Evidence from observational studies
A large number of small-scale observational studies assessing the links between agriculture and nutrition have been conducted in recent years, with many including a focus on cultivated (and reared) biodiversity at the household or farm level, and/or dietary diversity (Section 2.2.1 below). Some of these have also sought to assess the role of access to, and participation in, markets in mediating the relationship (Section 2.2.4). There has been significantly less attention to the role of wild food biodiversity (Section 2.2.2), and even less on the contribution of forests and agro-forestry systems to dietary intakes (Section 2.2.3), despite their potential to make a significant contribution to dietary diversity and nutrition in specific contexts.
2.2.7 Cultivated food biodiversity
Recent reviews of studies that have analysed associations between cultivated biodiversity (measured at the crop group or species level) and dietary diversity scores (as proxy indicators of diet quality, measured at the food group or food type level) have found evidence of a small, positive association, at least among poor, isolated households and communities with limited market access or imperfect market structures (Powell et al., 2015; Jones, 2017; Ruel et al., 2018; Sib- hatu and Qaim, 2018a). There is also limited evidence of an association between dietary species richness (in addition to dietary diversity) and adequate intake of multiple micronutrients where, for every additional species consumed, dietary nutrient adequacy increases (Lachat et al., 2018).
Given the small effect sizes observed, the nutritional significance of the association, even in settings of very low agricultural diversity, is unclear. Jones (2017) cautioned that potentially large and unrealistic increases in cultivated biodiversity may be required to have a nutritionally meaningful impact on dietary diversity.
The relationship between cultivated biodiversity and dietary diversity also appears to be highly context-specific (Jones, 2017; Ruel et al., 2018). At very low levels of biodiversity, a marginal increase in cultivated biodiversity has a much greater effect on diet diversity than on farms with moderate to high levels of production diversity. In contexts where on-farm biodiversity is already high, agricultural diversification may have minimal-to-no impact, and even a negative impact, on dietary diversity (Sibhatu et al., 2015; Jones, 2017; Ruel et al., 2018). Market access and participation also appear to be key factors that modify the relationship (see 2.2.2 below).
In general, this literature suffers from many of the same methodological weaknesses as the NSAP literature, including: a lack of appropriate controls or comparison groups, as well as accounting for potential mediating and confounding factors; and significant heterogeneity in measurement approaches and indicators used, making comparisons between studies difficult (Ruel et al., 2018).
2.2.2 Market access and participation
Even among subsistence-oriented households and within the context of traditional food systems, there is interplay between own production, sale to and purchase from markets and diversified consumption, with both own production and market participation being important to diets of smallholder producer families
(HLPE, 2017). Market participation refers to the market-based sale, purchase or acquisition of foods through retail markets (ranging from small informal local markets and small neighbourhood shops/kiosks to supermarkets, hypermarkets and ready-to-eat food retail), food transfer programs, bartering or food sharing.
Access to, and integration into, markets has the potential to improve food security, income, dietary diversity and nutrition and health outcomes (Darrou- zet-Nardi and Masters, 2015; Reardon et al., 2015; Bellon et ah, 2016; Sibhatu and Qaim, 2018b). On the other hand, greater market access has also been associated with the nutrition transition, which is characterized by greater consumption of processed, energy-dense foods, and the gradual shift from a high burden of undernutrition to rising rates of overnutrition, overweight and diet-related chronic diseases (Baker and Friel, 2016).
The small, positive association between cultivated biodiversity and dietary diversity appears to operate independently of market access and relative market- orientation of a household/farm. However, market access and participation (in addition to the extent of existing on-farm diversification), appear to be important effect modifiers ([ones, 2017; Ruel et ah, 2018). There is relatively consistent (albeit, again, limited) evidence that market access is independently, positively associated with dietary diversity, at least among poor, rural communities in low and middle-income countries (Remans et ah, 2011; Sibhatu et ah, 2015; Bellon et ah, 2016; Jones, 2017; Koppmair et ah, 2017). Evidence on the role of market access among other population groups and other settings is very limited, although it appears that the influence of market access is likely to be highly context-specific (Jones, 2017).
Market access and participation have largely been measured in this literature using simple proxy indicators of market access, such as travel time or distance to closest market, road or population centre. Relative market-orientation (or commercialization) of a household’s agricultural production may be a more direct proxy indicator of market participation, with some limited evidence that households with farms that are at least partly market-oriented (typically measured as proportion of cultivated area devoted to non-food cash crops) have more diverse diets than less market-oriented farms (Jones, 2017). However, the evidence is too limited to determine the nature, and even direction, of this relationship. A mixture of cultivated biodiversity along with enhanced infrastructure for improved market access is likely to be the optimum combination for improving dietary diversity (Ickowitz et al., 2019).
2.2.3 Wild food biodiversity
Food biodiversity can also be gathered from forests, water bodies and uncultivated areas within diverse agricultural landscapes. Wild food biodiversity includes plants, other forest products (for example fruit, nuts, mushrooms and bush meats), insects and aquatic species. Many wild foods are collected from within, or at the margins of, agricultural production systems, and some foods are considered semi-wild, or semi-domesticated (Halwart, 2006; Bharucha and Pretty, 2010).
