Future climate and land use changes challenge current dependencies on wild food harvesting by rural indigenous communities

Abstract

Traditional food systems support the livelihoods and well-being of rural Indigenous communities, particularly in remote, asset-poor areas. However, the diversity of wild foods is in global decline under the accelerating impacts of climate and environmental change with major but poorly understood implications for dependent communities. Here, we combine a comprehensive systematic household survey involving 400 households from 18 rural Indigenous settlements across the Republic of Sakha, a vast and climate change sensitive region in the Russian Far East, with species distribution models for 51 food species of animals, plants, and fungi to (i) profile current household dependencies on wild food harvesting; (ii) project future (2050s) changes in the regional distribution and local availability of wild foods under alternative climate and land use change scenarios; and (iii) discuss their combined potential implications. We find that current dependencies, understood as shares of the total food consumed and income by household, are on average relatively low across settlements, albeit with important regional variability. Remote and isolated settlements in the Arctic region of the Republic of Sakha have greater levels of dependency with stronger links to animal products, while those in the southern and central regions, which are better connected and closer to major urban areas, have lower levels of dependency and are dominated by nonanimal products (plant-based products and fungi). These dependency patterns contrast with projected changes in the regional distribution and local availability of food species, signaling major turnovers of species with important potential implications for dependent rural livelihoods.

Introduction

The traditional food systems of Indigenous peoples and rural communities (hereafter referred to as TFSs) play a key role in promoting and maintaining food security (1), supporting health and livelihoods (2), conserving biological diversity (3), ensuring the long-term sustainability and provisioning of ecosystem services (4), and conferring resilience to climate change and natural hazards (5). Strengthening the understanding, promotion, and preservation of TFSs and the traditional knowledge associated with them is critical for achieving sustainable development and climate mitigation on global agendas (6, 7). However, our current understanding of the multifaceted contribution of TFSs to rural indigenous livelihoods and how this might be compromised by future environmental and climatic changes remains incomplete and geographically biased (7).

TFSs are increasingly exposed to and impacted by the growing cumulative effects of climate change and other existing environmental impacts that are altering the habitats and resources of many wild and cultivated food species, leading to changes in the behavior, phenology, abundance, and distribution of these species (8, 9). Despite existing uncertainties and regional variability, there is wide consensus that such changes will intensify in the future, compromising the food provisioning of dependent communities (10, 11), ecosystem functioning and services (12, 13), and human welfare (14). These impacts will disproportionately be felt by rural communities in low-income and marginal areas whose well-being and livelihoods rely heavily on wild harvesting (15, 16).

As one of the global regions experiencing the fastest climatic and environmental changes, the Circumpolar North (covering Arctic and Subarctic regions) is particularly vulnerable to their impacts on TFSs (7, 17). The region is also home to a vast diversity of Indigenous Peoples (IPs) whose diverse TFSs, typically comprising species of plants and animals in the hundreds, represent a cornerstone of their livelihoods, cultural identities, and sense of place (18, 19). Indigenous ways of life are well adapted to the harshness and variability characteristic of the natural ecosystems in these extreme regions, and have an acute awareness of the need to ensure the sustainable harvesting, trade, and use of the wild species residing in them (19). However, climate change in combination with ongoing human-induced environmental changes (e.g. land transformation and fragmentation, environmental pollution) is increasingly altering the availability of and access to wild food species impacting their diets and traditional lifestyles (20–22).

Meanwhile, debates continue about the extent to which improved access to services, infrastructure, and markets can decouple well-being from nature and offset related disruptions and restrictions to wild harvesting (23, 24). While the association between wild harvesting and well-being is typically strong in remote and income-poor areas, with environmental income representing an average of 28% of total household income in developing countries (16), people's dependence on nature is contextual to many dimensions of well-being, including access to infrastructure, cities, assets, and skills (25). Indeed, a growing body of research suggests that mixed household economies are not transient but are becoming increasingly common in many remote rural areas from the tropics (25) to the Arctic (24). However, detailed assessments of the extent to which rural indigenous livelihoods depend on provisioning from wild ecosystems and the potential implications of changes in the local availability of those resources in the context of future environmental change are still lacking, particularly for remote, asset-poor rural areas. To this end, integrative multidisciplinary approaches can provide useful perspectives (6).

Here, we combine the results of two rounds of comprehensive systematic household surveys, comprising a general survey on 400 households from 18 rural settlements across the Republic of Sakha (RS) (Fig. 1, Table S1) followed by an in-depth survey on 185 households from two representative settlements with projections of the future redistribution of 51 wild food species (Table S2) using species distribution models (SDMs; see section Materials and methods) to (i) understand the current dependence on wild harvesting of rural households in terms of food consumption and income; (ii) project future (2050s) changes in the regional distribution and local availability of wild food species under alternative climate and land use change scenarios; and (iii) discuss the potential implications of those changes for livelihoods given existing current dependencies. The RS, the world's largest administrative subdivision (3.084 million km²) comprising approximately half of the Russian Federation's Far Eastern district, provides an ideal scenario for our analysis because of its diversity of Indigenous Peoples (comprising 54% of the total population; a large proportion of which lives in rural areas), rich cultural, ecological, and geographic diversity, and because it is a region experiencing rapid environmental changes related to climate change and land use transformation (26, 27). We apply the definition of TFSs provided by Kuhnlein et al. (28) as “all food from a particular culture available from local natural resources and culturally accepted” to the existing systems of traditional hunting, fishing, and gathering that complement those of farming and animal breeding to provide a subsistence base from available natural resources. Specifically, we focus on the wild food species of plants, fungi, mammals, and birds consumed by the rural communities of the RS. Recognizing that local species and varieties of crops and farmed animals also represent an important component of TFSs, we focus exclusively on traditional wild food species because we are interested in evaluating their contribution to the economy of rural households and understanding the potential impacts of future climate and land use changes to which they are particularly vulnerable (7).

