Globally, agricultural production occurs at vastly different scales, from massive corporate or government owned industrial agricultural farms to small-scale farmers working to produce sufficient food for their families. These small-scale farmers cultivating <10 ha (referred to hereafter as smallholders), often have access to the fewest technologies and financial safety-nets but still produce close to 30% of the world's food (Ricciardi et al., 2018). Here we propose an enhanced, quantitative engagement of agronomists and agricultural development experts with health and social science communities to consider the needs, desires, and behaviors of individuals alongside strategies to increase yields and reduce human labor inputs to quantify context and barriers to agricultural adoption. By incentivizing researchers to extend agricultural technical and conceptual approaches (hereafter ag-approaches) developed in agriculture science to the impacts experienced by households and individuals, we can accelerate the transformation of the global food system (Pretty, 1997).

Farmers employ countless management strategies to hedge some of the risk inherent in areas dependent on variable rainfall or other weather hazards (Bhatta and Aggarwal, 2016; Sibhatu and Qaim, 2017). These strategies are often based on tradition or experience which may, or may not, align with existing scientific knowledge on how to adapt to rapidly changing conditions. Unanticipated and unmitigated climate risk is a primary driver of both short- and long-term food insecurity, as it reduces uptake of new agricultural technologies and can lead to negative livelihood impacts well after a period of climatic stress (Hansen et al., 2022).

There are many ways to develop specific policies that support the transformation of climate information into multi-layered agricultural programs, including anticipatory action and index-based insurance which supports governments and individuals in adopting new agricultural practices (Hansen et al., 2022). For example, the World Food Program's R4 Rural Resilience Initiative provides access to subsidized drought and flood risk insurance products which are triggered with a precipitation index. The program allows vulnerable households to access agricultural insurance while supporting community-led disaster risk reduction and landscape restoration activities (Spiegel and Satterthwaite, 2013). Connecting agricultural technologies to insurance, increased yield or other development goals requires rigorous modeling and assessments of social, economic and productivity outcomes. However, many of these programs offload assessment of performance of their program to models of their choosing, without rigorous insight into their quality, relevance, or connection to livelihood improvements (Saltelli et al., 2020). The failure of agricultural development programs may come from incomplete model selection, unfounded assumptions, lack of quality or quantity of relevant local data, or lack of insight into the context in which the implementation may occur.

Agronomists and agricultural development entities are doing vital work on developing new, high yielding varieties and agricultural management strategies to more sustainably increase agricultural productivity. These agricultural interventions are being developed with a focus on environmental sustainability (Rockström et al., 2017; Pretty et al., 2018; Peng et al., 2020; Pilling et al., 2020). Ag-approaches can be used to understand how transformation of heterogeneous smallholder systems into those that regularly and consistently produce marketable food and sustainable rural livelihoods can be achieved. Increased agricultural production through intensification needs to be balanced with environmental and social considerations (Hoffmann et al., 2015). Despite impressive technical advancements made in agronomy expertise, downstream impacts on people's lives are difficult to consistently achieve and document, especially in low-income settings (Di Prima et al., 2022). Only a few examples have been documented on communities that have been able to attain long-term behavioral change in smallholder farm management (Cui et al., 2018; Stevenson et al., 2019), despite decades of investment. This is worsened by the critical lack of ground data observations (Saltelli et al., 2020) of agroecological and socioeconomic heterogeneity of smallholders needed to tailor farming practices, as not all practices are universally beneficial (Stevenson et al., 2019).

Social considerations, such as attitudes, preferences, and behaviors of stakeholders in rural areas dominated by smallholder agriculture livelihoods need to be examined while researching sustainable agriculture interventions (Ban et al., 2013). Stakeholders within agriculture settings include not only farmers and their families, but also the community, retailers, input providers, wholesalers and consumers (Brown et al., 2022). Explicitly considering stakeholders during research and planning of sustainable agriculture interventions should allow any suggested interventions to be more realistic and inclusive, informing complex choices by farmers on which crops (Lemos and Morehouse, 2005), with what inputs and with how much investment (Hirsch et al., 2011).

