Empirical evidence supports neither land sparing nor land sharing as the main strategy to manage agriculture–biodiversity tradeoffs

Abstract

Agricultural land-use change is a key driver of biodiversity loss. Two alternative strategies have been discussed to align biodiversity conservation with agricultural production in landscapes containing agriculture: (i) land sparing, with intensive agriculture strictly separated from natural land, and (ii) land sharing, a mosaic of low-intensity agriculture and natural elements. Sparing builds on high-yielding intensive production to provide more area for natural habitats; sharing aims to support biodiversity within agricultural landscapes by employing wildlife-friendly farming practices. A considerable body of literature addresses conceptual aspects of these strategies, but empirical evidence on how they support biodiversity is scarce. We assessed the empirical evidence by analyzing 57 peer-reviewed articles identified in a systematic literature review, of which only 17 allowed a comparison of the strategies. These 17 articles contained 27 cases of comparisons, of which 52% reported that context-specific solutions combining sharing and sparing performed best, and exclusively focusing on one strategy cannot balance the competing demands of food production and biodiversity. In 41% cases, land sparing performed best and in 7% land sharing. However, these 17 studies almost exclusively focus on specific contexts and metrics (e.g. species population density of tropical forest birds) and the other 40 studies lack important elements for a comparison, such as the assessment of agricultural production performance. The empirical basis is thus sparse and does not support statements claiming that, in general, either land sharing or land sparing strategies are unequivocally better. It rather highlights the importance of context-specific solutions for aligning agricultural production and biodiversity conservation.

Significance Statement.

We are facing a biodiversity crisis, threatening ecosystem integrity, nature's contribution to people, and through this human well-being. The debates on solutions are controversial and often stuck between opposing advocacy for land sparing or land sharing. Our findings provide a systematic synthesis of the empirical evidence on biodiversity within sparing and sharing strategies, which is needed to support policy recommendations. Limited data are available, often based on oversimplified methods and with biased coverage of species groups towards forest birds. In consequence, the available empirical data do not support advocating for one of these strategies only. Both are needed to provide context-specific solutions for better aligning agricultural production and biodiversity conservation.

Introduction

Biodiversity has declined rapidly over the past decades (1–3). Agricultural expansion and intensification are key reasons for this (1, 4), via habitat deterioration, fragmentation, and destruction (e.g. deforestation, clearance of structural landscape elements such as hedges), damage to natural ecosystems dynamics (e.g. changes in species compositions), and pollution (e.g. from pesticide use and overfertilization) (5). Biodiversity protection and the production of food, feed, biofuel, and other biomass to meet current consumption patterns are in stark conflict, and finding solutions is of paramount importance for sustainable future agricultural and food systems (6–8).

Evolving around how to best protect biodiversity in agricultural landscapes, there is a debate that is polarized around two conflicting strategies—land sharing and land sparing (9–11). Different definitions of these concepts are available in the literature, and here, we frame them in the following. Land sharing promotes both agricultural production and biodiversity targets within the same area; it is assumed to be associated with complex landscape structures including patches of natural habitats, low-intensity production patterns, as well as biodiversity-friendly practices such as the use of organic fertilizers and biological plant protection (11–13). As a result, agricultural production in land sharing scenarios tends to have lower yields than in conventional systems and thus tends to occupy larger areas. Land sparing, in contrast, focuses on providing as much contiguous undisturbed natural habitat areas as possible, thus promoting strict separation of biodiversity protection and agricultural areas. The latter are managed with a high production intensity to minimize their size and are therefore expectedly less biodiversity-friendly (12, 14). Land sparing is thus assumed to lead to smaller areas under intensive production with higher yields to achieve the same biomass production.

This debate was initiated in 2005 by a theoretical model on the relationship between biodiversity and agricultural yield (15). This model aimed to investigate whether and to what extent agricultural landscapes may support wildlife, and the role played by natural habitats (16, 17). The debate and its underlying framework have been substantially criticized in many ways. Fundamentally, it is criticized that framing biodiversity support in agricultural landscapes by such a dichotomy of sharing vs. sparing is overly simplistic and conceptually flawed (18–22). The critique points, for example, at the normative nature of the way sparing and sharing are assessed and a lack of discussions on whether the proxies measured and metrics applied reflect what society aims to achieve (18, 23). Similarly, part of the critique of the sharing/sparing debate highlights the importance of the choice of scale and system boundaries, including telecoupling and rebound effects, which are aspects that are often neglected in the sharing/sparing conceptualization (18, 24–26). Others highlight the narrow focus on production and protection, neglecting governance, the specific livelihood situation, and special rights of indigenous communities as well as the socioecological contexts in which such management decisions must be implemented (23, 27–30). Critique is also voiced on a technical level targeting the methods used for analysis. Baudron et al. (16), for example, criticize the use of density-yield functions as not being robust indicators for investigating species survival or for determining how species react to changes in yields if other landscape changes occur (20, 31).

