The socio-ecological dynamics of food insecurity among subsistence-oriented indigenous communities in Amazonia: A qualitative examination of coping strategies among riverine communities

: Despite the Amazon’s natural wealth, food insecurity is a major concern among indigenous 4 communities. Yet, little is known about the socio-ecological dynamics shaping the contributions of local 5 ecosystems to food security. In this study we examine how ecological features interact with normative 6 structures, lifestyles, and livelihoods to expose indigenous peoples to food shortages and how they 7 attempt to cope with worsening food insecurity conditions through participatory exercises with ten 8 indigenous communities along the Caquetá River, Colombia. Our results indicate that traditional food 9 systems are sensitive to human and natural capital disruptions. However, severe food insecurity is 10 prevented by the combination of a well-preserved environment and traditional social institutions, which 11 facilitates widespread access to wild foods, farmland, environmental knowledge, supportive relations, 12 and labour. Nevertheless, traditional adaptations appear insufficient when food insecurity results from 13 health shocks. Our findings highlight the need for interventions that pursue conservation objectives 14 whilst promoting social structures supporting resilience.


INTRODUCTION 35
Food insecurity, defined as the lack of "secure access to sufficient amounts of safe and nutritious food 36 for normal growth and development and an active and healthy life" (FAO 2017: 107), remains a global 37 problem. Currently, 815 million people endure chronic undernourishment worldwide (FAO 2017). In a 38 context of growing concern about climate change, the United Nations' Sustainable Development 39 Goals has commit signatory countries to end hunger whilst preserving ecosystems (UN 2015). There is 40 consensus that further research is needed to better understand the role natural ecosystems may play 41 in developing sustainable and resilient food systems that benefit vulnerable populations, such as the There is substantive evidence that wild foods and other non-timber forest products (NTPFs) support 45 the regular functioning of food and livelihood systems among the rural poor, generating 'daily net' 46 contributions in various forms that enhance dietary diversity and nutritional intake (Powell et al. 2015; settings. Whilst there have been recent efforts to promote related policy debates in response to 111 environmental change due to urbanisation, deforestation, and pollution (Ortiz et al. 2013;UNEP 2009), 112 such discussions commonly overlook the plight of communities yet to be directly affected by these 113 trends (Kuhnlein et al. 2013). Food insecurity is in fact a constitutive part of indigenous populations' 114 reality. For example, Brazil's and Colombia's health and demographic surveys from 2009 and 2011 115 found, in their respective Amazonas departments, that stunting 1 rates among children under five were 116 over twice their country-level estimates (Brazil: 40.8% against 7.1%; Colombia: 28.7% against 13.2%) 117 (Horta et al. 2013;Ojeda et al. 2011). In addition, hydrological data indicate that severe flooding and 118 droughts periodically occur in Amazonia, exposing residents to covariate shocks (Marengo et al. 2013). 119 Climate models suggest that seasonal weather patterns are changing in the region, reporting a 120 gradual intensification and lengthening of the dry-season (Joetzjer et al. 2013); while conservation 121 models highlight that biodiversity-rich areas may face extinction threats originating elsewhere, as 122 intensive deforestation can produce knock-on effects on sensitive mammalian species (Soares-Filho et 123

al. 2006). 124
We examine food insecurity through indigenous groups' own descriptions of disturbances affecting 125 access to food and the response mechanisms they adopt. We use the concept of 'coping strategies,' 126 defined as the range of responses that people adopt temporarily in the face of stressors or shocks that 127 take place within existing social structures, such as production, knowledge, and governance systems 128 (Ellis 2003;Eriksen et al. 2005). These structures include those institutionalised forms of adaptation to 129 the environment that help managing risk, known as 'adaptive strategies' (Berman et al. 2012;Eriksen 130 et al. 2005). Accounts of coping behaviours are considered suitable indicators of food insecurity. 131 Comparisons across countries found that coping strategies reflect worsening food insecurity 132 conditions, as people attempt to balance their food needs with income generation and assets 133 preservation (Davies 1993;Harvey et al. 2014;. When food insecurity is mild, for 134 instance, families typically draw on existing resources (e.g., wild foods). However, as scarcity worsens, 135 unsustainable means of accessing food become increasingly prevalent, like selling durable assets. 136 Survival strategies such as begging or skipping meals signal failure to cope. The range and times in 137 which such responses are adopted, moreover, are considered to reflect sensitivity and resilience to 138 food insecurity (Davies 1993;Ellis 2003). A rapid transition towards harsh measures in the face of 139 short-term shocks, for example, indicates a highly sensitive livelihoods system, as in the case of 140 landless casual labourers, who may rapidly experience hunger when unemployed for a few days. In 141 turn, the absence of destitution-type coping strategies despite experiencing shocks indicates that a 142 food system can recover fast enough to prevent the collapse of local livelihoods. 143 We frame our examination in the sustainable livelihoods framework (Scoones 1998) and privilege a 144 qualitative lens to account for the socio-cultural structures guiding residents' food provision activities, 145 which are then defined not only in economic terms but also in relation to lifestyles and forms of identity 146 (Rigg 2007). This approach helps to broaden our understanding of livelihoods from a set of measurable 147 stocks of money, labour, and goods to one that incorporates immaterial forms of capital, such as social 148 relations and forms of association (social capital) (Bebbington et al. 2006