Results

Dependency on wild food harvesting

Based on the data collected from the two-stage systematic household surveys (see section Household surveys in Materials and methods), we measured the average household dependence on wild food harvesting (i.e. wild food species hunted, fished, or gathered by rural communities) from a dual perspective of the dietary (based on consumption frequency) and income share of wild food species. Income in this study covers both the actual monetary incomes and self-consumption of wild and nonwild farming items so that the income-based dependencies reflect living standards, not just cash sources (see Dependency on wild food harvesting in Materials and methods).

In terms of the overall share of all food items consumed by the average household, we found a modest dietary (8.7 ± 8.4%; mean ± standard deviation) wild food dependence across all study settlements. Berries, nuts, and fish made up approximately 65% of the overall wild food diet share across settlements (Table S3). Nonetheless, patterns of dependency were subject to large variations across settlements (Fig. 2A, Table S3) according to reported household consumption patterns (Table S4). With a few exceptions, settlements located in the central and western regions of the RS showed low overall dietary dependence. Conversely, dietary dependencies were the highest in remote settlements such as northern settlements of the Arctic region. Stark differences in dietary shares among settlements were also apparent when examining relative contributions by food group. Whereas wild mammal (35.66 ± 23.46%) and/or fish (44.17 ± 27%) products clearly dominated the diet in Arctic settlements, patterns of share contribution by food group were much more idiosyncratic in the other regions, with no apparent trends other than perhaps a dominance of nonanimal food products in settlements of the central and western regions (Table S3).

Relating dietary dependency on wild food to the total food consumed by food group (as opposed to all foods consumed) revealed a greater contribution of wild foods to the diet of rural communities, highlighting their nutritional importance (Fig. 2C). Wild fish (77.5 ± 26.12% across all settlements) and fungi (80.18 ± 33.73%) contributed over half of the total fish and fungi consumed annually in the average household for nearly all settlements. Other food groups contributed less (mammals, 7.18 ± 11.77%; birds, 11.05 ± 16.3%; vegetables, 6.68 ± 16.71%; berries and nuts, 36.12 ± 29.27%) and were associated with greater geographical variation (Fig. 2C). For example, whereas wild mammals consistently represented approximately a quarter of the food of mammal origin consumed by Arctic settlements, their contribution to most settlements of the other regions was negligible. Conversely, wild plant-based foods had greater contributions to their respective food groups in settlements from the central and western regions than in those from the other regions.

Income-based dependencies were assessed during the second stage in-depth surveys in two selected settlements, Kharyalakh and Rassoloda, representative of the Arctic and non-Arctic regions, respectively (see section Study area and settlements in Materials and methods). Both settlements had similar income levels and a share of the total income by household strongly dominated by wage- (including businesses) and welfare-associated (e.g. social security and subsidies) income, which together represented more than 80% of their total income share (Fig. 2B). However, the distribution of the remaining share was clearly different between them. In Kharyalakh, an Evenks community in the Arctic region, wild foods represented a sizeable contribution (11.1%) to the total average household income. Mammals and fish accounted for 96.9% of the wild food share in terms of aggregated wild food revenue across all households (see Materials and methods), whereas farming accounted only for a 0.3% of share. Conversely, in Rassoloda, a Sakha community of the center region, these figures reversed to 3.6 and 13.7% for wild food and farm-related income shares, respectively (Fig. 2B). Here, nonanimal foods (mostly berries) represented over two-thirds of the wild food revenue, while mammals accounted for only 1.1%.

Future redistribution of wild food diversity

Modeled changes in the current distribution of 51 wild food species based on projected climate and land use changes by mid-century showed a strong decrease in species richness in the southern half of the RS (Fig. 3). Importantly, decreases were smaller, more geographically contained and accompanied by greater increases in richness in the northern half of the RS under the low-emission scenario (SSP1.26). Variability in projections was the highest in the southern half of the RS (Fig. 3), reflecting the underlying variability in climate parameters associated with the different general circulation models. Among the species groups, an SW‒NE pattern of species redistribution was apparent for all groups (Fig. 4 and Fig. S1). The largest and smallest shifts were respectively projected for birds (30.74 ± 21.41 and 44.87 ± 22.82 km/decade for SSP1.26 and SSP5.85 scenarios, including migratory and resident species), and fungi (13.37 ± 5.55 and 18.8 ± 6.51 km/decade) (Table 1), albeit with large variability at the species level within each group (Table S5).

