Forests play a role in many food systems, through direct and indirect provisioning for human nutrition and through ecosystem services. Globally, 800 million people remain hungry, while 2 billion are suffering from hidden hunger (micronutrient deficiencies). Consumption of wild foods from forest landscapes has a large potential to contribute to mitigating nutritional deficiencies ( Powell et al. , 2015 ). In most cases, these are due to a lack of diversity in the diet ( Lachat et al. , 2017 ) and a poor supply of micronutrients ( Black et al. , 2013 ). An increasing number of studies has documented the contribution of wild forest foods to the diet and suggested a link between forest cover and dietary quality ( Johnson et al. , 2013 ; Ickowitz et al. , 2014 ). Assessments of actual quantities of forest foods consumed have been carried out ( Termote et al. , 2012 ; Powell et al. , 2013 ; Rowland et al. , 2016 ) and it has been found that their contribution to the diet varies considerably. Semi-disturbed forested environments with anthropic influence and mixed land use types ( e.g. , swidden agriculture, agroforestry, natural forest, tree crop plantations) seem to be associated with higher frequency of consumption of food groups rich in micronutrients ( Ickowitz et al. , 2016 ).

Research on consumption of wild foods in dry forest areas, and on their nutritional importance, are very few ( Rowland et al. , 2015 ) despite the large extension of dry forests and their relatively large supply of edible products compared to alternative land uses in the same regions ( Wunder, 2001 ).

Seasonality is a key element of food availability, especially in the Sahelian countries, and the contribution of food tree species to the diet is particularly important during periods of food scarcity or extreme drought ( Faye et al. , 2010 ; Atato et al. , 2011 ; Agúndez et al. , 2016 ).

In West Africa, food tree species are valued by lo­cal people. Preferred trees are usually multifunctional ( Gijsbers et al. , 1994 ; Faye et al. , 2010 ) and sup­ply edible products (Kris­­tensen & Lykke, 2003; Kristensen & Balslev, 2003 ). When the land is cleared for agriculture, favoured trees are retained and their natural regeneration in farmed fields is protec­ted ( Ræbild et al. , 2011 ). Indigenous communities appear to have favoured edible-fruit-yielding species normally found in the wetter Sudanian and Guinean phytogeographic zones ( Maranz, 2009 ). Their density tend to be generally higher in farmed versus non-farmed areas ( Maranz & Wiesman, 2003 ).

Based on increasing evidence, dry forests in tropical areas are vulnerable to several threats. Tree species richness and density declined in the West African Sahel in the second half of the 20th century ( Gonzalez, 2001 ). About 97% of the remaining area of tropical dry forest is at risk from one or more of threats, with the largest effects attributable to habitat fragmentation and fire ( Miles et al. , 2006 ).

In West Africa, dry forest cover coincides with areas of high population density. Southern Burkina Faso has experienced rapid population growth, mostly determined by immigration of farmers. Cropland has expanded at an annualized rate of 0.46% at the expense of forest cover, which decreased by 0.57% per year. In the same areas, migrant population has increased from 3% in 1976 to 57% in 2007 ( Paré et al. , 2008 ; Ouedraogo et al. , 2009 , 2010 ). In some regions, this has determined a considerable exploitation of non-timber forest products (NTFPs) and an associated decline in the density of some tree species ( Boffa, 1999 ; Kristensen & Balslev, 2003 ; Lykke et al. , 2004 ; Paré et al. , 2010 ), with potential consequences on various ecosystem services, including nutrition security ( Jones et al. , 2017 ).

The views of local people are central in providing information on general trends in vegetation dynamics and particularly on changes in occurrence of rare species ( Lykke, 1998 ). Preferences of local people are also crucial in the definition of locally acceptable management solutions.

A first objective of this study was to identify tree species that provide edible products to rural communities during the lean season across two ecoregions of Burkina Faso. A second objective was to understand what household (HH) characteristics ( e.g. , ethnicity, origin, ecoregion, level of education) had most influence in determining the observed patterns of consumption of edible tree products. Finally, based on the perception of local key informants, the main threats affecting the most consumed food tree species were documented, with the objective to assess 1) how threats varied across the ecoregions and villages investigated and 2) whether species-specific threats could be identified. The ultimate objective was to single out priority food tree species, relate these to their main threats and determine if implementation of conservation and mitigation measures would require adaptation to highly diversified contexts.

