Exclusion of agricultural lands in spatial conservation prioritization strategies: consequences for biodiversity and ecosystem service representation

Abstract

Agroecosystems have traditionally been considered incompatible with biological conservation goals, and often been excluded from spatial conservation prioritization strategies. The consequences for the representativeness of identified priority areas have been little explored. Here, we evaluate these for biodiversity and carbon storage representation when agricultural land areas are excluded from a spatial prioritization strategy for South America. Comparing different prioritization approaches, we also assess how the spatial overlap of priority areas changes. The exclusion of agricultural lands was detrimental to biodiversity representation, indicating that priority areas for agricultural production overlap with areas of relatively high occurrence of species. By contrast, exclusion of agricultural lands benefits representation of carbon storage within priority areas, as lands of high value for agriculture and carbon storage overlap little. When agricultural lands were included and equally weighted with biodiversity and carbon storage, a balanced representation resulted. Our findings suggest that with appropriate management, South American agroecosystems can significantly contribute to biodiversity conservation.

1. Introduction

Conservation strategies increasingly face the challenge of combining the consideration of biodiversity and of ecosystem service distributions within a single planning framework [1–4]. Spatial conservation prioritization (SCP) techniques have been an important tool in this regard, as they allow for the identification of sets of priority sites which optimize the representation of both, taking into account levels of complementarity [5–7]. SCP techniques also allow for the explicit incorporation of potential trade-offs between biodiversity and ecosystem services, enhancing the suitability of the priority sites identified by decreasing a priori land use conflicts. The consequences, however, for biodiversity and ecosystem service representation when incorporating these trade-offs within prioritization strategies has not been well studied.

Most examples of where biodiversity and ecosystem service trade-offs arise concern provisioning services [8,9]. The Millennium Ecosystem Assessment [8] defined provisioning services as ‘products obtained from ecosystems' and found that trade-offs between these services and biodiversity have been the largest driver of biodiversity loss over the last 50 years [8]. This is due to the dramatic land transformation that provisioning services normally imply [9,10].

Agriculture is a vital provisioning service for human well-being and a key component of the global economy [11]. Nearly 40% of the Earth's terrestrial surface is covered by agroecosystems [12], and as agricultural practices can decrease biodiversity through multiple pathways [13–15], agricultural land use and biodiversity conservation have traditionally been viewed as incompatible [13]. Given the current scale of agricultural land use, however, an increasing number of studies are considering agriculture's contribution to biodiversity and to related ecosystem services critical for successful conservation in the future [16–19].

The divergent approaches within the conservation community of considering agriculture as either compatible or incompatible with conservation goals are reflected in the SCP literature. SCP studies strictly focused on biodiversity conservation tend to include agriculture as a trade-off in prioritization strategies (i.e. a negatively weighted feature; e.g. [6]), hence forcing its exclusion from priority areas even though these areas could host significant levels of biodiversity and ecosystem services [19–24]. This results in the selection of locations with two main characteristics: (i) areas which over-represent biodiversity components that are found outside agroecosystems, while penalizing those that fall within agricultural lands, and (ii) areas which tend to promote exclusive conservation use and are suitable for proactive approach strategies (i.e. priority areas with low vulnerability [25]). By contrast, SCP studies focused on both biodiversity and ecosystem service conservation tend to include agriculture as a ‘conservation feature’ (i.e. a positively weighted feature; e.g. [5]), thus promoting the inclusion of agroecosystems within priority sites. This leads to solutions that select a set of locations that can: (i) over-represent biodiversity and ecosystem services, or agriculture, or represent intermediate levels of all three, and (ii) promote multi-use lands, which require appropriate extractive practices compatible with conservation aims, and are suitable for reactive approach strategies (i.e. priority areas with high vulnerability [25]). While both SCP approaches might address similar conservation issues and interests, their results are likely to differ significantly. To what extent and with what consequences for conservation decisions has yet to be assessed.