The implications for the current dependencies of rural communities on wild food species hinge on how these broader regional shifts will eventually reshape the patterns of local diversity and the composition of the wild food species available to them. We calculated the projected changes in species habitat suitability (irrespective of projected species occurrence) and occurrence (based on suitability thresholds and dispersal constraints) for the assemblages of wild food species within a radius of 100 km around each of the study settlements, considered as the approximate maximum distance at which wild harvesting is practiced throughout the year (see Material and methods and Fig. S2). Arctic settlements (e.g. Pokhodsk and Kharyalakh) are projected to experience moderate, although statistically nonsignificant, local increases in species mean habitat suitability under both scenarios, with larger increases associated with the most extreme SSP5.85 scenario (Fig 5A and Fig. S3). These changes are projected to result in fewer extirpations of currently available species, while several new species are projected to expand their ranges and become established locally, resulting in greater net increases in species richness relative to other regions. Conversely, the current mean habitat suitability of local wild food assemblages in most settlements in the western, central and southern regions is projected to experience substantial decreases (e.g. Chappanda, Rassoloda, and Ugoyan in Fig. 5A), often statistically significant under the SSP5.85 scenario. As a result, greater numbers of extirpations than colonizations are projected for these areas. Despite these large changes in habitat suitability, the resulting projected net changes in species richness are small under both scenarios (1.89 ± 2.4 and 1.28 ± 4.07 for SSP1.26 and SSP5.85 scenarios, respectively) as the extirpation and colonization of species often, but not always, compensate for each other (Fig. 5A and Fig. S4). This situation suggests that, over and above the effect of changes in species richness, species replacements are likely to pose emerging challenges and opportunities for dependent local communities. For example, moose (Alces alces) is projected to expand its range into some of the northernmost Arctic settlements by the 2050s under both scenarios (e.g. Pokhodsk in Fig. 5B), likely generating new food resources for these communities, some of which are currently among the most dependent on wild mammals across the RS (Fig. 2). On the other hand, settlements in central regions may face the extirpation of important food species. For example, projected local strong decreases in habitat suitability and possible extirpations of several species of plants, such as blackcurrant (Ribes nigrum), wild onion (Allium schoenoprasum), or lingonberry (Vaccinium vitis-idaeaea), may have important implications for settlements such as Rassoloda (Fig. 5B) where nonanimal foods currently contribute disproportionately (71%) to the total income share by wild food harvesting (Fig. 2B) and represent a significant proportion of the total fungi (79.86 ± 33.98%) and plant-based (27.77 ± 27.07%) foods consumed by households (Fig. 2C). Although some new species, such as wapiti (Cervus canadensis) or black grouse (Lyrurus tetrix), are projected to experience improvements in habitat conditions and become established locally, it is difficult to anticipate whether they will compensate for the possible impacts on current dependencies generated by the loss of plant species (i.e. wild animal-based dependencies are currently minimal for Rassoloda; Fig. 2).

Discussion

Dependency on wild harvesting

Globally, availability of wild foods is in decline due to a combination of drivers, most notably land use change, habitat loss, and overexploitation (29, 30). Climate change is an emerging compounding threat that is already altering the abundance, distribution, and accessibility of wild food species, particularly in sensitive regions like the Arctic, as well as the traditional practices of food procurement and storage by dependent communities (22, 31). As these changes continue to unfold, it is important to understand and anticipate their effects on human–environment relationships and existing dependencies of rural communities (8). We found that dependencies were greatest in the northern Arctic settlements and lowest in the central and western regions of the Republic (Fig. 2A). This trend was also reflected in the average household income shares for Kharyalakh and Rassoloda, which can be considered representative of the Arctic and non-Arctic regions of the Republic of Sakha, where the share by wild foods in Kharyalakh was nearly 3-fold higher than that of Rassoloda. These results are likely a reflection of regional differences in transport infrastructure and access to manufactured goods and services as well as in the traditions and practices of their indigenous communities. While the central districts are the most populous of the Republic, in close vicinity and direct access to the capital Yakutsk, the western region has the highest gross regional product due to the concentration and rapid development of the extractive industries for diamonds, oil, and gas. Conversely, infrastructure and access to services in the vast Arctic region of the Republic of Sakha remain minimal and, importantly, especially vulnerable to climate change (32). For example, climate warming is causing reductions in the operating windows of winter roads that are critical in the region (33); disruptions that are projected to exacerbate in the future (32). Whereas infrastructure development in rural areas is associated with increased overall well-being (25) and can have a positive effect on incomes and decrease inequalities at the individual household level (34), remoteness, and poor accessibility are factors that tend to reinforce the reliance on, and contribution of, nature to rural household economies (25). In this sense, current household wild food dependencies in settlements of the Arctic region are likely to decrease in the future as a result of the planned expansion of the rapid developing extractive mining, fuel, and energy industries, which will bring both improved access to services and the development of more robust and reliable infrastructure to the region, as well as increased environmental impacts and ultimately changes to the traditional activities of the indigenous rural communities; something that has been reported in the past with the other more developed regions of the Sakha Republic (35).