Wild foods can be much richer in nutrients and food components than many domesticated species (Burlingame et al., 2009; Bharucha and Pretty, 2010; Ter- mote et al., 2014; WHO and CBD, 2015). They tend to be resilient to harsh conditions and have significant potential to contribute to the nutrient-density of diets year-round, provided they are available, safe, consumed in sufficient quantities and the nutrients are bioavailable. Wild foods play an important role in contributing to food security and livelihoods for millions of people worldwide (Bioversity International, 2017; Ickowitz et al., 2019), although their actual consumption, contribution to diets relative to other foods, and nutritional significance within diets is elusive and varies widely between populations, settings and seasons. Globally, consumption of wild foods appears to be declining due to a combination of urbanization, costs, habitat loss and changing dietary preferences (Tennote et al., 2012; WHO and CBD, 2015).
Evidence on the contribution of wild food biodiversity to diets and nutrition is limited by a severe lack of food composition data for many neglected and underutilized cultivated and wild foods (Termote et al., 2014; de Bruyn et al., 2016), as well as practical challenges in measuring wild food consumption and contribution to the diet relative to other foods (Bharucha and Pretty, 2010).
There is some evidence that wild fish in particular make an important contribution to dietary diversity and nutrition among poor populations in low- and middle-income countries for whom seafood products have traditionally been by far the most frequently consumed animal-source foods. In these contexts, the expansion of aquaculture (which is typically characterized by a limited number of large commercial species) has been associated with reduced diversity of fish species consumed, along with reduced intakes of essential micronutrients (including zinc, iron and calcium) and fatty acids (Belton and Thilsted, 2014; Belton et al., 2014). Bogard et al. (2017) similarly found that despite an increasing trend in fish consumption in Bangladesh, which has been attributed to commercial species (mainly carp and tilapia) replacing non-farmed species, there was a decline in intakes of iron, zinc and calcium from fish.
2.2.4 Forests and agro-forestry systems
Despite their diminishing significance in terms of land area and populations, the majority of the world's wild biodiversity is found in and around forests, particularly tropical forests. In addition to wild foods, forests and trees provide essential ecosystem services that support agricultural production (Millennium Ecosystem Assessment, 2005b) and potentially, therefore, cultivated food biodiversity.
Several studies that have paired satellite imaging with dietary data to assess the association between forest and tree cover and various metrics of diet diversity and quality, have identified positive associations between forest and tree cover, and dietary diversity in children (Johnson et al., 2013; Ickowitz et al., 2014). In 15 Demographic and Health surveys from sub-Saharan Africa, Jones et al. (2017), found an association between deforestation and lower diet diversity of children. In Indonesia, mixed mosaic/agro-forestry landscapes have been associated with more frequent consumption of the largest number of micronutrient- rich food groups (Ickowitz et al., 2016). However, the authors recognize that the associations are complex and were not evident until analyses were undertaken on provincial as compared to aggregate national level. There are also challenges in distinguishing between wild and cultivated food biodiversity from forests.
The hunting, harvesting and use of certain wild foods, particularly bush meats, forest products and fish, raises issues relating to sustainable natural resource management and conservation (Bharucha and Pretty, 2010; Powell et al., 2015). However, wild animal-source foods can be an essential source of micronutrients that might otherwise be lacking in local diets (Golden et al., 2011). Harnessing traditional knowledge and cultural practices relating to wild foods and landscape management has been proposed as a strategy to simultaneously support sustainable land management and food systems and improve nutrition (Powell et al., 2015), although there has been limited research in this area and questions remain as to how scalable this approach is to improve food security and nutrition in increasingly urbanized populations (Cogill, 2015).
Evidence from modelling studies
Diet modelling with linear programming has been used to identify optimal diets in a range of contexts and for a range of nutrition purposes, including: complementary feeding for infants and young children; emergency and therapeutic feeding and food aid; national food programs; dietary guidelines; promotion of the use of locally available nutritious foods, including wild foods; and modelling of population and individual diets that meet specific nutrient, environmental and cost constraints (Maillot et al., 2008; Maillot and Drewnowski, 2011; WWF, 2011; Macdiarmid et al., 2012; Horgan et al., 2016; van Dooren, 2018; Willett et al., 2019; Swinburn et al., 2019).
Food composition studies have demonstrated the significant variation in nutrient content both between and within species (for example between different types of fruits, such as mango, banana and apples), as well as between different cultivars and traditional varieties of a single species (for example bananas) (de Bruyn et al., 2016; Bioversity International, 2017). There is emerging work using optimization models to demonstrate how incorporation of this food biodiversity can improve micronutrient density of the diet (Levesque, unpublished; Ekesa et al., 2019; Wessells et al., 2019). To date, this evidence has been used to develop nutrition education materials. In Benin, for example, a recipe book has been developed to demonstrate the preparation and added nutritional value of complementary foods that incorporate diverse, locally available species such as niebe (Vigna unguiculata), a pulse common in Benin, and a wide range of locally available dark green leafy vegetables such as Corete/Crincrin (Corchorus olitorius), Moringa, (Moringa oleifera), Bidens pilosa or Grande morelle (Solatium macrocarpum) (Bodjrenou et al., 2018). However, impacts of these approaches, including on micronutrient adequacy of diets, are yet to be tested.