Conceptual model of agricultural land systems for holistic research supporting the SDGs

In this study, we conduct a quantitative analysis of the domain (environmental or social) and functional role (driver, management choice, or outcome) of agricultural system variables (Fig. 1) from the three perspectives: research, policy, and practice (farmers and extension services). We use the European Union (EU) as a case study because of its commitment to be a global leader in achieving the SDGs (24). For the research perspective, we analyze 69 peer-reviewed articles (22) from agricultural land systems research because of its integrated, interdisciplinary, and stakeholder-oriented agenda (21), which is well-positioned to lead research on agricultural systems for the SDGs. For the policy perspective, we analyze five policies: the global SDGs, the EU’s implementation of SDG indicators, the EU’s Sustainable Development Strategy, the EU’s Common Agricultural Policy, and the latter’s associated agri-environmental indicators. Lastly, we gauge the practice perspective based on a review of seven agricultural sustainability assessment tools that seek to inform the practices of farmers and their extension services (e.g., the Sustainability Assessment of Food and Agriculture systems,; see list of tools in SI Appendix, Table S1). We use agricultural land systems research as an illustration of how one research field important for contributing the SDGs currently aligns with agricultural policy and practice perspectives; other fields may have different foci.

We identify 32 variables across the environmental and social domains of agricultural systems that are shared by research, policy, and practice. An additional 207 variables, however, are considered from only one or two perspectives, indicating that taking an individual (e.g., research-only) perspective on trade-offs in agricultural systems limits the possibility to understand, incentivize, and achieve the SDGs. We statistically analyze current agricultural land systems research in Europe, identifying four dominant approaches, none of which fully encompasses the 32 variables of shared importance. We argue that the prevailing data and methodological paradigms, as well as the limited adoption of systems approaches, prevent current European agricultural land system research from more fully meeting the needs of policy and practice. We identify opportunities for research to integrate perspectives from policy and practice, particularly through systems approaches, codesign of research, and communication with policymakers and practitioners. Such integration will support research that better evaluates trade-offs and guides agriculture’s contribution to the SDGs.


Components of Agricultural Systems from Research, Policy, and Practice.

From cataloging all indicators measured or listed in a research, policy, or assessment tool (practice) document, we identified more than 800 specific indicators. We aggregated these to a final list of 239 more-general variables of environmental and social drivers, management choices, and outcomes (Dataset S1). The variables were relatively evenly distributed between the environmental and social domains (Fig. 2A), reflecting the balanced importance given to the environment and society in agricultural systems.

Fig. 2.

The focus of research (69 peer-reviewed articles), policy (5 policies), and practice (7 agricultural sustainability assessment tools) on different agricultural system variables in Europe. (A) Total number of environmental and social variables of agricultural systems classified by their functional role as drivers, management choices, or outcomes. (B) Emphasis of the three perspectives on functional roles of variables in the environmental domain. (C) Emphasis of the three perspectives on functional roles of variables in the social domain.

In terms of system function, social drivers comprised the largest number of variables, followed by environmental drivers and management choices (Fig. 2A). Policies can influence many of the economic and political drivers in the social domain of agricultural systems, whereas environmental drivers may be more difficult to adjust, depending on the temporal and spatial scales being considered. This is perhaps why policies and assessment tools contained so few environmental drivers (e.g., soils, topography, climate) compared with the research reviewed (Fig. 2B). The large number of environmental management choices compared with social ones (Fig. 2A) reflects the importance from all three perspectives of managing the land on which agriculture relies (Fig. 2B compared with Fig. 2C). Simultaneously, this finding may indicate that the importance of managing social aspects of agricultural systems has, to date, been underrecognized in Europe. Alternatively, there may simply be fewer social components that can be managed and manipulated in agricultural systems. Social outcomes of agricultural systems also appear to be underresearched (from a land systems perspective, sensu ref. 22), relative to the importance of social outcomes emphasized in policy and practice (Fig. 2C).

We found a small core set of 32 variables shared among research, policy, and practice, comprising less than 13% of all variables. This suggests a limited consensus among the three perspectives regarding what is considered important for agricultural systems. Consensus among the three perspectives was greatest in environmental management choices, followed by social drivers and environmental outcomes (Fig. 3). The majority (56%) of environmental variables shared by the three perspectives related to soil and biodiversity (which are additional categories described in ref. 22 used to classify agricultural land system components; SI Appendix, Figs. S1 and S2), including variables such as tillage, fertilizer use, soil erosion, and pesticides (Fig. 3). Six of the 14 shared social variables related to political drivers, including policies on climate, environment, and agriculture, as well as subsidies and land ownership (Fig. 3). Despite the relatively large emphasis on social outcomes in policies and assessment tools (Fig. 2C), only four variables in this category were also shared by research: income, yield, labor productivity, and un/employment rates (Fig. 3). Social outcomes of agricultural systems appear to be underresearched in European land systems research (22), and evaluation of trade-offs among outcomes from a research-only perspective may overlook many variables important to policy and practice.

Cataloging over 800 individual operationalized indicators from the reviewed research, policies, and assessment tools into a final group of 239 environmental and social variables (SI AppendixSupplementary Methods and Dataset S1), the variables were then characterized as drivers, management choices, or outcomes based on our interpretation of how they were conceptualized in the original source and using the classification in ref. 22. The relative importance of these variables (e.g., soil type vs. precipitation) may differ across scales, but we made no assertions in this regard (see ref. 39 for a discussion on scale sensitivity of drivers in particular). We identified variables that appeared in research, policies, and assessment tools and termed them consensus variables.

Although the number of research articles reviewed was far greater than the number of policies and assessment tools reviewed, we do not believe that this contributed to a bias in the alignment and comparison. On average, research articles contained four variables (ranging from 1 to 31 per paper), whereas the policies reviewed covered as many as 232 indicators (as in the SDGs) and the assessment tools covered an average of 60 (ranging from 25 to 116). Thus, it was necessary to review many research articles to reveal the full scope of variables considered in agricultural land systems research in Europe.

Lastly, we identified dominant approaches to agricultural land systems research in Europe using a suite of multivariate analyses. Two multivariate hierarchical cluster analyses were employed to determine distinct groups of research articles based on the variables included in each study. We then used three quantitative techniques to determine which categories of driver, management choice, or outcome variables were contributing to the distinction among cluster groups (see SI AppendixSupplementary Methods for full details). The statistical significance and validity of cluster solutions was also evaluated (SI AppendixSupplementary Methods). The focus of these dominant approaches to research was then compared with the categories of consensus variables identified to determine whether the prevailing research paradigms aligned with policies and assessment tools.