The contribution of nature to well-being and livelihoods in rural communities is typically strong, particularly in remote, asset-poor areas. For example, an exhaustive review of nearly 8,000 households in 24 developing countries placed the share of total household income accounted for by environmental income at more than a quarter (28%) (16). Of this share, wild foods represented a key category, globally comprising between 30.3% and 48.9% of the environmental income and 10.9% of the total household income; a figure virtually the same as our estimated income share for Kharyalakh (11%), but significantly larger than that of Rassoloda (3.42%). Environmental income primarily subsidizes rural livelihoods via three complementary pathways: supporting household consumption, generating cash income, and reducing poverty gaps by providing seasonal gap-filling and safety-net functions (16, 36). Consumption-related income is typically higher in remote rural areas where poor transport infrastructure, limited access to services, and the lack of alternative food sources (i.e. marketed foods being more expensive, less varied and prone to fluctuations in availability) contribute to stimulate subsistence consumption of wild foods and materials (16, 37). This is reflected in our results. Kharyalakh is an Evenk's community of the remote and isolated Artic region with forest-tundra reindeer husbandry as one of their main traditional activities (25.7% of the surveyed households also declared to hunt wild reindeer). Conversely, Rassoloda is a Sakha community in the center region near the capital Yakutsk with good infrastructure and access to services which main traditional activities are cattle and horse breeding (18.2% of households engaged in cattle breeding, whereas none did in Kharyalakh). Crop farming was also far more common in Rassoloda than in Kharyalakh. Agriculture in the boreal and Arctic regions has traditionally been perceived as marginal and insufficient to cover the needs of local communities (38). As part of our in-depth surveys, we also monitored the price and assortment of food and commodities from local shops and markets at both settlements. In Kharyalakh (June 2022), 33% of the 66 tracked commercial products were unavailable, with food products dominated by long-lasting and highly processed foods while fresh products such as beef, dairy products, and vegetables were unavailable. Conversely, 91% of the tracked products were available from shops in Rassoloda, including a much wider choice of foods. Further, remoteness and poor logistics also raise the price of available products. The affordability of basic types of food for Rassoloda was significantly higher than for Kharyalakh; the price gap for fresh vegetables (2 to 8 times) and fruits (2.7 to 3.5 times) being particularly large. Healthy, fresh food from local shops is often physically and economically inaccessible to residents of remote communities. Nonetheless, estimated dietary-based dependencies across our study settlements confirmed a strong and geographically consistent contribution of wild foods to household consumption. Wild berries and nuts (35.9 ± 8.5%), fungi (77 ± 18.5%), and fish (77.2 ± 11.8%) represented a significant share of their respective food groups consistently across all settlements (Fig. 2C). These results highlight the dietary and cultural importance of wild foods in rural settlements over and above their contribution to household economies.

Changes in the regional distribution and local availability of food species

Biodiversity has a crucial role in ecosystem processes that sustain production and provisioning from nature and enhances the resilience and capacity of the ecosystem to respond to impacts. The access of households to a diverse array of foods, distribution channels, and incomes is crucial for sustainable food security outcomes (16, 25, 39).

Our projections of a northern shift in the current distribution range of many boreal food species and declines in Arctic species agree with observations and projections from other studies. Climate change is driving the progressive greening of the Pan-Arctic tundra associated with the expansion of shrub vegetation, which is linked to a biome-wide increase in productivity (40). This process has facilitated the expansion of associated boreal animals such as moose, snowshoe hare, red fox, or American beaver (41, 42). Paradoxically, this same greening in Arctic North America has been associated with negative impacts on caribou populations despite increased plant biomass due to the deterioration of summer pasture quality, as shrub expansion involves plant species with strong antibrowsing defenses (43). Moreover, climate change contributes to population declines, range contractions and changes in the phenology and movement patterns of many Arctic species (44). In turn, these climate-driven changes in the composition and abundance of local species, fueled by other human-related impacts such as land conversion, overexploitation and invasive species, are contributing to changes in the customs, traditions, and nutrition of IPs. For example, IPs from the Arctic region of the Republic of Sakha recall an increased presence of predatory boreal forest species in the Arctic tundra, such as wolves, brown bears and sables, together with the arrival of new species, including migratory birds such as mallards, shovelers, lesser white-fronted geese, or plants such as dandelions or black currants (31), several of which are included in our study and projected to expand their ranges into the Arctic tundra. Increased predator presence can have a negative effect on the local wild species of mammals and birds they prey upon, some of which are important game species, and can impact people's livelihoods by, for example, predating on reindeer herds or competing for fishing resources and tampering or destroying fishing gear (31). On the other hand, some common local waterfowl breeding species, such as different ducks and geese, are in decline. Although this pattern seems to be mainly connected to impacts outside the Arctic, such as overhunting (42), it has a negative impact on the diet of local indigenous communities because the impacted species are important subsistence game species (31). Similarly, recent warming has pushed reindeer migratory routes northward in the RS (31) while posing increasing threats to traditional reindeer husbandry (45). These changes have significant implications for rural indigenous communities because reindeer is a keystone food and cultural species (19). Nutritionally, reindeer meat is rich in protein, minerals, and essential fatty acids, and its consumption together with that of blood and liver, which are culturally accepted in Sakha Arctic communities, helps prevent diseases such as chronic bronchitis, obesity, and arterial hypertension (20). On the other hand, the presence of new species may also lead to changes in hunting, fishing, or harvesting, either directly or indirectly, by local indigenous communities. For example, authorities have cataloged predators expanding into the Arctic tundra, such as wolves and brown bears, as hunting species and are using rewards to encourage local people to hunt them (31). With time, these changes may translate into future alterations in current consumption and income-related dependencies.

