Variation in Food and Nutritional Stability Among Amazonian Populations Living in a Context of Dramatic Seasonal Flooding

Central Amazonian rivers annually inundate farmland and expand aquatic habitats, making fish harder to catch. Little attention has been paid to date as to whether this hydrological regime impacts human dietary intakes, and, if so, whether impacts vary between communities in seasonally flooded (várzea) and non-flooded (terra firme) zones. To address this gap, we collected dietary data on over 8,000 meals in six várzea and three terra firme communities during the hydrological cycle, and calculated energy and macronutrient intake. Diets were dominated by fish (44%), the main source of protein and fat, and manioc flour (26%), the main source of energy and carbohydrates. Fish consumption fell as water levels rose. This resulted in reduced energy and macronutrient intake in the várzea, but not the terra firme, where wild meat and fruits helped maintain protein and fat intake, respectively. Our data demonstrate seasonal instability of dietary intake, with resilience provided by access to permanently unflooded land.


Introduction
In subsistence-based farming communities, food availability and access are highly dependent on the local landscape and temporal changes in environmental conditions.Food instability occurs when these changes temporarily prevent people from accessing adequate food, which is most evident when periodic natural shocks such as droughts result in crop failures and famine (e.g., Glantz, 2019).However, far more common is seasonal variation in food availability and access related to more predictable annual changes in temperature and precipitation.Most subsistence-based farming communities experience seasonal variability in both production and consumption (Arsenault et al., 2014;Ferro-Luzzi et al., 2001;Leonard, 1991;Wandel, 1989).In such contexts, wild foods (e.g., fish, wildmeat, fruits) can serve as a nutritional buffer (Coomes et al., 2010;de Merode et al., 2004).However, availability of and access to wild foods are also subject to seasonal cycles (Endo et al., 2016;Golden et al., 2019;Nyahongo et al., 2009), which may become problematic when they represent a dietary staple.
One of the places most dramatically affected by seasonality is Central Amazonia.This incredibly biodiverse and food rich environment for humans experiences a marked seasonal hydrological cycle (Fleischmann et al., 2022).Characterized by a smooth and predictable flood curve with one pronounced peak per year, it is one of the largest seasonal fluctuations in water level on earth.This 'flood pulse' raises river levels by up to 12 m on the Amazon-Solimões main stem, and by > 15 m in tributaries such as the River Purus (Fassoni-Andrade et al., 2021).It seasonally inundates vast areas of floodplain (~ 600,000 km 2 ) transforming much of the riverine habitat (Fleischmann et al., 2022).The mass of water is so large that it is thought to be responsible for the greatest regular sinking and rising in the Earth's crust (crustal oscillation), at 5-7.5 cm annually (Bevis et al., 2005).
Throughout the Amazon Basin, rural, riverine Amazonian people, known as ribeirinhos, rely on the local landscape to obtain much of their food (Murrieta et al., 1999;Piperata et al., 2011).Dietary staples tend to be farinha, a toasted flour made with cultivated manioc, and wild caught fish (Dufour et al., 2016).These staples are complemented by other crops (e.g., pumpkin, yam) planted on the fertile floodplains (várzea) and in upland forests (terra firme), wild and managed fruits, domestic livestock, terrestrial and aquatic wildmeat, and purchased food (e.g., chicken, processed meat, coffee, sugar, rice, beans, and crackers).The goal of this research was to understand the impacts of the dramatic, seasonal hydrological cycle on the dietary patterns of ribeirinho populations in Central Amazonia.
While research conducted in Central Amazonia has documented the ways the flood pulse transforms the riverine ecology and lives of ribeirinho people (Junk, 1997), the effects of seasonal flooding on food availability likely vary based on where households are located on the landscape.For this reason, crops grown in the várzea (e.g., manioc and watermelon) are largely limited to those that can be harvested before the annual water rise (Funatsu et al., 2019).Raising livestock (e.g., chickens, pigs) on the várzea is also challenging due to lack of dry land and difficulties in producing feed (Chibnik, 1994).
In terms of homes, many ribeirinhos construct floating houses atop large rafts made with tree trunks (flutuantes), which rise and fall with river levels (de Oliveira et al., 2021).Those living in stilted houses (palafitas) in the várzea are forced to either temporarily relocate, or continually raise the floorboards to avoid inundation during extreme floods, which are becoming more common with climate change (Barichivich et al., 2018).As river waters invade forested zones, fish and other aquatic animals spread out, making them harder to catch (Endo et al., 2016;Tregidgo et al., 2020).The flood pulse prevents many terrestrial species from occupying the várzea, while other species either move to the forest canopy (Ramalho et al., 2021) or migrate to terra firme zones (Alvarenga et al., 2018) where there is dry land year-round.While the most dramatic changes are experienced by those in the várzea, the numerous ways the fluctuating waters affect food production and the habitats of key species in the diet also have the potential to impact the food security of ribeirinho people living in terra firme zones.
There is mounting evidence of food insecurity and malnutrition among ribeirinho populations (Piperata et al., 2011(Piperata et al., , 2013(Piperata et al., , 2016)).Tregidgo et al. (2020) documented ribeirinho households skipping meals and experiencing entire days without eating during the high-water season, an indicator of severe food insecurity.Others have reported widespread malnutrition, including high rates of anemia among young children (Kempton et al., 2021;Torres et al., 2022).Recent research argues for the importance of local wild foods in supporting human nutrition in the region.For example, Torres et al. (2022) demonstrated a lower prevalence of anemia among Amazonian children who consumed more wildmeat.Moreover, Heilpern et al., (2021a, b) estimate that declines in fish stocks (defaunation) and dietary shifts from local wildmeat to purchased domestic meat (i.e., the nutrition transition) may have negative effects on dietary iron intakes.
The extent to which difficulties obtaining sufficient nutritious food due to seasonal flooding is affecting dietary patterns and, hence, potential nutrient deficiencies in ribeirinho communities in Central Amazonia remains poorly understood.To our knowledge, only one study reported dietary data across seasons (Murrieta et al., 1999) and was conducted in an eastern Amazonian tidal estuary where the hydrological cycle is far less pronounced (Vogt et al., 2016), and attributed seasonal dietary changes to periods of crop harvest and income, as opposed to changing water levels.