Ag-approaches are integral to developing policy-relevant scenarios to understand the impact of interventions like climate services (Hansen et al., 2007), ag-tech tools (Oyinbo et al., 2020), or agricultural insurance (Osgood et al., 2018). Farm management components that include both tactical and operational decision-making (Fountas et al., 2006) represent key factors where interventions and investments can be planned. Nevertheless, as Siddique et al. (2012) points out, much of the knowledge derived from agronomic crop science has often resulted in reductionist approaches to agricultural management, where a single management factor is modified which results in yield or crop quality improvements. Multiple factors across diverse settings are rarely managed and modified together in ways that simulate the complexity of smallholder systems. If farmers are to benefit from agronomic crop science focused on sustainability, it must be integrated into an overall crop management process, accounting for interactions between factors and incorporation of human, social and economic constraints.

We conducted two literature reviews focused on representing the most influential knowledge about agricultural interventions. We identify the most cited 50 publications and the most relevant 50 papers of the last 5 years using the methodology of Nagendra et al. (2018). Our objectives were to synthesize the literature on interventions in smallholder farming systems, and to summarize the current state of knowledge and identify the different domains from the Sustainable intensification Assessment Framework (SIAF) (Musumba et al., 2017; Stewart et al., 2018).

A search was conducted on April 21st, 2022 in the Web of Science database using the keywords “Interventions,” “Innovations,” “Agriculture,” “Smallholders,” “Food Security,” “Sustainable Intensification,” and “Modeling”. The results were filtered according to the following criteria:

To analyze the papers' contribution to interdisciplinary research on agriculture and nutrition, we evaluated the 100 papers on whether they included mention of the five domains of sustainable agriculture, as described below:

Figure 1 shows the lack of connection between research realms. There are significant connections between farming and human nutrition, and therefore human health but only a few approaches have tried to capture this interaction (Moberg et al., 2020; Di Prima et al., 2022). Furthermore, most of the publications that involved human domain focused mainly on nutrition, and especially on the calories consumed, disregarding the true complexity of utilization and access to a balanced diet with sufficient micronutrients and diversity (Lobell and Gourdji, 2012; Hasegawa et al., 2018). Lastly, most of the publications, when they included a social assessment, focused on economic concepts of market access and supply chains (Figure 1C; Horbach et al., 2012; Garrett et al., 2017; Ceballos et al., 2020). While Ag-economists focus on the cost/benefit and risk assessments associated with adopting new ag-methods, this kind of agriculture-oriented research is not well connected to broader social science and public health research (Griscom et al., 2017; Meemken et al., 2019; Adegbeye et al., 2020). In summary, research in this domain has neglected the real intricacy of the connections between biological systems and demographic aspects.

Researchers in social sciences and public health are considering the ways local food production impacts child health with a focus on expanding social science/health research in ways that consider the climate-food security linkage (Grace, 2017; Cooper et al., 2019). Anthropometric measurement (especially of children—e.g., stunting or wasting) allows for a quantitative assessment of an individual's health at the time of survey. From this information, researchers ascertain the level of undernourishment in a community and can use this information to spatially and temporally evaluate aspects of the food system that can be connected to biophysical data or agricultural models (Phalkey et al., 2015; Shively et al., 2015; Shively, 2017; Randell et al., 2021). Most studies that seek to connect environmental shocks and agriculture to human health use data (aggregated over space and time) on temperature, precipitation and vegetation anomalies, such as drought, heat waves and excessive rainfall (Brown et al., 2014; Randell and Gray, 2019; Sellers and Gray, 2019; Jain, 2020). Understanding how exposure to environmental shocks combined with the timing of exposure of individuals will help development programs to design appropriate interventions to protect human health and wellbeing (Grace et al., 2020; Hunter et al., 2021). This fairly general approach has been especially relevant in the context of climate change as scholars work to consider the linkage between the environment, food systems and health outcomes. However, the use of coarse environmental data fails to engage with the complexities and advancements in agronomic sciences and therefore misses an opportunity to advance cutting-edge science to support climate change adaptation.