Here, we evaluate the consequences for biodiversity and carbon storage representation when agricultural production is considered as either a trade-off or conservation feature in SCPs for South America. Using three different prioritization strategies, in which agriculture is weighted as positive, negative or neutral, we assess how the distribution, representativeness and extent of spatial overlap of priority areas change for biodiversity, carbon storage and agriculture. We also evaluate the benefits and penalties for biodiversity under different biodiversity–carbon–agriculture conservation strategies. Finally, we discuss the implications of these different strategies for conservation strategy recommendations.

2. Material and methods

(a). Priority area

A ‘priority area’ can be defined as a location or zone that optimizes the representation of a particular variable in the landscape. This variable can be relevant for either conservation or productive purposes. While a biodiversity priority area can represent relatively high species richness or endemism, a productive priority area can represent relatively high economic production or specific environmental conditions to generate a particular kind of product (i.e. Altiplanic climate for quinoa production). For both conservation and producer parties, the identification of biological and productive priority areas is critical in order to assess potential synergies and trade-offs between these, thereby promoting both practices. Promoting different practices can be done by exclusively protecting some biodiversity priority areas and exploiting others or by making compatible the coexistence of both practices when possible. Here, a priority area is considered as an area with relatively high value of a feature in the landscape, and not only an area with high conservation value for protection.

(b). Data

Data were processed and analysed using the South American Albers Equal Area Conic projection.

Following a similar approach to Durán et al. [26], agricultural production was calculated as the sum of the averaged gross production in US dollars (USD) between 2000 and 2010. Specifically, the agricultural production layer was calculated as follows: (i) the harvested area of 95 major crops (i.e. proportion of a grid cell that has been harvested for a specific type of crop) in 2000 [27] was multiplied by crop land cover (i.e. spatial distribution of agricultural lands) [28]. From this we obtained a second layer showing the area per grid (i.e. hectare) that was harvested for each major crop. (ii) The resultant layers for each major crop were then multiplied by their respective yields (tonnes ha⁻¹) for the year 2000 [29], obtaining tonnes of crops produced per grid. Finally, (iii) tonnes per grid of each major crop were then multiplied by the average price (USD tonne⁻¹) for 2000–2010 (FAOStat), to enable the resultant layers to be sensibly combined, and to obtain the USD value of agricultural production per grid (electronic supplementary material, figure S1-d).

The carbon storage dataset was obtained from Saatchi et al. [30]. Their above- and belowground live biomass carbon stock map was produced using a combination of data from 4079 in situ inventory plots, satellite light detection, ranging (LiDAR) samples of forest structure to estimate carbon storage, plus optical and microwave imagery (1 km resolution) to extrapolate over the landscape. This dataset is at 1 × 1 km resolution, and cells were aggregated by calculating their average in order to attain the same resolution as other layers (electronic supplementary material, figure S1-c).

We used global datasets on the distributions of amphibian (compiled by the IUCN Global Amphibian Assessment), bird [31] and mammal (compiled by the IUCN Global Mammal Assessment) species downloaded from the International Union for Conservation of Nature (IUCN) Red List of Threatened Species website (http://www.iucnredlist.org/) in January 2014. For comparative purposes, we used two groups of biodiversity data for analysis: all of the species that occur in South America (‘all species' hereafter) and just the threatened species. Threatened species were selected based on their conservation status—critically endangered, endangered and vulnerable—published on the IUCN website. For both all and threatened species, distribution shapefiles were clipped to a South American continent boundary shapefile, thus maintaining only South American native ranges. They were then rasterized to presence/absence grids. For the all species group, 6606 species were included (2308 amphibians, 3090 birds and 1208 mammals; electronic supplementary material, figure S1-a), whereas a subset of 1120 species (573 amphibians, 360 birds and 187 mammals) were in the threatened group (electronic supplementary material, figure S1-b).