STUDY AREA
The study was conducted in the corregimiento La Pedrera, department of Amazonas, Colombia. 2 It 157 comprises four indigenous reserves (resguardos), containing ten villages, two informal rural settlements 158 (veredas), La Pedrera town, and two State forest reserves (Fig. 1)  Fishing and hunting are also mostly for household consumption, although occasional sales are common. 195 Catfish species, like dorado (Brachyplatystoma flavicans) and pintadillo (Pseudoplatystoma tigrinum), and white-lipped peccaries (Tayassu pecari) and tapirs (Tapirus terrestris) are sold in the town market. 197 Bushmeat trading is illegal but widely accepted. Apart from some households that rear poultry, livestock 198 is non-existent. The extraction of timber for shelter is common in the region but its commercialisation 199 is limited. Foraging of wild fruits is mostly for subsistence and ritual use. Paid farm labour is non-existent. 200 When required, households rely on collaborative agreements (mingas) in which friends and neighbours 201 provide labour in return for food, drinks, and mambe (roasted coca leaves). The few forms of 202 remunerated non-farm labour reported include (i) non-professional services (e.g., cooks), (ii) 203 administrative work for public or civil organisations (e.g., schools or NGOs), and (iii) being a member of 204 a village authority. Apart from the last, employment is usually short-term.

METHODS 219
Data were obtained using participatory methods between March and June 2013. Three thematic group 220 exercises were conducted on (i) local livelihoods, (ii) diets, and (iii) coping strategies for food insecurity (Schreckenberg et al. 2016). During group discussions on livelihoods, participants described 222 predominant living conditions, economic activities, domestic roles, and seasonality issues. They also 223 produced a socioeconomic classification of residents using local criteria. During discussions of diets, 224 participants jointly produced a list of food items consumed locally, identifying their sources. These the strategies households adopt to deal with them. Food scarcity events were grouped into 'severity 230 scenarios' representing the extent to which households' diets were affected. These served to identify 231 which bundles of coping strategies were most commonly adopted according to different levels of 232 worsening conditions. Finally, participants described the resources mobilised for each coping behaviour 233 and the (in)formal rules affecting their implementation. 234 Ten villages participated in the study. 3 Five were interviewed separately and the remainder were 235 clustered into two groups due to their shared access to forest areas and similar livelihoods. Twenty-236 seven discussion groups were conducted (Table 1). A purposive sampling strategy was used for all 237 exercises. Discussion groups on livelihoods aimed to obtain depictions from different social groups, so 238 male and female residents of different ages and from different community areas were invited to 239 participate. Discussions on diets included only women who were responsible for preparing family meals. 240 Based on information from the livelihood discussions, coping strategy exercises were carried out 241 separately with 'better-off' and 'worse-off' residents as a proxy indicator for vulnerability to food 242 insecurity (Horta et al. 2013). Participants classified households according to landholding areas, health, 243 and age of household members, and access to cash income. In one village participants considered 244 everybody to be doing equally well, so only one coping strategy discussion group was conducted. 245 Two trained facilitators led each discussion group in Spanish. When needed, community-based 246 fieldwork coordinators helped with translation to local dialects. Discussions were recorded and free-247 hand notes were taken. Participation was voluntary and written consent was obtained from each 248 participant after an explanation of the project's objectives, participants' rights, and data uses. Thematic framework analysis was used to analyse combined free-hand notes and transcribed excerpts. 252 Two general themes were established: (i) food scarcity scenarios and (ii) coping strategies. The first 253 contained three categories according to the severity scenario established ( Table 2). The second theme 254 was divided into 14 categories, one for each coping strategy identified. These, in turn, contained sub-255 categories related to the livelihoods framework: capital endowments (financial, physical, social, cultural, 256 and natural) and institutional factors. 257 Finally, coping strategies were grouped according to their association with different severity scenarios 258 to ascertain patterns in capital mobilisation and institutional set-ups. A 'consensus ranking' approach 259 was adopted . Strategies were allocated to the severity scenario into which 260 they were placed by the greatest number of discussion groups. If a strategy was equally allocated to 261 two scenarios, the number of mentions in adjacent categories was used to determine the final 262 classification (Table 3). Fishing, hunting, and foraging (collectively known as rebusque) are typically conducted on a daily basis. 284 Aside from casabe and fariña, food is rarely stored due to a combination of factors: lack of electricity, 285 warm and humid climatological conditions, and residents' preference for fresh produce. Meat-smoking 286 techniques are mostly used as a culinary choice rather than a conservation measure. Easy-to-store non-287 indigenous products, such as rice and pasta (labelled 'white-people's food'), are occasionally added to 288 meals. 289