Our study presents novel information on the current contribution of wild food species toward supporting rural households in a region (the RS and the Russian Far East) underrepresented in the literature but very important to the issues at hand. Compared with other Russian northern regions, the RS has experienced a relatively small decline of its large rural population, comprising 35.9% of its total population mainly controlled by big state-run companies (26), due to its high proportion of the indigenous population, the highest among all Russian Far North regions, who traditionally live in rural areas where they engage in pastoralism, reindeer herding and other traditional natural resource uses, and the existing measures implemented by the regional authorities to support rural communities from the revenues obtained from lucrative extractive industries of natural resources such as diamond, gas and oil (26). Wild harvesting still represents a core traditional activity of many of these communities, which diet is characterized by a significant proportion of traditional foods (20). At the same time, their livelihoods are increasingly threatened by rapid environmental changes including the extraction of its rich natural resources and accelerating climate change (27). We contribute to this body of knowledge by putting current dependencies on wild food harvesting across the RS in context with the potential implications of the regional and local redistribution of food species under future climate and land use changes. As mixed household economies become increasingly established in Arctic indigenous communities (24), attaining a balanced preservation of local wild harvests with an increasing equitable access to infrastructure, markets and skills is seen as most beneficial for securing the well-being of these remote, isolated rural areas (25). This is important because transition from traditional subsistence economies into mixed household economies can also result in a reduction of consumption of wild foods and wild harvesting to commodity production. For example, in western Siberia increases in the export of traditional reindeer products (meat as well as nonedible products such as skins or velvet antler), have resulted in a decrease in the access to, and consumption of, reindeer by local Indigenous communities with associated health impacts (21).

Wild foods chiefly contribute to the food security, social equality, physical and mental health, cultural identity and sense of relatedness of a large proportion of the population worldwide (30), particularly for indigenous and rural communities in remote and asset-poor regions (46). We concur with the growing number of voices calling for a solution to the current global policy and legal limbo related to the conservation and sustainable use of wild foods (7, 47) by moving from a traditional science-based only approach to wildlife conservation to integrated conservation approaches that recognize the diverse values of nature for sustainability (6) and incorporate indigenous perspectives, rights, and traditional knowledge into management and conservation practices (7, 47).

Materials and methods

Study area and settlements

Our study was conducted in the Republic of Sakha in the Russian Federation's Far Eastern district (Fig. 1). The Republic is sparsely populated, with 0.32 people/km², 33.2% of whom live in rural areas, yet this region boasts rich cultural, ecological, and geographic diversity. According to the 2010 All-Russian Population Census, representatives of more than 129 nationalities/ethnicities permanently reside in the RS. Yakutia has a significant population share of Indigenous Peoples, including Chukchi, Dolgan, Even, Evenk, and Yukagir who are catalogued as Indigenous small-numbered peoples of the North, Siberia, and the Far East.

Eighteen rural settlements (Fig. 1 and Table S1) were selected to assess household dependencies on wild traditional foods (see below). Settlements were distributed across the five existing economic zones in the Republic of Sakha according to their extent and population: Arctic (5 settlements), central (7), southern (1), eastern (1), and western (4) Yakutia (Fig. 1). The population of these settlements in 2020 ranged between 127 and 2289 censused residents with mixed ethnic representation (Table S1).

Household surveys

Data to assess the dependency on wild harvesting of the average household by settlement (see below) were extracted from a large two-stage systematic household survey conducted for a separate international joint research project (48). The first stage covered 400 households from 18 settlements during June–December 2021 (three settlements were visited in the beginning of 2022 due to accessibility problems). These settlements were chosen based on multiple criteria including their distribution across the five recognized economic zones of Yakutia, settlement size, their traditional economic activities and ethnic composition according to the 2010 All-Russian Population Census, and logistics (accessibility during the field research period). The second stage comprised a follow-up, in-depth survey of two settlements, Kharyalakh (105 households surveyed between April and November 2022) and Rassoloda (80 households between March and September 2022), selected as representative of the Arctic and non-Arctic regions and because of their proximity to certified medical laboratories that were used for analysis of biochemical samples, which were also taken as part of our wider project (48). The first-stage survey collected detailed information on dietary patterns as well as basic background household information, such as demographics, education, and occupation. The second-stage survey further asked for details of income-generating activities, including wage labor, business, social security, farming, and hunting/fishing/gathering of wild species (see Table S6 for a list of the relevant questions from each stage of the larger surveys that were used here for the calculation of the dependencies). Participants comprised the family head or other members who played major roles in procuring and managing food in the household. The surveys were approved by the North-Eastern Federal University (NEFU) Biomedical Ethics Committee (approved Resolution No. 6, Report No. 33 on 2021 December 15). Prior informed consent was obtained from all survey participants. Specific details on the survey design and ethical standards are provided in García Molinos et al. (48).

Traditional wild food species

We compiled a list of traditional wild food species for the Republic of Sakha from several sources including technical reports (49–52) and relevant scientific literature (53–57). This list was used to inform both the wild food species in the household surveys and the selection of species for the distribution models (see below). For the latter, given the availability of expert-range maps needed for the development of our distribution models, we selected a final set of 51 species of mammals (7 species), birds (22), plants (14), and mushrooms (8) (Table S2). Overall, this list is intended as a nonexhaustive but representative sample of the most relevant wild plants and animals contributing to the TFSs of rural indigenous communities across the Republic of Sakha.

Given their importance, freshwater fish species were considered for inclusion, but we ultimately left them out of the analysis for several reasons. First, we were often unable to identify individual species in household surveys, as these species are frequently reported generically as “fish” or with broad common names comprising several distinct species (e.g. whitefish). Second, distribution maps were not available for many of these species or did not well reflect their distribution within the Russian Federation and the Russian Far East in particular. Finally, existing high-resolution databases of hydrological networks and associated basins are either missing from high-latitude regions (particularly the Arctic) or known to have low quality and limitations due to the lack of high-resolution ancillary data (e.g. 58). Nonetheless, although we could not include fish species in our model projections, we do consider this food group in our dependency analysis on wild traditional foods (see next section).

