Such a mechanistic understanding could help point to strategies for managing multiple ecosystem functions simultaneously , a key goal for agroecosystem management. The effects of biodiversity on multi-functionality are often context dependent, because different mechanisms govern different ecosystem processes. Therefore, managing for multiple agroecosystem services requires understanding the responses of individual services to changes in environment and management as well as trade-offs that exist among services. Given its mechanistic foundation, a traitbased approach could be used to develop agricultural and land-use management strategies to provide multiple ecosystem services that take into account such trade-offs . To develop generalizable principles of how agrobiodiversity impacts ecosystem processes and services, we propose a trait-based approach to agriculture that adopts recent advances in trait research for multi-trophic and spatially heterogeneous ecosystems . Given that traits can vary with environmental conditions, making the relation between trait diversity and ecosystem functioning context dependent, we argue that trait values should be measured across environmental conditions and agricultural management regimes. This knowledge will help predict how ecosystem services vary with agricultural practices and environment, and could be used to develop particular trait-based management strategies that can be implemented in farming systems to increase multiple ecosystem services as well as to manage trade-offs among ecosystem services in agriculture . A trait-based approach to the study of agroecosystems could transform understanding of the importance of agrobiodiversity from largely context specific and based on species identities to generalizable and predictive. For instance, although it is currently well established that intercropping can increase crop yields through niche complementarity, understanding of intercropping comes from examples of particular species interactions in particular contexts, rather than from principles that can be generally applied across different species compositions and environmental conditions.

The statement that intercropping maize with cow peas increases yield is less generalizable than the finding that,vertical hydroponics under conditions where plant-available NO3 concentrations are lower than a certain threshold, intercropping facultative N2-fifixing species increases staple grain seed set and protein content. The latter statement refers to well-defined, measurable traits , while the former refers to taxonomic affiliations that group multiple traits, thereby masking the mechanisms of how intercropping increases yield. Both approaches predict that intercropping increases yield, but the approach referring to functional traits can guide management strategies over a broad gradient of environmental conditions by pinpointing the general controls, such as abiotic and biotic , on rates of soil nutrient cycling and human nutrition .Important initial steps have been taken to apply a trait based framework to agroecosystems. The bulk of this initial research focused on using traits to understand how biodiversity in agricultural systems responds to environmental conditions and land management, rather than on understanding how biodiversity impacts agroecosystem services. Examples of trait-based response to environmentinclude weeds, pollinators , pasture vegetation, soil macrofauna, and soil microbes. Most work connecting species-based measures of biodiversity to agroecosystem services focuses on pollination and pest control ; however, most research using traits has focused on the plant trophic level, such as intercropping . Research on the contribution of intercropping to productivity has largely focused on functional group classifications. In a recent example, crops of broadly different functional types were planted in different combinations and shown to increase overall production. For weed control, the functional traits of weed seeds and cover crop grasses at the plot level are key predictors of weed seed interception by grasses that prevent weed establishment. Weed traits also have an important role in weed persistence and interaction with crop production, a reminder that functional traits can simultaneously contribute to agroecosystem services and disservices. Some initial work has also applied functional group classifications to pollination and pest control services, yet applying traits to mobile organisms remains a key research priority . The diversity of functional groups of bees explained more of the variance in pumpkin seed set than did species richness.

For pest control, functional group diversity of birds was a significant predictor of arthropod removal. However, in contrast with findings from a pollinator system, bird functional group diversity was not as strong a predictor of ecosystem services compared with species richness. Less work has considered how continuously varying measures of functional traits influence agroecosystem services. Gagic et al.provided an initial step by calculating functional trait diversity based on a mix of continuous and categorical traits to show that functional trait metrics are superior to taxonomic measures in linking diversity to several ecosystem functions. Although this study included some important agroecosystem services , it was not specifically focused on agriculture. In a forage production system, Laliberte´ and Tylianakis showed that resource addition and grazing strongly determine grassland functional trait diversity,which cascades to induce changes in grassland productivity, decomposition, and soil carbon sequestration. Abiotic and biotic factors directly impacted functional diversity, directly impacted ecosystem functioning, and indirectly impacted ecosystem functioning through changes in functional diversity. Wood et al. applied a similar approach to soil microbes on African farms and showed that, although microbial functional diversity can be strongly structured by farm management, functional diversity was a weaker predictor of ecosystem processes than were abiotic factors. This approach that simultaneously assesses the influence of biotic and abiotic controls enables ecologists to determine when functional diversity is a key control on agroecosystem services and when it is not. Many applications of trait-based research to agroecosystems have been conducted at the plot scale, while fewer studies have looked at larger or multiple spatial scales. Remans et al. showed that nutritional functional traits of crops are an important predictor of nutrition related health outcomes at a national scale. For animal nutrition, dry leaf matter content can be an important predictor of forage digestibility across broad climate conditions and management regimes. In pollinator systems, sociality is a strong predictor of pollinator response to fragmentation at the landscape scale. Such landscape fragmentation, and resulting distance between pollinator habitat and crops, can have significant negative impacts on yields.

