Quantifying the production and consumption of fruit and vegetable crops globally
We focus this work on fruit and vegetable consumption in the UK, India, and South Africa. These countries are not meeting their fruit and vegetable consumption guidelines, as is the case for many countries, globally. Only 28% of adults in England had their ‘5 a day’ in 201875, the average per person intake was 124 grams for caregivers of children averaging 99 grams in a peri-urban area of South Africa in 201176, and 160 grams was the average per capita intake per household in India in 2011-201277.
Our analysis is global in scope as we measure the pressure being placed on biodiversity by fruits and vegetables consumed in the UK, India, and South Africa and compare pressures associated with different international trade partnerships, as well as domestic production. The crop data we use78,79 are freely available, so our findings can, in future, be replicated in other countries and extended for other purposes in the focal countries.
We use Monfreda et. al.78,79 crop harvested area and yield data from circa 2000 because these include 50 individual fruit and vegetable crops also recorded in the trade data we used and are spatialised for these crops on a global scale using subnational-scale survey and gridded remote-sensing (satellite) data. These data also have a similar timeframe to available biodiversity data (described below). The original Monfreda et al.78,79 dataset includes 61 fruit and vegetable crops, but this number includes 6 groups of mixed fruits and vegetables (unspecified ‘berries’, ‘citrus fruit’, ‘fresh fruit’, ‘stone fruit’, ‘tropical fruit’, and ‘vegetables’). We further exclude 5 crops (green onions, ‘melonetc’ (other melons, including canteloupes), peppers, cashew apples, and string beans) which are not consumed by the focal countries and/or do not have associated records in the trade data described below.
The data we used mean that our results cannot capture land-use changes that have taken place in the last 15–20 years, but indicate which countries and regions are under greater relative biodiversity pressure than others, and which crops and trade relationships are driving this. Other freely available, spatialised and more recent data are not available for a variety of fruits and vegetables. For instance, SPAM80 has more recent global data (2010), but only covers four specific fruits and vegetables and three groups - tropical fruit, temperate fruit, and vegetables. Monfreda et. al.78,79 data are available for 31 fruits and 24 vegetables, as well as grouped categories we exclude from our analysis due to their non-specificity. The resolution of the Monfreda et. al.78,79 data selected is 5 arc minutes (about 9 km at the equator), and the data cover the period 1997–2003. We used the R81 ‘raster’ package82 to ensure that the extent and projection of the crop and biodiversity data were consistent, conducting bilinear resampling of the crop data to the 5 arcmin (~ 10km at the equator) resolution species data. Production data are mapped in Extended Data Fig. 1, for reference.
Dry-weight production values were used, calculated for each fruit and vegetable analysed using water-content averages from food composition data from the USDA National Nutrient Database for Standard Reference83 (averages used as given in Table S2). Dry-weight production values were calculated by multiplying the raster for each fruit and vegetable by 1-water content(grams)/100.
In our work, consumption is defined as production (tonnes) plus imports (tonnes) minus exports (tonnes), also commonly termed ‘supply’. We processed trade data from the FAO to measure consumption. To determine the overall movement of food and agricultural products, processed foods were converted into primary crop equivalents, as in Dalin et al.67, based on FAO product categorization. FAO production and trade values were adjusted using a balancing algorithm, to link final demand to origin of primary product, created by Kastner et al.67,84. This algorithm converts secondary products, derived from primary products, into their primary equivalents using data on: primary product production and bilateral trade; secondary products derived from these primary products; calorie contents; and extraction rates67,84,85.
For our analyses, we used average traded tonnes of primary equivalent crops (metric tonnes, 1,000 kilograms) for 1997–2003 (to match the crop data period represented by Monfreda et. al.78,79). We also assessed changes in trade since 2003 (up to 2017, which was the most recent year for which country trade data were reliable and complete) to produce Fig. 5.
Mapping species ranges and estimating vertebrate richness
We represent biodiversity using species richness - the most readily available biodiversity metric globally for the largest number of vertebrate species. In this study, we focus on vertebrate (mammal, bird, amphibian, and reptile) species richness, as these are the best sampled taxonomic groups globally and have been the focus of studies which have demonstrated significant negative relationships between human land uses (including cropland) and species richness, which is a key assumption of our work. We use 10-km resolution, global-scale, spatialised species-richness estimates from Etard et al.50,87,88 comprising species records contributed in circa 2012. These maps were produced by summing species-range distribution maps from BirdLife International87 and the IUCN88. These organisations also provide the information to facilitate the exclusion of areas outside of known elevational limits for species, carried out for these data. Using range maps to estimate species richness is appropriate for representing the broad distribution of species across the globe and within a given country, despite known limitations89 (e.g., assuming that a species is present across its entire range and excluding species which have become extinct since ranges were last recorded). Vertebrate richness data are shown in Extended Data Fig. 2, for reference.
Our work makes the necessary assumption that the production of fruits and vegetables places a pressure on vertebrate richness, as our new biodiversity pressure indicators (described next) quantify the overlap between cultivated areas and species ranges. This assumption is also made45 and tested8 in previously published studies.
Measuring the biodiversity pressure of fruits and vegetables by country and crop
We created a new measure of ‘affected species range’ (species.hectares (ha)) to encapsulate the actual area occupied by specific crops in a country and the number of species in this area potentially being put under pressure by this. This is a per grid-cell and per crop measure, rather than per tonne, so it reflects the overall biodiversity pressured area associated with a given crop.