Given our biodiversity dataset is based on species extent-of-occurrence range maps, for which there is increasing uncertainty in species presence at high-spatial resolutions, we processed and analysed all our maps at three different resolutions—10 km, 0.5° (approx. 56 km) and 2° (approx. 224 km)—and compared the results. Given that there were no substantial differences in the key findings reported here (electronic supplementary material, table S2, and figures S2 and S3), the results for the finest resolution are presented in the text.

(c). Zonation framework

The analyses were carried out using Zonation [32], a spatial conservation planning tool that produces a hierarchical prioritization of the representation value of a gridded landscape. ‘Hierarchical’ here implies that the most valuable 5% of the landscape is within the most valuable 10%, the top 2% is in the top 5% and so on. The Zonation algorithm operates by successively removing those cells whose loss results in the smallest reduction in the value of a feature in the remaining landscape, thereby producing a ranking of the contribution of each cell. The removal order of cells depends on the cell removal rule, which determines which cell leads to the smallest marginal loss of a feature value [27,28]. In our analyses, we used the core-area cell removal rule, in which each species distribution is considered separately, securing locations that gather a high proportion of a species's geographical distribution, thus favouring the rarest species in the resulting priority area. We used this rule in order to generate complementary priority sites that contain high-priority features (i.e. rare species), which is considered a better approach to target conservation efforts in comparison with species richness (i.e. ‘additive benefit function’ cell removal rule in Zonation).

Using only one feature the strategy exclusively benefits cells that include that unique variable (‘single-criterion strategy’ hereafter), while using more than one feature the strategy optimizes the representation of all the variables at the same time (‘multi-criterion strategy’ hereafter). Features can be given numeric weights making it harder to remove cells that contain features with greater weightings. Thus, features that have been assigned greater weightings will dominate within the top percentage of the landscape that has been prioritized. The latest version of Zonation has been expanded in order to consider simultaneously both positively and negatively weighted features from the perspective of conservation [6]. Areas with positive conservation features are retained in the top fraction of the priority ranking, whereas areas with negative features (e.g. industrial areas) are removed early in the prioritization, thereby receiving a low-priority ranking. This variant of Zonation produces a spatial priority ranking that reduces the interference between competing land uses.

(d). Spatial prioritization analyses

We evaluated the performance of different prioritization strategies, in which the solution units were: (i) averaged proportion of the range of each species contained within the priority area for biodiversity (applies for both all and threatened species analyses), (ii) tonnes of carbon biomass for carbon storage and (iii) USD for agricultural production. We first evaluated the performance of single-criterion strategies—‘biodiversity only’, ‘carbon only’ and ‘agriculture only’—in order to calculate to what extent these would represent one another. In order to evaluate the consequences for biodiversity and carbon representation when agriculture is considered either as a trade-off (i.e. negatively weighted, hence excluded from conservation priority areas) or as a conservation feature (i.e. positively weighted, hence included within conservation priority areas), we assessed the performance of three multi-criterion strategies in which agriculture was, respectively, weighted for each of the three cases: 0 (hence ignored), −1.0 and +1.0. All species were weighted equally (w = 1/6606 all species; w = 1/1120 threatened species) and carbon was weighted 1.0. This implies that species were jointly equal to the carbon and the agriculture values (when agriculture =±1.0). Following the Strategic Plan 2011–2020 of the Convention on Biological Diversity [33], we used a cut-off of 17% to define the spatial extent of our high-priority areas [34]. Using the resulting prioritization maps from the single- and multi-criterion strategies, we evaluated the extent of their spatial overlap by calculating the percentage of overlapping grid squares for the highest priority (hereafter ‘top’) 17% of the landscape [6].