The list of 25 indigenous tree species, mentioned as edible and consumed during the lean season shows that, during the lean season, edible products derived from trees are a significant supplement in the diet, mainly based on cereals. Leaves and fruits are particularly effective in diet diversification due to their rich content in micronutrients ( Bvenura & Sivakumar, 2017 ).

The knowledge available in the literature on important food tree species in Burkina Faso is usually derived from small-scale case studies. The sampling in this research was designed to capture a representative diversity of contexts across phytogeographic regions and provinces with different ethnic composition. Nevertheless, the priority species that emerged from this study seem to be aligned with those derived from other ranking exercises carried out in Burkina Faso ( Kristensen & Lykke, 2003 ; Lykke et al. , 2004 ; Thiombiano et al. , 2012 , 2014 ). A total of 12 out of the 25 species identified as important in this study are considered among the top most useful plant species in Burkina Faso ( Zizka et al. , 2015 ), all characterized by multiple uses (up to eight different uses for A. digitata ). Interestingly, only a few species emerge for their considerably greater importance in the diet, A. digitata in particular, and their edible products are consumed by a very large percentage of HHs. Overall, only three tree species ( A. digitata, P. biglobosa and V. paradoxa ) are consumed by a large majority of HHs.

This gap between availability and ac­tual con­sumption has been observed in other envi­ronments ( Termote et al. , 2012 ; Boedecker et al. , 2014 ). An explanation for this has not been provided but the pattern observed could be due to the fact that highly consumed species are actively favoured and maintained in the landscape by human activities, thus they are easier to access and collect, and their edible products can also be found in the market. On the contrary, various factors may constrain consumption of less common wild species and contribute to reduce their usefulness. For example, the excessive distance to be covered to harvest less common wild species ( Thomas et al. , 2009 ; Termote et al. , 2012 ), the significant workload needed to process them into edible products, the progressive change in dietary habits in younger generations who move away from traditional foods ( Lykke et al. , 2002 ), the declining knowledge about traditional plants, the progressive loss of habitat of some wild species ( Barucha & Pretty, 2010 ) and the lack of information on their nutritional properties ( Waswa et al. , 2015 ), are all factors that may contribute to limit interest and consumption.

The factors that most closely explain the patterns of consumption of edible tree products in this study are phytogeographic region and ethnicity. The average number of tree products consumed was slightly higher in the Sudanian villages (5.5 tree products) compared to the Sahelian villages (4.35 tree products), however the average number of tree species used was quite similar (4.14 in the Sudanian and 4.17 species in the Sahelian zone). HH hunger score was significantly higher in the Sudanian compared to the Sahelian zone and this could be possibly due to the different quality of the last harvest, which was rated as good by villagers in the Sahelian zone and bad by those in the Sudanian zone. The average number of tree species used did not depend on any of the independent factors entered in the Poisson regressions (ecoregion, socio-economics, HFIAS, HHS), while the number of tree products used was significantly correlated with phytogeographic region as well as ethnicity. This indicates that the number and type of species/products consumed in each village depend not only on local occurrence of a species and seasonality of the parts consumed as food, but also on a mix of cultural factors that are correlated with the spatial distribution of ecoregions. This is in line with results from a study about differences in the perceptions and valuation of woody plant species in three ethnic groups (Fulani, Mossi and Samo), in the sub-Sahelian region of Burkina Faso. The latter study revealed that the most important species, identified for conservation priority, based on the informants' perspective, were food species with a high socioeconomic value: V. paradoxa, A. digitata, Tamarindus indica, P. biglobosa, L. microcarpa and Ziziphus mauritiana ( Sop et al. , 2012 ) and preferences were similar for all ethnic groups, though plant use differed among ethnic groups, with age of respondents influencing plant knowledge in all three ethnic groups.

Some species are progressively disappearing. A review across West African countries (Burkina Faso, Mali and Niger) by Wezel & Lykke (2006) revealed that according to local perceptions 79% of the woody species mentioned were classified as having decreased or disappeared. An assessment of local perception of changes in distribution of socio-economically important tree species in the Sahelian region of Burkina Faso in three ethnic groups (Mossi, Fulani and Samo) ( Sop & Oldeland, 2011 ) revealed a decline in occurrence of more than 80% of the 90 listed species; 40% of these were considered under threat, mainly from drought, deforestation and bushfires. Undoubtedly, the combination of future climate scenarios and land use changes is expected to have a strong negative impact on the flora of Burkina Faso ( Heubes et al. , 2013 ).