290
Participants from 'better-off' and 'worse-off' discussion groups reported that there had not been 291 instances of starvation in the previous two decades and that residents, in general, 'always had something 292 to eat.' However, they identified various factors that affected residents' capacity to obtain food in the 293 quantity, variety, or quality desired (Table 2). First, they stated that hunting and fishing constituted 294 unpredictable activities: while they generally managed to obtain food, they could not be sure about the 295 amount or type of game or fish they would obtain on a daily basis. This uncertainty was perceived as 296 more acute during adverse environmental conditions. Specifically, at the beginning of the rainy season 297 many fishing and hunting grounds became flooded and small fish species dispersed in the floodplains, 298 and game moved deeper into the forest, becoming more difficult to hunt. Friaje events (short-term 299 decreases in temperature, strong winds, and heavy rains) commonly occur between May and June, 300 further limiting hunters' mobility and game availability. During the rainy-to-dry transition, households 301 that fished in streams and ponds were negatively affected by the decreasing water levels. Extreme dry 302 or wet seasons exacerbated these challenges. 303 A second factor affecting food availability concerned short-term household labour unavailability due to 304 travel or minor illness, which interrupted families' daily harvesting and rebusque activities. Finally, health 305 emergencies, particularly those requiring constant care or hospital treatment, meant households could 306 face mid-or long-term labour unavailability, affecting productive activities as well as generating 307 financial pressures due to the need to purchase food to substitute natural resources and to cover travel 308 and subsistence expenses if seeking treatment in cities. Deaths of adult family members affected 309 households' access to food on a more permanent basis. 310 Building upon these descriptions, informants outlined different scenarios of growing inability to cover 312 food needs. All except one group proposed a 3-tier severity gradient, which was thus adopted: 313 1) Low-severity: short-term disruptions in the provision of animal protein, when residents could not 314 obtain fish or bushmeat for a few days either because of unsuccessful hunting and fishing, 315 weather conditions, or temporary household labour unavailability. 316 2) Medium-severity: a more prolonged period of limited access to animal protein; when diets 317 became more reliant on farm produce. This responded to severe flooding or declining water levels 318 or times when labour unavailability extended for a week or longer. 319 3) High-severity: Periods when households could no longer cover their food needs through 320 traditional rebusque activities, commonly due to health crises. This resulted in poor diets, given 321 residents' limited capacity to purchase food in markets due to scarce cash income and medical 322 expenses, particularly if they travelled to cities to receive treatment. 323