Following a similar approach to Thomas et al. [7], we assessed the variation in biodiversity representation in combined biodiversity–carbon–agriculture prioritization strategies. Relative priority weightings were assigned to carbon and agriculture, but the weight for each species was kept equal to 1.0. Agriculture was also weighted both positively and negatively. Thus, by assigning relative weightings to carbon and agriculture, it is possible to evaluate the extra carbon value gained for a given percentage of agriculture loss, and vice versa, and how much representation of biodiversity is achieved within these combined strategies. The relative weightings ascribed to carbon and to agriculture were defined in units of n, where n was the total number of biological species in the analysis (n = 6606 all species; n = 1120 threatened species). Similar weightings were assigned as for Thomas et al. [7]: 64n, 32n, 16n, 8n, 4n, 2n, n, 0.5n, 0.25n, 0.125n, 0.0625n, 0.0312n and 0.0155n.

For the combined strategies in which agriculture was considered as a conservation feature, both carbon and agriculture were assigned positive relative weights, and thereby both variables were retained in the top percentage of the landscape for conservation priority sites. For those combined strategies in which agriculture was considered as a trade-off, carbon was assigned positive weights and agriculture negative weights. This results in the retention of carbon in the top percentage of the landscape, but in the early removal of areas that contain agricultural lands, hence its exclusion from priority sites.

For both combined strategy approaches (agriculture positive and negative), relative weightings were ascribed reciprocally. This means that when carbon received the maximum weight (64n), agriculture received the minimum (0.0155n), and all combinations were tested through to agriculture receiving the maximum (64n) weight and carbon the minimum (0.0155n) (electronic supplementary material, table S1). Thus, when agriculture is positive, as carbon loses priority, agriculture gains priority. However, when agriculture is negative, as carbon loses priority, agriculture loses priority too, due to agriculture's weights becoming increasingly more negative (e.g. −64n), resulting in its earlier removal from the landscape.

3. Results

Priority areas for biodiversity, carbon and agriculture differ in their locations (figure 1). High-priority biodiversity areas (i.e. areas that represent a relatively high species occurrence) for both all and threatened species are concentrated to the west of South America, southeast of Brazil and south of Chile (figure 1a,d). Carbon priority areas are highly aggregated in the northwest and northeast of the Amazon forest (figure 1b), and high-priority agriculture areas occur in northern and southern South America (figure 1c). Using all species (n = 6606), the ‘biodiversity-only’ strategy represents 56.2% of biodiversity, 18.4% of carbon stock and 28.7% of agricultural production within the top 17% of the landscape. For the ‘carbon-only’ strategy, the top 17% of land captures 19.0% of biodiversity, 42.0% of carbon and only 8.1% of agricultural production. Alternatively, an ‘agriculture-only’ strategy would maintain within the top 17% of the landscape 27.1% of biodiversity, 12.0% of carbon stock and 88.0% of total agricultural production. For threatened species, ‘biodiversity only’ represents in the top 17% of the landscape 86.4% of biodiversity, 16.3% of carbon and 27.2% of agricultural production. Single-carbon and single-agriculture strategies represent 13.0% and 36.7% of threatened biodiversity, respectively.

Figure 1.

Priority maps for South America based on single-criterion Zonation analyses: (a) biodiversity only, (b) carbon only and (c) agriculture only using all species (6606 species of mammal, amphibian and bird), and (d) biodiversity only using threatened species (1120 species of mammal, amphibian and bird).