In spite of these critiques, the debate has quickly gained importance in both science and policy (13). Land sparing is often strongly promoted (10, 21, 32), and it is claimed that only sparing can protect large natural habitats (e.g. forests) from human influence and maintain biodiversity. This argument is prevalent in policy and societal debates favoring agricultural intensification to support further yield increases (e.g. with intensive monocultural systems) rather than low-intensity systems (e.g. organic systems complemented by biodiversity-friendly measures or high nature value farming) (33, 34). In several recent initiatives related to ecosystem protection and biodiversity conservation, agricultural intensification plays an important role, such as in half-Earth (35, 36) or 30 by 30 (36, 37). It should be noted, though, that these initiatives also mention many other aspects, such as agroecology, reduction of food waste, and sustainable consumption patterns.

Nevertheless, land sparing has been frequently contested from the start, by both researchers and NGOs (e.g. 11, 19, 38). It is argued that almost no undisturbed habitats are left and that impacts from intensive agriculture cannot be confined, unavoidably resulting in adverse impacts on biodiversity and ecosystems beyond the spatial and temporal boundaries of agricultural activities (39, 40). Highly intensive production in sparing regimes strongly degrades soil quality and ecosystem services required for agriculture, such as pollination, thus endangering yields and food security in the longer term (41). Agricultural intensification and increased yields do not necessarily lead to spared land, without a policy framework in place (11, 27, 42); in fact, it can even lead to further expansion of agricultural land (a phenomena exemplified by Jevon's paradox or rebound effect) (43). Some critics also suggest that land sparing supports agricultural systems that are problematic in terms of justice, equity, local food security, and other societal and governance aspects (e.g. 9, 23, 44, 45). Furthermore, the support for land sparing in many studies is implicitly based on the presence or absence of natural habitats such as primary forests, which are not generally found in agricultural landscapes (46).

In general, adopting context-specific solutions and avoiding “one-size-fits-all” approaches seem crucial for sustainable land use (47). This is also reflected in a strand of literature, which goes beyond the dichotomic approach of land sparing vs. land sharing. Grass et al. (48), for example, argue that a combination of both strategies is needed, embedded in well-connected landscapes: sparing is needed for species that are incompatible with agriculture and as refuges, and sharing is needed for connectivity (e.g. for preventing isolation and subsequent extinction of populations in protected areas) and provision of ecosystem services as the basis for productive agricultural landscapes. Relatedly, other authors argue for nuanced scale and context-specific options or mixed sharing/sparing strategies (see e.g. 11, 24, 49–51).

Accordingly, there is much literature on the sharing/sparing debate, including substantial warranted criticisms that fundamentally challenge the appropriateness of the sharing/sparing conceptualization. As has been stated by e.g. Fischer et al. (18), the sharing/sparing debate is based on normative choices on what to focus on regarding biodiversity conservation and agricultural food production. Thus, before policy advice can be based on results from this debate, the priorities of society regarding agricultural food production and biodiversity conservation need to be defined. Next, it should be determined whether these priorities are sufficiently captured in sharing/sparing assessments. If these requirements are met, policy recommendations related to this topic should then be based on the most encompassing quantitative empirical knowledge available. However, the current literature synthesizing the debate focuses on conceptual aspects, analyzing what is covered or lacking in various studies (e.g. 16, 18, 22, 24, 52); focuses on specific geographies (e.g. the global tropics in Luskin et al. (46)); excludes important aspects (e.g. the meta-analysis of Barral et al. (53) does not consider agricultural production); or does not provide a systematic review including the latest evidence (11). A systematic review of the empirical basis on how sharing and sparing strategies compare regarding biodiversity protection is thus direly needed to summarize the current evidence and reveal potential gaps. Here, we present the results of such a systematic review. The aim is to provide a robust basis for decision-making, built on the synthesis and critical analysis of the quantitative results from comparative empirical studies investigating the relation between biodiversity and sparing vs. sharing land management.