Dependency on wild food harvesting

We quantified the average household dependence on wild food species (hereafter referred to as “dependence on wild food harvesting”) by settlement via two complementary quantitative approaches based on (i) the annual food consumption frequency of wild food species over all consumed foods (i.e. dietary-based dependency), and (ii) the share of income of wild species over all household incomes (i.e. income-based dependency). The dietary-based dependency was calculated for each of the 18 settlements based on the data of the first-stage survey. Respondents were asked to provide the consumption frequencies and origin (obtained from nature or cultivated/bought in the market) of food items commonly consumed in the Sakha Republic including the traditional wild food species precompiled by our team referred to above (see Table S7 for a list of all food species and their corresponding categories). This information was used to calculate dependencies based on consumption (see section Dietary-based dependency). The income-based dependency was then calculated for the two settlements, Kharyalakh and Rassoloda, based on the second-stage in-depth survey. In this survey, respondents were asked to provide the Russian or local names and quantities of all food species obtained from nature or cultivated in their households to accurately characterize the contributions of wild foods toward household incomes (see section Income-based dependency).

Definition of wild and domestic foods

Food items frequently reported to be collected from nature in our household surveys were categorized as wild foods for the purpose of our dependency analysis if they were included in our compiled list of traditional food species for the Republic of Sakha referred above (Table S7). Our classification is therefore based on the origin of the species, where a food species is deemed wild if it is locally available from nature and reportedly collected/hunted/fished by rural residents as food. Given we were asking for the recollection of food species consumed over the year, we did not conduct any collection of specimens to check the reported species names. We therefore assume that the local people who frequently collect and consume these common wild food species are sufficiently familiar with them to be able to name them accurately. We also note that, to account for situations where respondents were unsure about the specific name of the food species in the surveys (or where the species was not among those listed in the first survey), we also provided the option to report them as generic groups (e.g. game birds, fish, mushrooms) rather than individual species (Table S7), which were then used as such in our dependency analysis. However, we note that this classification does not account for the possibility that some households may obtain some of these species from local markets (although most wild foods sold at local markets are themselves hunted/collected/fished locally from nature) or that some wild foods may be obtained from both wild and domestic sources (e.g. reindeer herding as opposed to hunting wild reindeer). On the other hand, it is also difficult to ascertain the extent to which human intervention in practices such as open-range herding, common with reindeer in the Republic of Sakha, can influence the wild status of a species. Although these situations most likely affect, to some degree, the households in the surveyed settlements, we could not make these distinctions based on the data collected in the surveys.

Dietary-based dependency

Let T_j represent the complete set of foods consumed in household j from settlement s and W_j the corresponding set of wild foods hunted, gathered, or fished (HGF) from nature by that household, such that $Wj⊂Tj$⁠. The household dependence on wild food harvesting in terms of dietary consumption $Djdiet$ (i.e. the share of wild foods in the household diet) can then be defined as the ratio:

where $zi,j$ is the annual consumption frequency for food item i by household j calculated from the household surveys as explained below. The average household dependence on wild food harvesting in settlement s in terms of dietary consumption is then calculated from all n households surveyed in the settlement as:

In addition to the overall index, to have a better insight into the dietary contribution of wild foods by specific food groups, we calculated dependencies for individual food groups by calculating the shares of wild mammals, birds, fish, plants (vegetables), berries and nuts, and fungi to the total (wild and nonwild) of the corresponding food group category.

Annual consumption frequencies (⁠$zi,j$ in Equation (1)) for each identified food item (Table S7) were estimated from frequency categories as recorded in our household survey. First, these categories were converted into an equivalent number of consumptions per month (Table S8). Subsequently, they were multiplied by the number of months each food item was consumed by each household (i.e. accounting for seasonality). The results of these calculations indicated the annual consumption frequencies for each food item. Although examining both consumption frequencies and quantities would provide a more precise measure of dietary-based dependency, logistic and time constraints made it unfeasible to collect data on consumption quantities of all individual items in our case. Focusing on consumption frequencies, on the other hand, has the benefit of minimizing potential recalling errors in respondent's answers. Of the 400 households included in the first-stage survey, one household in Kharbala was excluded from the analysis as it did not report any dietary information.

Income-based dependency

We define dependency on wild food harvesting in terms of income sources as the share of incomes from harvested wild food species, which can be defined as the ratio of the annual income obtained by any household j from hunting, gathering, and fishing wild food species $yjW$ to the total annual income $yjT$⁠. The average income-based household dependency of settlement s over the n household surveyed can then be defined as follows:

The total household income includes the income obtained from the harvesting of wild food species as well as wage and business incomes, farming income (e.g. cattle, poultry, and crops), and any other reported sources of income (e.g. social security, subsidies, financial income, and remittances). Wage and business incomes and other sources of income were based directly on data collected from the household survey. Income from farming and wild foods was calculated as revenue minus reported costs (see Extended Methods in the Supplementary Material for details). Following customary practice, we account for both monetary and nonmonetary revenues. The latter refers to the domestic consumption of farming and wild foods by each household as well as donations to other households. Thus, we shed light on the values of farming and HGF productions rather than the actual cash revenue.