The affected species range metric, ‘ASR’, (unit: species.ha) is defined as follows:
Equation 1:\(ASR = {HA}_{c,l} \times {SR}_{l}\)
Where HAc,l is the harvested area of crop ‘c’ at location (i.e. 5 arcmin grid cell) ‘l’ (average number of hectares harvested per land-area of a grid cell, from Monfreda et. al.78,79; cell area = 0.0833332 degrees2, or ~ 1,000 hectares at the equator) and SRl is species richness at location ‘l’ (number of species, computed as described in ‘Mapping species ranges and estimating vertebrate richness’).
We mapped the ASR in each grid cell using the ‘raster’ package82 in R81, enabling us to identify pressure hotspots associated with fruits and vegetables being grown in our focal countries and their trade partner countries. We also quantified the weighted-mean ASR (weighted by dry-weight production in tonnes) for each fruit and vegetable for each country in the world (except Greenland and Antarctica) using the ‘exactextract’ package90 in R81.
A new measure of the pressure placed on biodiversity per tonne of a given crop
We developed a new ‘biodiversity pressure’ intensity metric (‘BP’) which builds on the affected species range to capture the number of vertebrate species (or species richness) potentially impacted by cropland per hectare per tonne of a given crop. This enables us to compare crops, considering their pressures according to global harvests in general, rather than specific pressures placed by a given country’s consumption. We designed this metric to be fully scalable but also relativised to the amount of crop being produced in the area for which it is quantified. This is important as we are assessing the trade-off between wild animal health and human health; a greater amount of fruit and vegetable produced implies more potential benefit to human health by supporting a supply of healthy foods, while a large extent of cropland in a biodiverse area will put more pressure on wild animal health. Accordingly, a fruit or vegetable that puts a high pressure on biodiversity (high ASR) but provides nutrition for many people (high production value) would have a lower BP value than one which has the same ASR but is associated with low levels of food production.
The biodiversity pressure metric, ‘BP’, at the grid-cell level (unit: species.hectares (ha) per tonne) is defined as follows:
Equation 2:\({BP}_{c, l}= \frac{\left({HA}_{c,l} \times {SR}_{l}\right)}{{P}_{c,l}}\)
where HAc,l is the harvested area of crop ‘c’ at location ‘l’ (average number of hectares harvested per land-area of a grid cell, from Monfreda et. al.78,79), SRl is species richness at location ‘l’ (number of species, computed as described in ‘Mapping species ranges and estimating vertebrate richness’), and Pc,l is the production of crop ‘c’ at location ‘l’ in tonnes. Production was calculated using the product of yield (tonnes/hectare) and harvested area (hectares) at location ‘l’, both of which are datasets available at the same resolution and for the same crops with global, spatialised coverage via Monfreda et. al.78,79.
As well as mapping BP per grid cell using the ‘raster’ R package81,82, we quantified the average pressure for each fruit and vegetable for each country in the world (except Greenland and Antarctica) using the ‘exactextract’ package in R81,90. The average of BP (weighted by dry-weight production) across all grid cells in each country, was computed as follows:
Equation 3:\({BP}_{c,i}= \frac{\sum _{l in i}{P}_{c,l} \times {BP}_{c,l}}{\sum _{l in i}{P}_{c,l}}\)
where BPc,i is biodiversity pressure (species.ha/tonne) of crop ‘c’ in country ‘i’ and Pc,l is the dry-weight production of crop ‘c’ at location ‘l’ in tonnes, so this equation sums over cells ‘l’ in country ‘i’ to get the country-level, weighted-average BP, presented in figures summarising imports (Figs. 2 and 3, panel b). The weighting means that, when combining with international trade data (Eq. 5), we consider that the origins of exported crops are distributed in the same way across grid cells as production in a country (i.e. exports are more likely to come from locations where production is the largest).
The production-based biodiversity pressure (‘BPprod’) sums biodiversity pressure over a country, as follows:
Equation 4:\({BPprod}_{c,i}={BP}_{c,i}\times {P}_{c,i}\)
The BPprod is presented in Figs. 2 and 3 (panel b) for each focal country’s domestic component of BP.
Measuring the consumption-based biodiversity pressures of fruits and vegetables
Finally, we developed a consumption-based measure of biodiversity pressure, ‘BPcons’, to represent the pressure on biodiversity embedded in the consumption of a country (here: the UK, India, and South Africa). BPcons accounts for pressures associated with a country’s consumption, made up of domestically produced and imported food.
BPcons is calculated as follows:
Equation 5:\({BPcons}_{fc, c} = {C}_{fc, c}^{dom} \times {BP}_{fc, c}+ \sum _{i}{T}_{i, fc, c} \times {BP}_{i,c}\)
where: BPcons is the consumption-based biodiversity pressure of focal country ‘fc’ and crop ‘c’ (unit: species.hectares (ha)); Cdomfc,c is the local, domestic supply of the crop ‘c’ to the focal country ‘fc’ (tonnes; production - exports); Ti,fc,c is the import of crop ‘c’ from country ‘i’ to ‘fc’ (tonnes ; where ‘i’ is an import partner); and BPi,c is the average biodiversity pressure computed at the country level as described in Eq. 3.
By providing multiple measures of biodiversity pressure, and data on consumption and trade, we can start to see whether a high pressure is due to high biodiversity in a country overall, large amounts of a crop being produced/consumed, or because the risk of a crop is particularly high because of where it is grown within a country, relative to species ranges within/across countries.