Results

General results

We identified 57 eligible studies that empirically compared land sharing with land sparing strategies in an agricultural setting and assessed biodiversity (see Materials and methods and Section S1.2 for further details). We split the studies into cases, where results could be differentiated into distinct species groups or regions, resulting in 105 cases. These cases are very heterogeneous and cover different species (Figs. 1g and h and S1c), climatic zones (Figs. 2 and S1), natural ecosystem types (Figs. 1i and j and S1d), and agricultural systems (Figs. 1e and f and S1b). Furthermore, they employ a wide number of different methods to measure agricultural intensity (Figs. 1c and d and S1a) and biodiversity (Figs. 1k and l and S1e), to characterize land sharing and land sparing, and to determine agricultural production (Table 1, also see Dataset S1). We analyzed the studies and categorized them according to whether they contained all the information needed for an analysis of how land sharing and land sparing compare (i.e. assessment of agricultural production performance, clear specification of the biodiversity metrics used and of the natural ecosystem in the study region, and clear definition of what the sharing and sparing production systems are). Only 17 studies (27 cases) included all of this information and were assigned to categories 1 and 2 (Table 1). Although these 17 studies fulfilled basic criteria to compare a sharing with a sparing strategy, they were still far from being comprehensive. We found that many of these studies followed the debated framework proposed by Green et al. (15) and therefore limited their assessment to yield and species population density (Fig. 1d and l), disregarding other important aspects such as profitability, scale, or societal context. Furthermore, these studies were strongly biased in species coverage towards tropical forest birds (19 out of 27 cases in categories 1 and 2). The other 40 studies (78 cases) lacked at least one reviewed aspect. Most often, an agricultural production metric (e.g. yield) was missing, and thus, the link between different strategies and agricultural production performance was not possible (30 studies, 45 cases). These studies addressed habitat suitability for different species in different agricultural systems but did not allow for an assessment of how biodiversity conservation and agricultural production targets could be achieved simultaneously. Twelve studies lacked clear definitions of land sharing, or the allocation of the production systems to one of the two strategies was unclear. For seven studies, a clear identification of the spared land was not possible (cf. Dataset S1 for further details). Some studies (54, 55) also defined protected areas as land sparing, without the associated area for intensive production, and therefore compared biodiversity in low-intensity agriculture (sharing) with protected areas. This disregards the intensive production areas that would be part of the land sparing strategy when considering the original definition by Green et al. (15).

Fig. 1.

Overall results a), b) on the performance of the different strategies in the studies and the performance differentiated by different data items c–l). The panels to the left cover all 105 cases (57 studies), while the right-hand panels cover the studies of highest quality only (17 studies of categories 1 and 2, 27 cases). “both”: either strategy, depending on the circumstances, or a combination of both performed best; “Sharing”: land sharing performed best; “Sparing”: land sparing performed best, “not significant”: no significant difference was found for these cases; size of circles and numbers: number of cases reporting the respective result. (cf. Section S1.5 for further description of categories in e)–h), k), and l); four studies (seven cases) measured an additional agricultural metric next to yield (number of cases remains 105, but numbers in c) therefore add up to 112); crop-livestock are systems with a combination of arable cropland and livestock production, forest includes temperate (five cases), subtropical (six cases), and tropical (46 cases) forests; “other” in i) is the study of Tello et al. (56), including several species groups but not giving individual results. HANPP, human appropriation of net primary production; species pop. density, species population density; c, cases; s, studies).

Fig. 2.

Global distribution of reviewed studies (map) and distribution of cases within climate zones (bar plot). Approximate location of sampling sites of 102 cases (56 studies, one study covered 27 sub-Saharan countries giving only an overall result and is therefore not pictured on the map (57)), climate zone classification, and best outcome of studies in different colors and symbols: “both” (circle): either strategy, depending on the circumstances, or a combination of both performed best; “Sharing” (triangle): land sharing performed best; “Sparing” (square): land sparing performed best, “not significant” (rhombus). Three studies (seven cases) are classified in different climate zones by the authors than they appear on the map (54, 58, 59), in the analysis and bar plot they have been treated as classified by the authors. Points are indicative for the approximate location of the cases; some coordinates have been changed to improve readability of the figure, by spreading the points around their location to avoid overlap and adding a line to each point indicating to which location the point belongs. Barplot segments are ordered from top to bottom as: “both”, “sharing”, “sparing”, and “not significant” (the last category occurs only in temperate and tropical regions). Background map adapted from (60).