The multi-criterion strategy in which agriculture was considered neutral (i.e. weighted zero, hence ignored) maintains 33.6% of biodiversity, 40.0% of carbon stock and 10.9% of agricultural production considering all species, and 63.1% of biodiversity, 38.4% of carbon and 13.5% of agriculture considering threatened species. High-priority areas for this strategy, for both all and threatened species, are located mainly in the Amazon forest, where high levels of carbon also occur (figure 2a,d). When agriculture was weighted negatively, the representation of biodiversity and agricultural production fell to 13.3% and 0.1%, respectively, in the top 17% of the landscape for all species, but carbon representation remained similar at 39.8%. The drop in biodiversity and agriculture representation was more dramatic using threatened species, falling to 7.9% and 0%, respectively. Carbon representation remained high at 38.4%. This strategy has the characteristic that the bottom 17% of the landscape (i.e. lowest priority) has the high-priority areas for biodiversity and agricultural production, whereas the top 17% has high-priority areas for carbon storage (figure 2b,e). Assigning positive weight to agriculture increases the representation of the three variables with 38.9% of biodiversity, 31.8% of carbon stock and 64.3% of agricultural production using all species, and 65.2% of biodiversity, 30.3% of carbon and 64.2% of agricultural production using threatened species. This strategy combines areas from the far north and south of South America, and from west Amazon forest, achieving a balanced representation of the three features (figure 2c,f). Biodiversity-only areas (25.9%) overlap with agriculture-only areas in the top 17% of the landscape, whereas only 16.7% overlap with carbon-only areas (table 1). This is consistent with cells that contain high proportions of species' distribution ranges tending to co-occur with agricultural lands. The corresponding overlap between carbon-only and agriculture-only areas is low, overlapping only 7.3% of the top 17% of the landscape. Multi-criterion strategies that weight agriculture 0 and −1.0 present the highest overlap with carbon-only sites. This indicates that combined prioritization strategies in South America that ignore or exclude agricultural lands would strongly promote carbon storage representation (table 1). However, the resulting priority areas from the strategy in which agriculture was weighted equal +1.0 showed a 58.4% overlap with carbon-only areas, suggesting that the inclusion of agriculture in prioritization strategies still balances a set of areas that represent high-priority carbon areas. Moreover, this multi-criterion strategy where agriculture is weighted positively overlaps 28.7% with biodiversity-only sites, supporting once again that the inclusion of agricultural lands in prioritization strategies captures areas that represent a high relative proportion of species' distribution ranges. For threatened species, the same overlap relation was observed (table 1).

Figure 2.

Priority maps for South America based on multi-criterion Zonation analyses. All species are weighted equally, carbon 1.0, and agriculture weighted (a,d) 0, (b,e) −1.0 and (c,f) +1.0. (a–c) All species; (d–f) threatened species.

Table 1.

Spatial overlap between multiple features (biodiversity, carbon and agriculture) in South America. Percentage of overlapping grid squares for the top 17% of the landscape using all species (below, white) and threatened species (above, grey). ‘Feature only’: prioritization with each variable alone. ‘all, agri. x’: all species weighted equally, carbon 1.0 and agriculture 0, −1.0 and +1.0, respectively.

variable	bio. only	carbon only	agri. only	all, agri. x 0	all, agri. x −1	all, agri. x 1
biodiversity only	—	13.4	23.8	26.5	11.0	30.4
carbon only	16.7	—	7.3	84.6	71.3	54.0
agriculture only	25.9	7.3	—	11.8	0.2	44.7
all, agri. x 0	23.5	91.9	9.7	—	62.9	65.2
all, agri. x −1	11.4	71.6	0.2	67.1	—	39.4
all, agri. x 1	28.7	58.4	44.6	63.5	41.9	—

Potential conflicts and synergies between land uses are also apparent from performance curves (figure 3). Prioritizing for ‘carbon only’ (figure 3b,e) carries a slightly higher penalty for biodiversity representation than prioritizing for ‘agriculture only’ (figure 3c,f), and this penalty is higher for all species than threatened species. Threatened species are more effectively represented than all species when a low proportion of the landscape is allocated for conservation (‘biodiversity only’ strategy), although this carries a slightly higher cost for agricultural production (figure 3a,d). Excluding agricultural lands (‘All, agr. x −1’; figure 3h,k) carries a higher cost for biodiversity representation than when these are included (‘All, agr. x +1’; figure 3i,l). By contrast, when agricultural lands are excluded, there is not a high cost for carbon representation (figure 3h,i,k,l). Threatened species can be relatively better represented than all species when agricultural lands are included in the strategy (figure 3i,l).

Figure 3.