Given that traits determine the movement of species through a landscape, as well as their effect on that landscape, more research is needed to understand how the influence of a community on ecosystem services scales up to the landscape . A trophic approach can also be crucial to understanding agroecosystem services because many services provided by agriculture are determined by activity within, and interactions across, multiple trophic levels. Some studies apply a trait-based framework to understudied trophic levels, such as birds.Storkey et al. showed overlap in the traits affecting the response of plants to tillage and the effects of plants on abundance of phytophagous invertebrates. Plant communities characterized by ruderal traits were associated with greater invertebrate abundances, suggesting that growth strategy can be linked to plant response to disturbance effects and other trophic levels .Agroecosystems range in complexity of the spatial arrangement of crop varieties, species, fields, and landscape types. This heterogeneity can have important effects on agroecosystem processes by determining the persistence, distribution, dispersal, and interactions of farmland biodiversity. These population- and community-level processes can in turn affect ecosystem services through effect traits. It is well established that the spatial partitioning of agroecosystems has an important consequence for ecosystem services. For example, pollination and pest control services depend on the spatial arrangement of vegetation in the farm scape, where farm scapes with spatial heterogeneity in vegetation types can have higher yields because pollinators and pest predators can access more of the cultivated area of the farm scape. However, pests can also rely on non-crop vegetation types to complete their life cycles; therefore,hydroponic vertical farming systems understanding pest traits could additionally provide valuable insights into ecosystem disservices that can compromise farm yields. Many of these studies on spatial structure implicitly evoke interactions between spatial structure and functional traits, but do not measure those traits explicitly. The research showing the importance of forest habitat for coffee yields assumes, and treats as static, the dispersal traits of pollinators. Given the important inter- and intraspecific variation in response and effect traits, the impact of spatial arrangement on agrobiodiversity and ecosystem services can be highly dependent on trait values and distributions. Thus, explicitly including trait measurements into existing spatial approaches to agricultural research is key. Lonsdorf et al. used a trait-based model to predict pollinator abundances in a spatially complex environment, but did not connect these predicted abundances to ecosystem services.

To integrate traits and spatial scale, trait-based data could be integrated into existing spatially explicit models of ecosystem services. These modeling approaches would first identify the landscape patches important to the provisioning of certain ecosystem services. Services, key functional traits, and abiotic properties would then be measured in each of the components of the spatially structured landscape. Spatial configuration metrics could then be calculated to determine how space influences functional trait control of ecosystem services. For instance, Biswas et al. demonstrated that fine-scale responses of plant functional trait diversity to environmental disturbance exhibit greater unexplained variance and evidence of local-scale competition than did coarse-scale patterns. Combining such spatial metrics with data on traits and abiotic characteristics would enable the development of spatially explicit models of ecosystem services that use point data to predict the landscape distribution of ecosystem services. Models with and without trait data could then be compared to determine the importance of traits vis-a`-vis environmental properties to particular ecosystem services. Such a spatially explicit representation of traits and ecosystem services would also be important because functional traits, and associated services, can vary through the farm scape over time. For instance, plant matter of N2- fixing plants is often relocated from one field to another to improve soil fertility. Sampling vegetation and soil nutrient status in single plots would fail to identify the effect of N2 fixation on soil nutrient availability in the broader farm scape by ignoring this transfer of plant matter between farm fields.In addition to being focused on small spatial scales, most research on biodiversity–ecosystem functioning has been conducted on single trophic scales. Yet, the ecosystem services provided by agriculture often depend on activity within multiple trophic levels and interactions across trophic levels. For example, rates of symbiotic N2 fixation are determined by the activity of several trophic levels. Leguminous plants regulate carbon and oxygen flow to roots that symbiotic N2-fifixing microorganisms use to fix atmospheric N2. Root-feeding nematodes can suppress N2 fixation by feeding on roots and decreasing the number of root nodules for N2 fixation. Similarly, for pest control, consumptive predator activity traits affects pest populations , which in turn affect crop yields. Thus, it is crucial not only to apply traits to understudied trophic levels, but also to understand the interactions among multiple trophic levels. A trophic, trait-based framework of ecosystem functioning requires quantifying the traits involved in the responses of species to the abiotic environment, effects of species on the environment, and the effects of species on, and their responses to, the presence and activity of species at other trophic levels. Within a given trophic level, traits determine the effect of that trophic level on an ecosystem process and/or service; the response of that trophic level to higher trophic levels; and the effect of that trophic level on lower trophic levels. These latter two types of trait can inform how trait interactions across trophic scales might improve inference about the relation between agrobiodiversity and ecosystem services. Applying trait-based research simultaneously across multiple trophic and spatial scales is essential for predicting ecosystem services because of interactions between trophic and spatial scales. For instance, large monocultures may be worse for pest control when the pest is a better disperser than the predator because the pest can out disperse the predator into the middle of the crop field and then increase in abundance. In this case, response traits interact with mobility traits, landscape context, and trophic traits to determine an ecosystem service. Although previous work in ecology has proposed the adoption of either trophic or spatial approaches to trait research, we argue that predicting agroecosystem services requires both because of interactions between these two frameworks.Previous efforts to integrate functional trait research into ecosystem service assessments have been proposed, but these have stopped short of creating tangible management targets that can be practically implemented by managers.