Table 1.

Categories of cases/studies identified in the eligible studies, their description, and the number of cases and studies in each category.

#	Category name	Description	# of cases (studies)
1	Full comparison, density	Full comparison of a sharing strategy (low-intensity production system) with a sparing strategy (high-intensity production system and natural habitat) Agricultural production measured (all studies use yield as metric) Density-yield function (or very similar method) to assess biodiversity in both production systems and natural habitat	19 (10)
2	Full comparison, other metrics	Full comparison of a sharing strategy (low-intensity production system) with a sparing strategy (high-intensity production system and natural habitat) Agricultural production measured (all studies use yield as metric) Other methods than density-yield function (e.g. species richness) to assess biodiversity in both production systems and natural habitat	8 (7)
3	Sharing vs sparing, no yield	Comparison of land sharing and land sparing Well-defined strategies No yield or agricultural performance measured	14 (6)
4	Incomplete strategies	Comparison of agricultural systems Natural habitat or sparing system missing Mostly without measurement of agricultural performance	30 (12)
5	Scenario modeling	Modeling approach based on empirical data Comparison of landscape diversity, heterogeneity, and sharing/sparing using scenarios	12 (7)
6	Expansion vs. intensification	Agricultural expansion impacts vs. agricultural intensification impacts	4 (2)
7	Landscape level	Biodiversity and landscape diversity at a regional/landscape level Correlation of HANPP (human appropriation of net primary production) and human disturbance with species richness	6 (2)
8	Carnivore occurrence	Separation vs. coexistence of large carnivores and humans Habitat and human settlement/disturbance overlap	4 (3)
9	Others	Individual comparisons e.g. species occurrence vs. specific measures, practices, etc.	8 (8)

Categories 1 and 2 include the most complete studies allowing for an adequate comparison of sharing and sparing (cf. SI for full description of categories).

Given the strong bias and the large heterogeneity in the studies, we refrained from undertaking a quantitative meta-analysis (cf. Section S1.3 for further explanations). We thus provided a descriptive assessment of the study results to show whether they concluded that sharing, sparing, or a combination performed best for balancing agricultural production and biodiversity (cf. Section S1.5 for the detailed definitions of these strategies).

The subset of the 17 most complete studies and the full data set draw a similar picture, namely that in a large share of cases and studies a combination of both sharing and sparing was best depending on the specific context (52% of the cases in the most complete studies and 34% of all cases), as depicted in Fig. 1a and b. Sparing performed best in 41% of the most complete cases and 36% of all cases, sharing in 7 and 19%, respectively. Eleven percent of all cases and none of the most complete cases had a nonsignificant result, meaning that the statistical analysis performed in the studies did not find significant differences between the performance of sharing and sparing for these cases (cf. Section S1.5 for further details). The relatively poor performance of pure sharing strategies also reflects the fact that many studies in the sample focused on species (e.g. tropical forest birds or mammals) that depend on large natural forest areas being present (cf. also Fig. 1g and h). In these instances, sharing strategies performed poorly, as the habitat for these species was not present. Overall, sparing was not generally the better strategy, rather both sharing and sparing seem to be needed, depending on the context, to support biodiversity either equally well as sparing individually, or even better when considering only the most complete studies in categories 1 and 2.