Performance curves for different prioritization strategies using all species and threatened species separately as biodiversity features. The x-axis represents the proportion of land that has been removed from the entire landscape, and y-axis represents the proportion that remains for that particular feature (when x = 0 everything remains in the landscape). For biodiversity, the performance curve is an average across individual species curves. (a–f) ‘Feature only’: prioritization with each variable alone, for (a–c) all species and (d–f) threatened species. (g–l) All species weighted equally, carbon 1.0 and agriculture weighted (g,j) 0, (h,k) −1.0 and (i,l) +1.0. (g–i) All species; (j–l) threatened species.

Figure 4 shows how biodiversity representation varies with relative priority weightings for carbon and agriculture, for the top 17% of the landscape. By weighting agriculture positively, the relative proportion of species' ranges is better represented when prioritization benefits agriculture over carbon, using both all and threatened species (figure 4a,c). Adding positive weight to agriculture in the prioritization strategies results in a rapid increase of biodiversity and decrease of carbon representation, with the increase in biodiversity being higher for threatened species (figure 4c). Considering all species, biodiversity representation reaches a maximum of 38.9% when agriculture is weighted n, and 66.5% for threatened species when agriculture is weighted 2n. However, as agriculture positive weight keeps increasing, biodiversity representation starts to decrease, ending in a representation of 27.1% for all species and 42.4% for threatened species when agriculture equals 64n (figure 4a,c). Assigning the same positive weight to carbon and agriculture (n), priority sites in the top 17% of the landscape represent 38.9% of biodiversity, 31.8% of carbon stock and 64.3% of agricultural production when all species were considered, and 65.2% of biodiversity, 30.3% of carbon and 64.2% of agricultural production when threatened species were considered (figure 4a,c). Alternatively, when agriculture is weighted negatively, as its negative weight increases in magnitude (hence its priority decreases), biodiversity representation remains roughly the same, starting with a maximum representation of 19.0% and ending with 13.1% when agriculture weight equals −64n (figure 4b). The variation of threatened species representation is larger, starting with 12.9%, reaching a maximum of 22.9% and ending with 8.7% (figure 4d). The variation of agriculture representation among negative weighted strategies is small using either biodiversity group, with variation of no more than 8% between the highest priority (−0.0155n) and the lowest (−64n) for agriculture (figure 4b,d). The reduction rate of carbon representation is smaller when agriculture is weighted negatively (figure 4b,d). Finally, assigning the same magnitude weight to carbon and agriculture (negative), priority areas represent 13.3% of biodiversity, 39.8% of carbon and 0.1% of agriculture for all species, and 7.9% of biodiversity, 39.1% of carbon and 0% of agriculture for threatened biodiversity.

Figure 4.

Relative weightings given to carbon versus agriculture, while biodiversity weight was kept constant at 1.0. y-axis represents what proportion remains for that feature within the top 17% of the landscape. (a,b) Using all species, agriculture is weighted positively and negatively, respectively. (c,d) Using threatened species, agriculture is weighted positively and negatively, respectively.

4. Discussion

In this study, we evaluated for the first time what the consequences for biodiversity and carbon storage representation are when agricultural lands are both incorporated and excluded from a prioritization strategy in South America. Our results show that the incorporation of agricultural lands in the prioritization strategy increases biodiversity representation (i.e. relative proportion of species' ranges), although it does not promote carbon storage representation (figure 4a,c). Also, assigning a relatively high weighting to carbon decreases biodiversity representation in the resulting priority areas (figure 4a,c). By contrast, when agricultural lands are excluded from priority sites, biodiversity representation decreases while carbon storage increases (figure 4b,d). We consider the basic and applied implications of these results.