324
Participants detailed their responses to the food insecurity scenarios outlined (Table 3). 'Better-off' and 325 'worse-off' groups largely reported similar coping strategies. The most common coping behaviours 326 comprised travelling farther to less frequently used hunting and fishing grounds, purchasing food on 327 credit, asking for donations from relatives or friends, as well as rationing the frequency and amount of 328 food consumed. Destitution-type strategies, such as selling assets, sending children away, or theft, were 329 mentioned only once each and so were excluded from the analysis. The sole dissimilarity observed 330 across socioeconomic groups consisted of an apparently greater reliance among 'worse-off' residents 331 on exchanging tools or farm produce for food. 332 Low-severity food insecurity was associated with responses focused on managing locally available 333 resources. These consisted of investing more time and energy in rebusque activities, with affected 334 residents travelling farther than usual to reach fishing or hunting grounds; increasing consumption of wild foods that are less-preferred due to their taste, like mojojoy (Melolonthidae beetle grubs) and arriera 336 ants (leafcutter ants, Attine tribe); or relying more heavily than usual on forms of collective consumption, 337 such as sharing meals with or borrowing food from neighbours for a few days until rebusque activities 338 could be resumed. 339 Under a medium-severity scenario, strategies shifted towards adopting rationing measures, mobilising 340 additional household resources, and accessing external food sources. Rationing involved reducing the 341 size of food portions and the variety of diets, with meals increasingly comprised of chagra produce 342 (cassava, chilies, and fruits). The mobilisation of additional resources involved women and children 343 engaging in rebusque activities, chiefly minor foraging and fishing activities. Accessing external food 344 sources entailed trading cassava-based products for bushmeat or fish with villagers from other 345 locations; conducting a form of tool 'rental' whereby households provided hunting or fishing gear to 346 neighbours who then shared their catch or game; or trading farm produce with colonists, traders, or 347 town residents in exchange for easy-to-store processed items, like pasta or canned tuna. Residents also 348 reported buying these items from traders on credit. 349 Responses to high-severity food insecurity were aimed at accessing cash income while adopting 350 stringent forms of rationing. Rationing entailed children receiving larger meals (cassava and fruits) than 351 adults. Informants reported, however, that adults rarely had to skip meals. Improving access to cash was 352 mostly achieved via unconditional financial support from relatives or selling bushmeat in town markets. 353 To obtain bushmeat, residents resorted to hunting mainly in community conservation areas where 354 wildlife stocks are likely to be high.

Coping strategies and forms of capital mobilisation 335
In low-severity food insecurity scenarios, local responses relied chiefly on human, cultural, and social 336 capital as inputs to access missing protein sources (Table 4). Human capital involved healthy household 337 members walking for long distances to access distant fishing or hunting grounds whilst enduring mud, 338 heat, humidity, and insects. Apart from shotguns, required physical capital was basic artisanal foraging 339 and fishing equipment (e.g., harpoons or straw baskets). Alongside these material resources, informants 340 Given the lack of labour availability, households resorted to mobilising women and children for 352 rebusque activities. However, this mostly involved minor foraging and fishing since women and 353 children were generally considered to lack the knowledge, skills, and physical strength of experienced 354 male hunters and fishers. The repertoire of rebusque techniques and equipment was hence more 355 limited. At this level of need, chagra produce became more central to diets, mainly cassava-based 356 products and cultivated fruits that did not require labour-intensive harvesting. 357 Community members also engaged in different forms of trade. Households provided hunting or 358 fishing gear to villagers who, in turn, shared the bushmeat or fish they caught. The loan periods varied 359 according to the amount and type of food expected and relationship between parties. Another form 360 of exchange involved exchanging cassava-based products with colonists, traders, and townspeople for 361 easy-to-store 'white people's food' (e.g., canned tuna). Financial capital, moreover, appeared for the 362 first time as a consideration for obtaining food, in the form of credit in local stores. Although better 363 and worse-off informants alike reported being able to obtain credit, the amount of food they could 364 obtain depended on their monetary income. Only those with regular wages (e.g., village authorities) 365 were in a position to cover their food needs through this strategy. For all these transactions, social At the highest level of need, the range of equipment, knowledge, and skills required for hunting were 371 more limited than in previous scenarios because hunting was now specifically aimed at game with 372 market value, such as agouties and peccaries. If possible, hunting was conducted in wildlife-rich 373 conservation areas so that human capital investments (time and energy) were not too demanding. Social 374 capital was key for dealing with severe shocks as well, since having a good relationship with traders was 375 valuable for bushmeat commercialisation. In addition, having relatives in urban areas, with easier access 376 to cash income, was considered advantageous to provide financial support to families facing health 377