Species, ecosystem types, climate zones, and biodiversity metrics

The studies covered a broad spectrum of species groups, but the majority considered bird species (Fig. 1g and h). Overall, no clear differences in the performance of the strategies could be identified. However, some trends can be seen for certain species groups. For birds, sparing and a combination of both strategies performed better than sharing, and for plants and insects, sharing performed similarly to sparing or the combination of both. Only two cases considered soil organisms (earthworms in Gabriel et al. (61). and soil arthropods in Segre et al. (62)), although soil health and biodiversity are crucial for sustainable agriculture. There were no studies on microbes or fungi. The species spectrum narrows down to birds (19 cases), insects (two cases), and plants (six cases) when only considering the 27 most complete cases. Several studies evaluate species dependent on natural habitats (e.g. obligate grassland species (63–65), forest-dependent species (66, 67), or large carnivores needing undisturbed natural habitat (68–70)), excluding others that play a pivotal role for important ecosystem functions, such as those that comprise the soil microbiome and support soil health. Seventeen of the 57 reviewed studies considered specialization and habitat preferences, of which 15 looked at birds. Mobility of species was rarely considered, but Steffan-Dewenter et al. (71) noted that within forest-based species, plants were more affected by a switch from forest to agroforestry than the more mobile birds and insects. Michael et al. (54) found that for mobile species such as birds, sharing could be a conservation strategy, while for less mobile species such as amphibians or reptiles, more species-specific management was needed and heavily forest-dependent species such as arboreal marsupials needed protected forest areas. Birds are generally rather mobile, but 20 cases (18 studies) identified sparing as the best strategy for them. This may reflect specific habitat preferences of the birds considered in these studies (e.g. contiguous forest; 24 of 45 bird cases are in forest ecosystems, 22 of which are in tropical forests) and is also reflected in the generally better performance of sparing in tropical regions and forests (Figs. 1i and j and 2), as previously also suggested by Ramankutty and Rhemtulla (49, 50).

For all other natural ecosystem types, depending on the context, a combination of both strategies gave the best outcomes (Fig. 1i). Large areas of primary or secondary forest still exist in tropical regions with species that are highly reliant on this type of ecosystem and unable to live in agriculturally managed landscapes. Accordingly, the good performance of sparing in the tropics is likely due to many studies in the tropics analyzing such species (e.g. 67, 72). One may expect more ambiguous results if more studies on other species and from other tropical biomes become available, such as tropical seasonal forests or savannas. These examples show that conservation and production strategies need to consider individual preferences and habitats of species and regional contexts, but many specialists will nevertheless rely on protected areas.

Overall, the reviewed studies neglected relevant aspects of biodiversity. This is reflected in the limited number of species groups covered and the dominance of birds therein (see above) and in the biodiversity metrics employed: species density and richness were predominant (with 43 cases in 23 studies and 44 cases in 22 studies, respectively, see Fig. 1k), while only five cases (two studies) employed functional and phylogenetic diversity. Further metrics such as the diversity of interactions between species relevant for ecosystem stability (73) and many essential biodiversity variables (e.g. species traits, genetic and ecosystem composition, community and ecosystem structure, etc. (cf. 74)) were missing.

Representation of agricultural practices

As with the species and biodiversity metrics, agricultural metrics were not adequately represented in most of the reviewed studies and often not considered at all (Fig. 1c and d). Forty-five cases (30 studies) did not include any assessment of agricultural productivity or intensity (e.g. yield), thus hindering a thorough assessment of biodiversity in relation to agricultural productivity. Forty-one cases (22 studies) reported yields (four thereof also profitability), and only a few studies covered other agricultural performance metrics without yield (profitability–3, human appropriation of net primary production–2, see Fig. 1c and d). For land sparing, some studies stated that the agricultural areas need to be managed “sustainably,” but it is not further specified in the studies what this actually means and how sustainable agriculture in a sparing system should look like. Furthermore, ecosystem services such as pollination or soil health, characterizing performant agricultural areas besides yield as a provisioning service, were hardly addressed. A few exceptional studies reported on pest control (62, 71), pollination (75), litter decomposition (71), carbon storage (71, 76–78) or recreation (79) (for details, see Datasets S1). Interestingly, the studies looking at services relevant for agricultural production (pest control, pollination, litter decomposition) recommended sharing or a combination of both strategies. This indicates that sharing or a combination is needed for supporting biodiversity while also providing ecosystem services relevant for agriculture and thus long-term food security, a finding also backed by the evidence from many studies addressing the performance of agroecology regarding ecosystem services provision and biodiversity (e.g. 80–82).