A more robust alignment between agronomists and social science research communities is still needed to create more effective food system investments (Yaro, 2006). Here we propose enhanced, quantitative engagement that considers the needs, desires, and behaviors of individuals alongside strategies to increase yields and reduce human labor inputs to quantify context and barriers to agricultural adoption and connect them to food security outcomes. Therefore, we propose a new approach in agricultural, technical and conceptual research that links agriculture science to health, nutrition, and demographic impacts experienced by households and individuals. This can be achieved by encouraging researchers to Integrate process-oriented crop models with demographic and health models that can enable exploration of the impacts of specific interventions through attention to the biophysical and field management aspects affecting food production (Figure 2).

An integrated quantitative modeling framework that allows the use of information across all five domains in sustainable intensification will enable the identification of barriers, develop new insights and scenarios that could test the likely results of policy changes and intervention investments on food security and nutrition. Figure 2 shows how the community could use interdisciplinary engagement to accelerate planning for investments, highlighted in the models with red coloration. Primary outputs of crop models are estimates of crop yield and biomass at a certain site and year (Holzworth et al., 2018; Hoogenboom et al., 2019). From these, crop models can derive other outputs of interest to the food security community, including total calories, nutrient content, cost, and complexity of the agriculture system (Valin et al., 2014; Grafton et al., 2015). In this context, crop growth models arise as an effective tool to summarize how the biophysical environment affects a community, and focuses these results on human health information such as nutrition and health outcomes. By broadening the focus beyond process-based agronomic interventions, a more holistic approach can be promoted.

Integrated frameworks and models hold the promise to plan, implement and measure outcomes across a variety of contexts. Management factors in process-oriented Ag-Approaches are integral to developing policy-relevant scenarios to understand the impact of interventions and include crop and varietal choice, planting date, fertilizer and manure usage, weeding practices, field preparation, seeding rate, and sowing techniques (Cooper et al., 2009). For example, the impoverished women farming peanuts on undeveloped plots in urban Ouagadougou, Burkina Faso do not use the modern peanut varieties developed by local agronomists—rather, they rely on known and trusted approaches to grow peanuts as quickly and cheaply as possible to sell peanut butter in the market to meet their household budget (Juana et al., 2013). Ag-Approaches can be used to determine if an improved legume introduced to these women farmers will grow well under variable rainfall, high temperature or other weather scenarios, while acknowledging that the woman farmer will not always be able to plant on the idealized planting window. We could also determine if the new legume variety will still perform when the woman farmer loses access to other inputs due to macro considerations such as fuel prices, inflation or drastic changes in the input supply chain, or how it might be affected by changes in labor availability caused by a catastrophic health concern. The woman farmer's decision-making regarding legume choice involves productivity and flexibility simultaneously, among other concerns. If the crop model answers the first well, then investment might help with the second. Models could also show how productivity investments alone without simultaneous health and input investments will doom our woman farmer's cash crop.

Coupling agricultural process models with social and demographic models would enable improved exploration of how to transform low-input subsistence agricultural systems to achieve food security without replicating the unsustainable systems seen elsewhere (Schaller, 1993). This transition requires careful attention to the equity and political processes in the affected communities, the scope and potential of policy interventions, and the practices that result in nutrition and health outcomes. A more integrated modeling framework would allow for the use of modeling scenarios to evaluate potential outcomes and identify the unexpected outcomes that might emerge from biophysical or policy interventions.

Publicly available datasets were analyzed in this study. This data can be found here: https://doi.org/10.6084/m9.figshare.19763044.v1.

MB, KG, ME, AC, IC, and JN conceptualized and wrote the paper. AC and IC conceived and designed the analysis and collected the data. AC, IC, and ME performed the analysis. All authors contributed to the article and approved the submitted version.