emergencies. 378
Coping strategies at the institutional interface 379 Villagers' responses to varying levels of food insecurity required them to engage with different 380 normative structures. When facing short-term food shortages, coping strategies were guided by 381 community-based rules. These included residents' rights to exploit the natural resources within the community's territory and their obligations to respect the conservation areas delimited by local 383 management plans. In addition, traditions of reciprocity and gift giving enabled food sharing and 384 borrowing practices; while the extraction of natural resources largely followed traditional gender-based 385 roles, with major fishing and hunting considered male activities. 386 Under a medium-severity scenario, residents dealt with a mix of local and external rules. At the 387 community level, a strong reliance on farm produce for home consumption underscored the importance 388 of traditional entitlements to chagras. Informants' observation that they 'always had something to eat' 389 centred on the expectation that all residents, independently of their wealth, would have at least one 390 chagra and that the cultivation of perennial crops would ensure a basic but constant access to food that 391 could be complemented through foraging and food-sharing practices. Complementary response 392 mechanisms, however, relied on equivalent exchanges, rental schemes, and purchases on credit, which 393 entailed dealing with market-economy considerations. Informants then experienced difficulties in 394 negotiating satisfactory terms of exchange. In a market context, fellow-indigenous residents became 395 competitors, each of them offering similar produce (fish, fruits, and cassava-based products). Moreover, 396 seasonality effects on traditional production functions hindered a favourable integration of traditional 397 productive activities into market transactions since periods of 'abundance' implied low market prices. 398 Finally, market prices were also affected by a range of external factors, such as fluctuations in fuel prices 399 and inflation that were difficult for residents with limited access to mass media to track. Furthermore, 400 some difficulties were specific to women, who reported being offered lower prices than men and facing 401 male traders' reluctance to engage in price negotiations with them. 402 This combination of local and external norms was also observed in high-severity scenarios. When food 403 insecurity followed health emergencies, residents relied upon their access to chagras for basic 404 sustenance and their rights to exploit forest areas to obtain bushmeat for sale. If hunting was to take 405 place in

LIMITATIONS 508
There are constraints associated with the participatory methodology we used in this study. First, while 509 the research team tried to include testimonies from multiple sources, participants did not constitute a 510 statistically representative sample. As previously noted, the study accounted for gender and 511 socioeconomic differences and reported only those findings consistently found across discussion 512 groups. However, it remains uncertain if hidden forms of bias may have affected participation or data-513 elicitation (e.g., friendship networks) (Mosse 1994). This limits the generalisability of our results to the 514 entire region. 515 Our methods, in addition, did not utilise standard food insecurity measures, such as stunting or caloric 516 intake (FAO 2017). Participatory approaches instead centre on local forms of knowledge and 517 representations (Schreckenberg et al. 2016). Thus, we examined food security as perceived and 518 understood by participants themselves. The highly-contextualised nature of depictions, however, 519 affected the possibility of integrating our results into technical nutritional assessments. Nevertheless, 520 our qualitative results identified a wide range of human-nature relationships that can inform future 521 quantitative food insecurity assessments. Moreover, they also provide contextual data that help to 522 explain indigenous populations' policy preferences regarding conservation and nutrition (Kuhnlein et 523 al. 2013). 524 Finally, participatory appraisals have been conceived to render detailed accounts of living conditions in 525 a short timeframe (Mosse 1994). An associated drawback is that they are unable to generate the same 526 level of in-depth understanding of knowledges and perceptions as ethnographic approaches. Reported 527 narratives of scarcity events and coping strategies hence provide an initial portrayal of a complex and 528 dynamic process. Longer-term research efforts could provide additional insights. Similarly, the focus on As discussed, such assessments cannot be confined to environmental, nutritional, or economic 548 measures. Careful attention should be given to the social structures that allow for the incorporation of 549 natural resources into local diets and promote resilience of the local ecosystem. To be viable, any 550 ensuing policy decisions must incorporate Amazon indigenous populations' own understandings of 551 food insecurity conditions. 552

ACKNOWLEDGMENTS 553
We thank Erwin Palacios (Conservation International) for coordinating fieldwork and the Association of 554 Indigenous Authorities of La Pedrera and villagers for granting us access to their communities. We are 555 grateful to Daniel Giraldo, Sandra Cardona, Catalina Angel, Sara Ramirez-Gomez, and Lina Gallego for 556 collecting field-data, and Charlie Shackleton for his inputs to early drafts of the text. 557

DATA AVAILABILITY 558
The data analysed for the study are available from the corresponding author on request. 559

COMPLIANCE WITH ETHICAL STANDARDS 560
Funding 561 The study was conducted under the 'Attaining Sustainable Services from Ecosystems using Trade-off 562 Scenarios' project (NE-J002267-1), funded by the Ecosystem Services for Poverty Alleviation program 563 (ESPA). 564