It is yet not clear what drivers of change (whether climate change, human exploitation or livestock grazing) are leading to the disappearance of tree species and a senescing of savanna in the Sahel. In this study, overall, the most prominent threats cited by the experts interviewed tend to be less directly linked to the influence of human exploitation ( e.g. , pests and diseases, aging of trees and loss of soil fertility). However, these threats are indirectly associated to particular practices ( e.g. , the lack of regeneration is most likely related to grazing pressure).

It is important to note that those species considered highly threatened are the most important in the diet. The high value of these priority species, combined with a considerable competition for their exploitation, may induce a perception of scarcity and availability lower than the demand, as illustrated by Thiombiano et al. (2013) , who showed how market channels largely mitigate lack of local supply of the most demanded species.

It is interesting to note that none of the species indicated as scarce or disappearing by the villagers involved in this study is considered critically endangered in Burkina Faso at country level, according to the checklist recently developed by Schmidt et al. (2017) . These species may be locally scarce or lost due to localized overexploitation. The findings from this study indicate that threats vary considerably across sites; forest clearing for agriculture is most pronounced in the South of Burkina, while livestock grazing in posing great pressure in the northern part of Burkina Faso, especially where pastora­lists have become progressively more sedentary ( Lykke et al. , 2004 ).

Regarding predominant threat types emerging from this research, the situation is indeed spatially patchy, as revealed by the diversity of threats profiles by village. In the Sudanian ecoregion, particularly in the Southern and South Western part of Burkina Faso, the main threat appears to be fire. The villages where fires is the most significant threat, Péni in particular, but also Barcé and Sara, are situated in the region neighboring the second largest city in Burkina Faso (Bobo-Dioulasso). These results seem to be validated by evidence from records on fires frequency made available by the NASA (2012) . Furthermore, a spatially explicit threat analysis for food tree species in Burkina Faso ( Gaisberger et al. , 2017 ) revealed that the areas with the highest incidence of fire are in the Southern and South Western parts of Burkina Faso. The high population density in these areas is associated to large immigration, especially from drier areas in the north. This is confirmed by the ethnic composition of the three selected villages located in the Sudanian ecoregion, characterized by a high presence of immigrants ( ca. 68%) significantly greater than in the Sahelian villages.

Immigrants normally occupy forested areas and clear them using fire, to make space for new cropland, thus this practice could explain the patterns observed. Forest fires are considered a significant factor contributing to deforestation in Burkina Faso, despite the large attention to this challenging issue in the existing legislation ( Kalame et al. , 2009 ). Specific management strategies are needed to contain the effects of fire on the vegetation, such as, for example, the adoption of longer intervals between fires. Droughts and aging are critical factors in the Northern part of the country, where precipitations are less and regeneration is generally more seriously compromised by grazing, due to a higher presence of pastoralists. High value species may experience particular threats across most of their range; for example regeneration of P. biglobosa is largely compromised by intensive seed harvest, so that only very few seeds germinate and survive browsing by animals ( Ræbild et al. , 2011 ). For various useful tree species, it has been observed that where regeneration is present, in all land-use types (forest, fallow, field), small trees seem unlikely able to develop into mature tree individuals due to grazing, so a demographic bottleneck is affecting most species ( Ouédraogo et al. , 2015 ).

This study highlights the important role of edible products harvested from trees in diversifying diets of rural communities in Burkina Faso during the lean season. The number of edible products consumed varies according to phytogeographic region and ethnic composition of the community living in a particular site. This study shows that overall a few tree species play a disproportionally greater role in the diet and are characterized by a very high frequency of consumption by the majority of households in both phytogeographic regions and across ethnicities: Adansonia digitata, Parkia biglobosa and Vitellaria paradoxa . Although these species are not listed as critically endangered or vulnerable at country level, this study reveals that they are perceived as threatened and could become scarcely available locally due to the great competition for their consumption but also due to a variable combination of threats. Some tree populations with critical traits adapted to specific environmental conditions may be lost and limit future adaptive capacity of the species. Our findings indicate that effective conservation efforts are highly needed for these priority species. However, considering how nutrient poor is the diet of the communities investigated, it would be important to explore the nutrition potential of largely underutilized tree species. Fire seems to be a critical driver of change in the savannas of the Sudanian phytogeographic zone, while drought is affecting indigenous tree species in the Sahelian zone. Our results suggest that threats vary across regions and mitigation measures need to be highly diversified and adapted to local pressures.

Estelle Bader (external consultant, nutritionist) provided great support in cleaning and reorganizing the dataset and carried out preliminary statistical analyses.