How data were acquired The data were generated by overlaying species distribution maps of mammals, birds, and amphibians with global maps of agricultural areas and yields Data format Analysed Parameters for data collection All species and crops available in the primary data sources were included. Description of data collection The data were generated based on existing species distribution and agricultural datasets Data source location Primary data sources: Species-specific range distribution and habitat data of birds [2] , and mammal and amphibians [5] , available from: http://datazone.birdlife.org/species/requestdis and https://www.iucnredlist.org , respectively. Global croplands and pastures [10] , and crop-specific harvested areas [8]

Value of the Data
• Despite widespread consumer concern about the biodiversity footprint of global food production, consistent estimates of the impacts of different crops on the world's species are not available. Our dataset fills this gap by providing global and country-level impact estimates based on three common biodiversity measures. • Our data provide important quantitative evidence to inform the decision-making of conservationists, policy makers, and food consumers, aiming to reduce food-related biodiversity impacts. • Our estimates can be combined with crop-specific nutrient level data to rank crops and producing countries according to how efficiently they produce nutrients relative to the biodiversity footprint. This can facilitate the development of biodiversity-related food labelling systems, enabling consumers to reduce personal impacts and encourage shifts towards a sustainable food production. • Our data also open the space for in-depth analyses of the heterogeneity in the local biodiversity footprint of specific crops within and across producing countries. A better understanding of these patterns will help inform where the future expansion of growing areas of specific crops should be prioritised in order to minimise biodiversity impacts.

Data Description
Our dataset ( Table 1 ) contains the distributions of local biodiversity footprints across pastures and the harvested areas of 175 crops in the year 20 0 0, based on three different biodiversity measures -species richness, threatened species richness, and range rarity -, at the global (cf. Fig. 4 ) and national level. Distributions are characterised in terms of 5th-95th distributional percentiles. In particular, for any given country and crop, the median value (50th percentile) of these distributions provides an estimate of the average number of species, average number of threatened species, and the average range rarity, that has been lost, compared to the scenario of natural habitat, on the growing areas of the crop of interest in the country of interest. Analogous data are provided for the distributions of the ratio of local biodiversity footprints to local crop yields. These data are relevant, in particular, when linked to crop-specific nutritional data: multiplying them with by amount of a given nutrient in one unit of crop produce makes it possible to rank different crops according to how efficiently they provide the given nutrient relative to their biodiversity footprint.
In addition to the distributional data, our dataset contains spatially aggregated measures of the total biodiversity footprint of each crop in each country. These data make it possible, for example, to compare the contribution of specific crops or countries to the global biodiversity footprint of agriculture (cf. Fig. 5 ).
The three types of data are available as .xlsx files on the Figshare repository ( https://figshare. com/s/0695ac6fa97892225687 ).

Local biodiversity footprints of crops and pasture
We used the method previously described by [6] and [1] to estimate the geographical distribution of all known mammals, birds and amphibians under four different land cover types: natural vegetation, arable land, plantation, and pasture. In the following, we summarise the approach. We used species-specific extents of occurrence of mammals, birds, and amphibians [2 , 5] , which we rasterised from their original spatial polygon format to a 5-arc-minute grid (~10 km at the equator). These data represent spatial envelopes of species' maximum geographic ranges, and do not account for the distribution of natural or artificial land cover within these areas. Extents of occurrence were refined by incorporating species-specific habitat preferences [2 , 5] , which include one or more biome categories in which each species is known to occur. In each grid cell contained within a given species' extent of occurrence, the species was estimated as being present under natural vegetation if its list of habitat categories contained the local potential natural vegetation type, for which we used a 5-arc-minute global map [11] . In the same  way, a species was estimated as being able to occur in a grid cell under arable land, plantation or pasture land cover, if its list of habitat categories included the relevant one of these three IUCN artificial land cover categories.
For the case of natural vegetation and for each of the three artificial land cover categories, maps of species richness, threatened species richness, and range rarity ( Figs. 1 -3 ) were derived as follows. Local species richness in a given grid cell is given by the number of species estimated as being present in the grid cell under the relevant land cover type. Threatened species richness was obtained in the same way but included only species whose Red List status is vulnerable, endangered, or critically endangered [2 , 5] . Range rarity [3] in a grid cell was calculated as the sum of the inverse natural range sizes of all species present in the cell under the relevant land cover. Thus, for this measure, species with a narrow geographic ranges are weighted more heavily than geographically widespread species [3] .
Finally, for a given biodiversity measure m (representing species richness, threatened species richness, or range rarity) and a grid cell x , the local biodiversity footprint β m c (x ) associated with a crop c was defined as the difference between the local potential natural biodiversity and the biodiversity for the land cover type corresponding to the crop (either 'arable land' or 'plantation', as defined by [4] ). Analogously, the local biodiversity footprint β m Pasture (x ) associated with pasture was defined as the difference between local potential natural biodiversity and biodiversity under pasture.

Crop-and pasture-specific distributions of biodiversity footprints at country level
Based on the derived biodiversity footprint maps ( Figs. 1 -3 ), we first determined the distribution of biodiversity footprints across the harvested areas of each crop c. We denote by H c (x ) the harvested area (in ha) in the year 20 0 0 of a crop c in a 5-arc-minute grid cell x , which is available for 175 crops [8] . These maps of harvested areas represent the latest consistent global dataset containing all crops included in our dataset. More recent maps, based on different methods of spatial allocation of cropland, are available only for a much smaller number of crops; for consistency, we did not include these here, but used the 20 0 0 data throughout our approach. We  Crop-specific percentiles of biodiversity footprints across the harvested areas of a crop c in a specific country i , denoted b m c, i (p) , were computed in the same way, but weights corresponding to grid cells located outside the country's borders [9] were set to zero.
Percentiles of biodiversity footprints across global pastures, b m Pasture , World (p) , were computed analogously, based on the set of local biodiversity footprints of pasture { β m Pasture (x ) } x and weights Percentiles of biodiversity footprints across pastures in a given country i , denoted b m were again computed by setting non-relevant weights to zero.

Crop-specific distributions of biodiversity footprints per unit yield at country level
In addition to estimating the distributions of local biodiversity footprints on agricultural areas, we used crop-specific global maps of fresh-weight yields (in Mg ha −1 year −1 ) for the year 20 0 0 [8]

Crop-specific total biodiversity footprints at country level
Thus far, we considered the distributions of (absolute and per-yield) local biodiversity footprints across agricultural areas. In addition, we derived spatially aggregated estimates of the biodiversity footprints of crops and pastures at country level by integrating local footprints over the relevant areas. For pasture, we used the physical area (in ha) covered by pasture in a grid cell x in the year 20 0 0 [10] , A Pasture (x ) , to define the spatially-aggregted global biodiversity footprint of pasture as In the case of crops, we used the physical area (in ha) covered by cropland in a grid cell x in the year 20 0 0 [10] , A Cropland (x ) , and the harvested area (in ha) in the year 20 0 0 of a crop c in a grid cell x [8] , H c (x ) . Generally, A Cropland (x ) is not the same as c H c (x ) , the sum of all local harvested areas. This is because H c (x ) does not represent physical area but harvested area which increases if the crop is harvested multiple times per year [8] . We accounted for this by calculating the spatially-aggregted global biodiversity footprint of a crop c as

Declaration of Competing Interest
The authors declare no conflict of interest.