Besides the dietary-based dependencies, we also calculated the breakdown of the income-based dependency by food groups (i.e. the contribution of each wild food group to the household income). To do so, we aggregated the total revenue over all households in a settlement and then calculated the share by food group. These calculations differ from that of $D¯sinc$ in the following two points. First, they are based on revenue instead of income because the survey's information on costs did not allow us to separate them into food groups. Second, we calculated the group contributions to the total harvested wild food revenue share using settlement-level aggregated values, instead of calculating average household values, to better reflect the real distribution of income contributions by food group to revenue within a settlement. Whereas a small number of households engaged in large-scale hunting and earned most of the revenue from hunting, most households typically did only small-scale gathering (mostly berries and fungi) and earned marginal revenue from wild food harvesting. In this situation, attaching the same weights to all households by calculating the partition of the share by household and then taking the average (as in Equation (3)), would fail to illustrate the real importance of food groups in terms of revenue. For example, in Kharyalakh the share of plants over the aggregated revenue is only 2.4%, but this share becomes 27% if calculated as the average share by household. Incomes per species group (monetary and nonmonetary) were estimated based on survey data, expert consultation, and local (producer) and official (retailer) price statistics, as detailed in the Extended Methods section of the Supplementary Information and Table S9.

Of the 185 households included in the second-stage survey, 14 households were not considered for analysis of the income-based dependencies due to either refusal to disclose income information (12 households) or reporting inconsistently large wild food harvesting and farming costs compared to revenues (2 households). The resulting rate of refusal (6.5%) can be considered acceptable for a household survey.

Species distribution models

Range maps and occurrence records

Expert-revised range maps for each species, which were used for fitting the distribution models (see below), were obtained from several sources (Table S2). Despite our exhaustive search, we were unable to find distribution maps for our eight fungi species. Therefore, we used the minimum convex hull generated from the full set of occurrence records as a substitute. Compared to the range maps, the convex hulls represent a more conservative (inclusive) assessment of the species range, but we consider it to be acceptable given that the purpose of using the range maps is to generate species pseudo-occurrence records for fitting the models, as explained below.

The species occurrence data used for validating model performance (see below) were obtained from multiple biodiversity repositories (Table S2; a full account of individual retrieved datasets is also provided in Table S10). We restricted the search geographically to Eurasia and temporally between 1970 and 2020. This time window spans 10 years before/after the baseline period of the climatic predictors used to build our models (1980–2010). The retrieved records were then filtered using a 500-km buffer around the distribution map of each species, which was used as our model calibration extent. Filtered occurrences were subsequently quality checked using the R package “CoordinateCleaner” (59) to remove records that were duplicated, had invalid or zero coordinates, or were in close (<10 km) vicinity to country capitals, country centroids or biodiversity institutions. To reduce spatial clustering of the records, we further performed spatial thinning of occurrences by using a 10 km thinning distance and the function “thin” of the R package “spThin” (60). The retained occurrences were then used as an independent dataset to validate the range map-based models built for the prediction of the distribution of the species (see below).

Environmental data and selection of predictor variables

We used a combination of bioclimatic, topographic, and land use predictors (Table S11). We considered two periods representing baseline contemporary (1981–2010) and future mid-century (2041–2060) bioclimatic and land use conditions and projections for the latest Coupled Model Intercomparison Project Phase 6 (CMIP6) coupled scenarios (the SSP-RCP scenarios). We focused on two of the highest-priority (Tier 1) scenarios recommended by the CMIP6: the low GHG emission, sustainability-focused SSP1–2.6 and the very high emission, fossil-fuel based development SSP5-8.5. See the Supplementary Material Extended Methods for further details on the environmental predictors and data sources, climate models and scenarios.

We used statistical and ecological criteria to select taxon-specific subsets of uncorrelated (|r| ≤ 0.7) environmental predictors (Fig. S6 and Table S11) that represent the range of regional climatic conditions well and are likely to influence species distributions (see Supplementary Material Extended Methods). Our bird species included several migratory species (Table S2) that breed from late spring in our study region until they return to their wintering grounds in Southeast Asia during the fall. Therefore, after correlations were accounted for, these species were modeled using only bioclimatic variables representative of conditions during the breeding season (summer) (Table S11).

Distribution modeling

SDMs based on occurrence localities (presence only or presence/absence data) assume that sampling is conducted uniformly across the entire occupied range of the species and properly represents the environmental conditions within. When this is not the case, there is a risk for incomplete sampling and oversampling bias, which can affect the estimation of model parameters and lead to poor predictive model performance (61). These issues are of particular concern in our case, as available occurrence records for most species were strongly clustered within northern Europe, while few to no records were available for the Russian Far East and the Republic of Sakha in particular (see Fig. S5 as an example). Therefore, we used an alternative modeling approach in which expert-derived range maps were employed to generate pseudo-occurrence records within a species range to fit the models (62, 63). Although map-based SDMs have been shown to perform similarly to occurrence-based SDMs at resolutions similar to those in our study (63), the probability that locations sampled randomly across a species range map represent true occupancy conditions decreases with finer resolutions (i.e. range maps tend to overestimate real occupancy) (64). To minimize this issue, we restricted the sampling of pseudo-occurrences for each species to their associated area of suitable habitat within its range (sensu 65). Our modeling procedure included the following steps (Fig. S5): (1) generation of pseudo-occurrences and background points from areas of suitable habitat within the distribution range, (2) selection of predictors, (3) model tuning and calibration, (4) model evaluation, and (5) model predictions for contemporary and future periods. All analyses were conducted in R v. 4.3.1. Below, we briefly describe each of these steps and provide further details in the Supplementary Material Extended Methods.