(a). Distribution and representativeness of priority sites

Priority areas identified by the three single-criterion strategies—biodiversity, carbon and agriculture—differ in their distribution and level of representation (figure 1). The agriculture-only strategy represents 8.1% (all species) and 36% (threatened species) more biodiversity than the carbon-only strategy, and a high positive weighting on agriculture results in a dramatic increase in biodiversity representation (figure 4a,c). This indicates that the highest carbon environments in South America do not host a relatively high occurrence of species, but that agricultural lands co-occur with a high proportion of species' range distributions. A similar result was obtained by Dobrovolski et al. [34], where excluding forecasted agricultural lands for the twenty-first century from prioritization strategies (agrosolution) resulted in a significant reduction of carnivorous mammal representation, compared with strategies in which agricultural lands were included (biosolution). Like our study, this suggests that the benefits of this conflict alleviation (by excluding agricultural lands from SCP) come at a biological cost [34]. While Dobrovolski et al. [34] also used species extent-of-occurrence data, they carried out the analyses at a coarser resolution (e.g. 0.5°), hence decreasing the overestimation of species presence within their distribution ranges, and therefore within agricultural lands. However, that we did not find significant variation in our key findings among analyses at three spatial resolutions (10 km, 0.5° and 2.0°; see Material and methods) suggests that these results are quite robust to such concerns. Thus, that these two studies indicate a relatively high overlap between biodiversity and both current and forecasted agricultural lands reinforces how critical it is to identify existing agricultural lands that co-occur with a high proportion of species distributions (by including agricultural lands in SCP), and thereby targeting these lands with appropriate farming management actions.

Multi-criterion strategies that consider agriculture as a neutral or negative feature also indicate that areas with relatively high occurrence of species and with high crop production levels often overlap. This can be observed in the reduction in biodiversity representation when agricultural lands are excluded (figure 4b,d), and in the high penalty to biodiversity representation that is carried by the negative weighting of agricultural lands (figure 3h,k). Contrastingly, carbon representation increases with these strategies, which indicates that high-priority carbon areas tend not to co-occur with high-priority biodiversity and agricultural lands (figure 1). The multi-criterion strategy that includes agricultural lands presents a good balance among the three variables, covering 38.9% of biodiversity, 31.8% of carbon and 64.3% of agriculture using all species, and 65.2% of biodiversity, 30.3% of carbon and 64.3% of agriculture using threatened species. This suggests that, if systematic protection is applied to carbon and biodiversity highest priority areas, together with biodiversity-friendly farming in those agroecosystems that overlap with priority biodiversity areas, simultaneous and effective representation of carbon, biodiversity and agricultural production could take place in South America.

While our results show a high proportion of agricultural areas overlapping with biodiversity priority areas, the actual proportion of species occurring within agrosystems is likely to be an overestimation, mainly because the analysis assumes that all agricultural land is suitable for every species for which this falls within their distribution bounds. For conservation purposes such overestimation is in some senses significantly less costly, as missing the opportunity of promoting appropriate farming management in agricultural areas that could potentially host a high number of species (given the disproportionally high overlap with species' ranges) hampers the ongoing efforts of balancing agriculture with conservation.

(b). Spatial overlap, weightings and trade-offs

In an extraction versus conservation competing scenario, two approaches have been suggested [13,35]: (i) land sparing (protect some land very strictly and exploit the rest intensively) and (ii) land sharing (protect less land but exploit the remainder with friendly practices). Given that an increasing number of empirical studies show that friendly farming practices can support high biodiversity while achieving moderately high yields [36–38], a land sharing approach seems a potential solution for those locations where high extraction and biodiversity priority areas overlap. This could be the case for the 25.9% of overlap between biodiversity-only and agriculture-only priority sites in the top 17% of the South American landscape (table 1). By contrast, a land sparing approach seems more suitable for the low 7.3% overlap between carbon-only and agriculture-only priority areas, as the outputs of these two activities are rather incompatible.

The strong conflict between agricultural lands and carbon storage can be seen in the penalty for carbon when an agriculture-only strategy is carried out (figure 3c,f), and in the rapid decrease in carbon storage as agriculture gains higher positive weighting (figure 4a,c). However, as agriculture receives higher negative weighting (figure 4b,d), carbon representation also decreases, suggesting that some carbon priority sites co-occur with agricultural lands. This co-occurrence is supported by the 58.4% overlap between carbon-only priority areas and the multi-criterion priority sites that include agricultural lands in the top 17% of the landscape (table 1).