Discussion

The empirical evidence is too scarce to justify general support of sharing or sparing strategies

This systematic review shows that the empirical basis for assessing the performance of land sharing and land sparing strategies regarding biodiversity in agricultural landscapes is relatively weak. Only 105 cases in 57 studies provided an empirical assessment and only 27 of these cases (17 studies in categories 1 and 2) provided enough detail to allow for a comparison of these strategies, meaning that they included a full definition and assessment of both strategies (see Table 1 categories 1 and 2) and measured biodiversity and agricultural production performance. Furthermore, the studies were strongly biased in species coverage (strong focus on birds) and very context-specific (e.g. covering one species group or limited spatial scope), which provides valuable insights for their context, but the results and empirical evidence derived from them should not be extrapolated. From our review, a main conclusion is that the scientific evidence is currently insufficient and does not justify generally and globally valid statements supporting one strategy or another without considering the specific context. Similarly, Luskin et al. (46) found that results are context-specific by analyzing how recommendations change in regard to authors' backgrounds, scope, study methodology, and geographical focus and scale. However, Luskin et al. (46). also considered recommendations from perspectives and reviews, not just empirical studies, and focused only on tropical and global studies.

Many empirical studies assessed whether agricultural systems could be suitable habitats for native species or not, but the tradeoffs between biodiversity conservation and agricultural production in general could not be addressed by their methods. There is thus an urgent need for additional empirical and comparable data to provide a better scientific foundation for assessing biodiversity strategies for balancing agricultural production and biodiversity conservation goals in agricultural landscapes. Importantly, additional studies would need to be conducted in adequate quality regarding conceptualization of sparing and sharing strategies as well as agricultural production metrics, and insights are only possible if a larger body of research is developed in an unbiased manner and with broad coverage regarding biodiversity metrics.

This lack of a broad empirical basis for the sharing/sparing debate also relates to the narrow focus and normative choices associated with the sharing/sparing framework proposed by Green et al. (15) and the focus on few agricultural and biodiversity metrics. A strong focus on yield is given by the framework and can also be seen in the limited selection of agricultural intensity metrics in the studies, disregarding other important aspects of agricultural production such as profitability and food supply stability as well as yield quality (e.g. 83). Similarly for biodiversity, the studies mainly focus on the presence/abundance of species; however, there are many more biodiversity metrics available, such as functional diversity, endemic species richness, or proxies for ecosystem quality from different impact categories (e.g. water consumption, climate change, eutrophication). So far, there are no recommendations from organizations such as United Nations Environment Programme (UNEP) or the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services regarding which metrics are best practice for biodiversity and ecosystem quality assessments. However, the UNEP is in the process of developing such suggestions (e.g. 84 or 85 for life cycle assessment).

Lacking consensus in definitions limits comparability

The studies reviewed adopted a wide range of different definitions for land sharing and sparing, which limits the comparability of the studies. Sharing can, for example, be defined as the total area used agriculturally with the lowest yield possible to still achieve a given production target (e.g. 17, 59, 86), or as systems including a variety of wildlife-friendly agricultural practices and natural elements (e.g. 58, 64, 65). These sharing strategies usually go along with reduced yields, but in many cases, yields are not reported. Because of these differing definitions, the same system was sometimes classified differently. Silvopastures (i.e. pasture-based livestock systems including trees), for example, were considered a mixed system covering both sparing and sharing aspects in some studies (87, 88), but were considered land sharing in others (89). Similarly, the definition of sparing can depend on scale and utilization of spared land (24); lightly grazed land was considered spared land in some studies (59), and in others, it was the sharing production system (90).

Furthermore, the definition of sharing in some studies was purely based on low yields, assuming that there are no biodiversity-friendly practices that achieve intermediate to high yields, even though there is ample evidence for such practices (e.g. 91–93). This assumption was prevalent in many of the studies in category 1 that employed the species density-yield functions. These studies found that sharing rarely performs well, as land competition between agricultural production and biodiversity becomes very strong due to these assumingly unavoidably low yields. In fact, pure sharing, according to such a definition via lower yields, means that the total area is cropped with low intensity, i.e. a homogenous agricultural landscape similar to intensive agriculture just with a very low yield. In this case, only species that are adapted to such agricultural landscapes may survive. This is also reflected in the assessment of “winning” and “losing” species from agricultural and natural habitats usually reported in the studies from category 1, based on species density-yield relations (17, 94). However, as a matter of fact, systems typically considered as sharing (e.g. shaded plantations (95), polycultures (96), organic systems (97)) are often heterogeneous farmland mosaics comprised of a mix of agricultural and natural elements. Therefore, sharing should rather be defined through biodiversity-friendly practices than purely by the yield.