Generation of pseudo-occurrences and background points

Generation of pseudo-occurrences for each species was conducted by randomly sampling 10,000 locations (cells) containing suitable habitat within each species’ range map with a minimum separation of 10 km between selected cells. This was done by comparing information on suitable level 2 IUCN habitat categories for each species (Tables S2 and S12) with available distribution maps of those habitats (66) (see the Supplementary Material Extended Methods for details).

Following recent recommendations (67), we used a random selection of 50,000 background locations within the calibration extent (i.e. a 500-km buffer around the distribution range) after excluding pseudo-occurrence locations, with a minimum separation of 10 km between locations.

Model tuning, calibration, and evaluation

We modeled the current and future relative probability of presence for each species via the maximum entropy (MaxEnt) algorithm (68). MaxEnt is widely used and has been shown to consistently rank among the best performing presence-only methods available (67). We tuned the MaxEnt parameters using a 4-fold spatial block cross-validation approach on the training dataset (69) and selected the parameter combination (regularization value × feature class) yielding the best performing model as determined by the area under the receiver operating characteristic curve (AUC). These results are provided for each species in Table S13 (see the Supplementary Material Extended Methods for details).

Model evaluation within calibration extent was conducted using the available occurrence records collected for each species from biodiversity repositories (Tables S2 and S10), which represent independent out-of-bag datasets of true occurrence locations for evaluating our distribution map-based models. We used the continuous Boyce index as a presence-only evaluation metric to assess model performance (Table S13). The Boyce index is a threshold-independent metric that measures how model predictions differ from a random distribution of the observed presences across the prediction gradient (70). The index ranges between −1 and 1, with values close to 0 indicating that predictions are not different from random outcomes and increasing positive/negative values indicating that presences are increasingly more frequently predicted in areas with higher/lower predicted suitability values than expected by chance alone. Given that small sample sizes may render comparisons between predictions and random expectations meaningless, we alternatively calculated Boyce indices using the pseudo-occurrences generated from the range maps via 4-fold spatial block cross-validation for those species (n = 5) with less than 50 occurrence records (Table S13).

Prediction of current and future distributions

We analyzed changes in species richness under different future scenarios by transforming the continuous habitat suitability predictions into binary (presence/absence) range maps for each species. This was done by using the 10^th percentile training presence as a threshold. This is a common and robust classification threshold that omits regions of habitat suitability lower than the suitability values associated with the lowest 10% of occurrence records (71). Using the resulting range maps, we calculated the proportion of future range loss/gains relative to current distributions for each species and the shift distance (km/decade) and direction (bearing) of projected changes in the distribution range centroid.

To account for the role of dispersal as distribution ranges expand under future climate change, we limited projected future distribution ranges by the distance that each species could theoretically cover over the projection period. We considered their current distribution limits and reported observed rates of range shifts in response to climate change, which were calculated as the 90^th percentile of all latitudinal range expansions reported in the Bioshifts database (72) for the corresponding family or the nearest closest higher taxonomic level available (Table S5). Although shift rates at higher taxonomic resolutions (e.g. species or genus) were available for some species, we decided to use the family as our reference level for consistency and to increase the number of available observations given that shift rates can be highly context-specific to local conditions. Nevertheless, the estimated expansion rates are intended as coarse approximations of the rates at which our species may expand their ranges in response to future climate change given dispersal limitations and are used here to provide more realistic and conservative projections than those provided by unrestricted expansions. Results assuming unrestricted dispersal showed similar overall patterns but with considerably larger expansions of species ranges and greater accrual of species into northern areas (Fig. S7).

As a threshold independence analysis, we also monitored changes in predicted habitat suitability for each species and taxonomic group at the local (settlement) scale. To do so, we used a circular buffer centered in each settlement of 100 km radius according to the reported maximum distances typically covered by rural residents of our study communities when engaging in gathering, fishing, and hunting wild food species (Fig. S2).

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This work is performed under the East Asia Science and Innovation Area Joint Research Program (e-ASIA JRP) for the Climate Change Impact on Natural and Human Systems call supported by the Japan Science and Technology Agency (JST SICORP Grant Number JPMJSC20E5) and the Russian Foundation for Basic Research (RFBR project number 21-55-70104).

Author Contributions

Conceptualization, J.G.M.; Methodology, J.G.M., D.Y., D.N., and T.G.; Formal analysis, J.G.M., D.Y., S.K., and V.P.; Investigation, T.G., V.P., V.G., and A.M.; Writing—original draft, J.G.M. with input from D.Y.; Writing—review & editing, all authors; Funding acquisition, J.G.M. and T.G.; Supervision, J.G.M., T.G., D.Y., and D.N.

Data Availability

Data used to generate the reported results is included as Supplementary material to the manuscript including complete citations to all published datasets used in our study (i.e. environmental predictors and species occurrence records). The raw household survey data (in Russian) cannot be included because it contains sensitive information on the participating households. Relevant summary tables are provided as supplementary tables and reasonable requests for these data may be further addressed to Prof. Tuyara Gavrilyeva ([email protected]), co-author and data holder.

References

Author notes