Moreover, our results show that for all species, 28.7% of biodiversity-only and 58.4% of carbon-only priority areas overlap with the multi-criterion strategy that includes agricultural lands (table 1), and which represents 64.3% of South American crop production. This shows that some areas identified by this multi-criterion strategy prioritize biodiversity, carbon and agriculture separately, whereas others prioritize both biodiversity and agriculture. While biodiversity representation increases rapidly as agriculture receives higher positive weighting, it starts to decrease when agriculture's weight approaches its maximum (figure 4a,c). This indicates that a proportion of agriculture priority areas do not co-occur with high-priority biodiversity areas. In this regard, neither land sparing nor land sharing seems the appropriate strategy for this set of priority sites, rather a mix of both. Balancing both strategies also applies to those agricultural lands that co-occur with biodiversity priority areas. While for some of them promoting biodiversity conservation may be particularly feasible and land sharing can be applied (e.g. multiple cropping or polycropping [20]), for others, given viability constraints, promoting a more suitable habitat for biodiversity conservation may be economically unfeasible and a land sparing strategy more appropriate. Thus, depending on the extent of overlap between ‘competing’ priority sites, how compatible the overlapping activities are (i.e. biodiversity versus farming), and how many priority sites without overlap are identified, would indicate the type of approach to adopt.

(c). Implications for conservation

Excluding agricultural lands (and any other form of land use) from SCP analyses is an arbitrary decision that arises from the conceptual framework on which a study is based. This exclusion predetermines what type of management approach (proactive versus reactive) can be applied on identified priority areas, as excluding lands from the prioritization process modifies a priori the nature, extent and number of variables represented. If agricultural lands are excluded, the identified priority areas will be more likely to be suitable for a single management approach, where only strict conservation activities are promoted (i.e. single land use). By contrast, if agricultural lands are included, the spectrum of management approaches widens. More biodiversity-friendly management approaches need to be included in agricultural regions in order to promote the representation of competing activities simultaneously (e.g. conservation agriculture [39]). In this regard, our study suggests that using trade-offs comprehensively in SCP aids in the identification of appropriate strategies required for the different types of priority area (i.e. single- or multi-use). This corresponds with other studies [40–42], which suggest that different areas need different conservation strategies. For instance, reactive approaches require the identification and protection of remnant natural areas in landscapes dominated by agriculture and other human uses [40]. Reactive approaches also often need strategies of coexistence of agriculture and biodiversity conservation. Proactive approaches might need strategies that protect large areas as these lands are available and tend to be cheaper [43].

Combining agricultural practices with biodiversity conservation is particularly relevant in the light of agricultural expansion, which is expected to impact about 1 billion hectares of land if current trends of agricultural intensification continue [44]. Promoting strategic agricultural intensification on existing agricultural land can reduce the extent of agricultural expansion (and hence the new conflicts with biodiversity priority areas), but can also foster the re-colonization of intensively exploited land that currently co-occurs with a high relative proportion of species' ranges but probably does not contain many species given the inadequate conditions.

Finally, even though our results highlight the increase of biodiversity representation when agricultural lands are included, we are not suggesting that all conservation strategies should be adopted within agricultural lands. Rather this study aims to raise awareness about the important effects that trade-offs in SCP can have on the representation of conservation features, and predetermines the spectrum of management approaches applied on the priority areas identified. Thus, when an SCP analysis is carried out, the consequences of excluding a particular land use, and the resulting implications for policy recommendation, should be considered.

Funding statement

A.P.D is supported by a Chilean studentship under the Becas Chile program of the Comisión Nacional de Investigación Científica y Teconológica, Gobierno de Chile (CONICYT).