The narrow focus on natural habitat needs of native species fails to capture the complexity of the agricultural context

Several studies (66, e.g. 67, 98, 99) focus heavily on forest-dependent species that rely on specific features (e.g. hollow elements for arboreal marsupials), which only a forest habitat can provide (54), which, in these studies, is only present to sufficient extent in sparing. It is therefore unsurprising that in these studies sparing is the best strategy for these species. However, sharing can also contain significant portions of natural habitats, and fragmented habitat patches can make a major contribution to improve biodiversity (100). More differentiated studies reveal that low-intensity agriculture and heterogeneous landscapes can promote various species, for example, generalists or species with a preference for heterogeneous habitats (89, 101, 102), but also specialists and species relying on more open habitats such as carnivorous and scavenging birds (103). In the studies reviewed here, but also in general, the discussion of sharing vs. sparing is dominated by the question of what habitat is needed for native species. However, focusing only on the needs of native species in a debate about the tradeoffs between biodiversity and food production is too narrow, especially considering that the debate is framed in an agricultural context, with sustainable food production as a main goal besides conservation. For a thorough assessment, such bias must be avoided and species related to agricultural areas and relevant for healthy fertile soils (soil microbiome (cf. also 104, 105)), pollination and biological plant protection should also be assessed to an equal extent with species that depend on the presence of natural habitats such as forest birds. Furthermore, a dualist approach which sees sparing and sharing at the extremes as in the polarized debates referred to in the introduction tends to implicitly assume that all species have the same need, overlooking their differences (e.g. specialist vs. generalists, species living on edges vs. species living in closed forests, etc.). For example, a recent meta-analysis shows that complex agricultural landscapes host more biodiversity with potential benefits for agriculture than less complex landscapes (82). Rasmussen et al. (106) show that diversification in agriculture and the combination of strategies can be beneficial for environmental and social conditions, and in areas shaped and influenced by agriculture, heterogeneous landscapes should be included in wildlife conservation management strategies to prevent biodiversity loss in and around land used for agriculture (58, 93).

The discussion on sharing and sparing needs to be context-specific

Reviewed studies advocating for land sparing usually declared that the intensive production needed for high yields on agricultural areas should be sustainable. However, these studies rarely defined what sustainable agriculture or sustainable intensification means and what effects this might have on agricultural practices and yields. An uncritical support for intensive agriculture based on the land sparing argument is problematic. For example, degradation of soils can lead to decreased yields over time (41), with water erosion and soil compaction alone being estimated to lead to productivity losses of 10–20% in the long term (107). Other areas would then be required to reach production targets, thereby potentially endangering the hitherto spared land. Studies disregarding these negative impacts, focusing on current yields only as the single performance measure for agricultural systems, and advocating for a pure sparing strategy can lead to misinformation, when replicated in general media and used for the justification of environmentally harmful intensification in agriculture, cf. the publication by Bateman and Balmford and the related debate (10 and e.g. 108 and 109). Depending on the context, it is also possible that environmentally friendly alternatives can perform better than intensive agriculture and are much more suitable in areas prone to drought or other impacts of climate change (110, 111). Such systems can be more resilient and maintain stable biodiversity, and some smallholder production systems can even be more productive than industrial farming due to beneficial effects of biodiversity (19). The analysis of the relation between performance of the strategies and temporal/spatial scale and additional context variables that may serve as covariables in a meta-regression would contribute to such context-specific discussion. We refrained from doing such, given the currently available data with its bias and high heterogeneity (cf. also Section S1.3 for further explanations).

Finally, the societal context is central. The livelihood of many communities and smallholders depends on natural resources attained from their land (112, 113). Systems managed by indigenous communities and smallholders are generally considered land sharing and are crucial to manage biodiversity hotspots (28), to ensure food security, and to provide ecosystem services, such as pest control and pollination (114, 115). These aspects are rarely covered in the studies reviewed (and not at all in the most encompassing studies in categories 1 and 2). The prevalent framework limited to yield and the population density of species disregards these relevant aspects and switching to industrial agriculture could not only endanger the livelihoods of local people, but also the resilience of the local agroecosystems and threaten biodiversity hotspots. The inclusion of local people and regulations (e.g. indigenous people) is therefore of utmost importance for sustainable agriculture.

Overall, the complexity of the nexus between food production and biodiversity conservation calls for more holistic studies, including societal aspects, broader selections of species, biodiversity, and agricultural metrics, and detailed analysis of the relevant wider context.

Conclusion

Our results show that empirical evidence on how sharing and sparing strategies compare in supporting biodiversity is very scarce. The existing empirical evidence does not support that either sharing or sparing is unequivocally better, but rather either one or a combination of them can be better under different circumstances. It is thus important to move from an ideological debate to a more integrative view, where both strategies are included according to their strengths and in relation to the specific contexts, as also previously illustrated in theoretical modeling approaches (20, 51) and argued for in conceptual discussions (e.g. 18, 23, 108, 116). Furthermore, the debate between these two strategies often overlooks the underpinning problems associated with the dualistic model itself (e.g. 11, 16, 18, 117, 118). We conclude that a combination of environmentally friendly agricultural practices, improvement of ecosystem service provisioning, and the protection of natural habitats are needed to resolve the potential tradeoffs between agricultural production and biodiversity. Ultimately, this debate has always been embedded in the context of producing enough food as one cornerstone of food security. The consumption side thus needs to be addressed as well, due to the high levels of food loss and waste and animal source food consumption, leading to high shares of cropland for feed production for animals that could instead be directly used for food production for humans. Changing the consumption patterns and reducing the total demand of goods and land could thus compensate for potentially lower yields of environmentally friendly agricultural production without requiring more land (cf. 119).

Materials and methods

Literature review

To derive an evidence-based answer to the question whether land sparing or land sharing performs better to balance the tradeoff between biodiversity and food production, we systematically reviewed existing scientific literature. We searched Web of Science and Scopus for relevant articles that had to include a term of each of the three following search term categories: (i) term related to sparing, (ii) term related to sharing, and (iii) a quantification term (see Section S1.1 for full search term string). The search resulted in 445 studies, which were then further screened following the PRISMA protocol according to these inclusion criteria: empirical studies, comparison of sharing and sparing, agricultural context, and including an assessment of biodiversity (see also Fig. 3). Studies were included regardless of their definition of the sharing and sparing strategies. In the end, 57 studies were left for further analyses (cf. Section S1.2).

Fig. 3.

PRISMA flow diagram. The literature search process according to the PRISMA protocol and including the literature search, screening, exclusion criteria, and number of included studies (n = 57). Adapted from (120).

Included here, as stated above, are only articles focusing on a comparison of the two approaches land sharing and land sparing. There is an additional vast body of literature generally assessing biodiversity in various agricultural systems. However, this literature has not been considered for the present review, since it focuses on the state of biodiversity in different systems or on the effect of various agricultural practices on biodiversity but does not compare strategies on how to best promote biodiversity while also ensuring food production.

Analysis

We categorized the 57 studies according to several criteria for completeness and methodological quality into nine different categories (Table 1). All studies were analyzed together and further also a subgroup of categories 1 and 2 separately. Categories 1 and 2 contained the highest quality studies with the most complete comparison between the two strategies “sharing” and “sparing,” i.e. production systems for both strategies were defined, biodiversity was assessed within these systems as well as in natural habitat to complete the sparing strategy, and additionally, agricultural productivity was assessed. All studies in these two categories assessed agricultural productivity as yield. This was not due to deliberate exclusion of studies using other agricultural intensity metrics, but because all other studies either did not assess agricultural production performance at all or were incomplete in another way (e.g. no natural habitat was included). Overall, there were very few other agricultural intensity metrics used (see Fig. 1c and d). For the analysis, the studies were further separated into cases, where different locations were included or different species groups with clear individual results were covered, resulting in 105 cases (cf. Section S1.2 for further details). To analyze the cases, we grouped them according to characteristics such as climate zones, natural ecosystem types, or species groups covered and evaluated the outcomes of the studies (sharing, sparing, both) per category of these characteristics (see Figs. 1 and 2). No statistical analyses were performed, given the strong bias in species coverage and the large heterogeneity of the studies and cases (see Section S1.3 for further details and explanation).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This research was conducted as part of the project “Deliberative Diets: Connecting Producers and consumers to value the sustainability of Swiss food systems scenarios” funded by the Swiss National Science Foundation, Sinergia Funding Programme, grant no. CRSII5_202300.

Author Contributions

E.A., A.F., A.G., Al.M., F.R., S.P., and Ad.M. designed the research. E.A., A.F., and Ad.M. performed the research and analyzed the data. E.A., A.F., A.G., Al.M., F.R., S.P., and Ad.M. wrote the paper.

Data Availability

All data associated with this article are freely available and included in the article or Supporting Information.