Research Hypotheses
To address these knowledge gaps, we investigate the extent to which the flood pulse influences dietary patterns and the intake of energy and macronutrients among Central Amazonian communities by comparing household dietary patterns across the hydrological cycle, and between várzea and terra firme households.We hypothesize that (H1) the consumption of fish and local crops will be lower during the high-water season, with resultant shortfalls in macronutrient intakes and, thus, energy.While we expect H1 to be true across várzea and terra firme environments, given the greater abundance of large terrestrial mammals (preferred by hunters) (Haugaasen & Peres, 2005) and yearround access to unflooded agricultural land in the terra firme, we hypothesize that (H2) seasonal declines in energy and macronutrient intake will be more pronounced in várzea than in terra firme households.
We begin by characterizing the diet of the ribeirinho population in the study area, including their main sources of carbohydrate, protein, fat, and energy.This is followed by an analysis of household-level dietary energy and macronutrient intake across the hydrological cycle (H1) and between várzea floodplain and terra firme communities (H2).

Study Area
We conducted our research in the Mamirauá and Amanã Sustainable Development Reserves, located in the Mid-Solimões region of Central Amazonia, Brazil.Mamirauá and Amanã are Brazil's first two Sustainable Development Reserves, which aim to ensure the protection of the natural environment while allowing traditional populations to live on and use these lands for subsistence purposes.The reserves are well studied due to two decades of research on socio-biodiversity, conservation, and human quality of life (de Andrade et al., 2021;Nascimento et al., 2019).
Based on the census closest to the time of this study (IDSM, 2011), the Mamirauá Reserve had a population of approximately 10,386 people, living in 184 settlements, made up of communities (comunidades) and smaller rural settlements (sítios).Amanã had approximately 3,860 residents living in 86 settlements.We selected communities where the Mamirauá Institute had the strongest rapport, and where the residents had the greatest understanding of our research, particularly that it would not be used for environmental enforcement purposes.This was essential to maximize the chance of honest reporting, particularly in reference to the consumption of sensitive foods, like wildmeats.These were communities where a long-term wildlife hunting monitoring program was ongoing.As such, data for this study come from nine ribeirinho communities, six of them located in várzea (São Francisco Do Aiucá, Barroso, São Raimundo do Jarauá, Boca do Mamirauá, Sítio Fortaleza and Nova Jerusalém) and three in terra firme environments (São José do Urini, Bom Jesus do Baré and Boa Esperança) (Fig. 1, Table S1).Note that each terra firme community, while built on permanently unflooded land, is located on the banks of a major water body (river or lake).
The main economic activities in the reserves are fishing, agriculture, the exploitation of timber and non-timber forest products (e.g., açaí fruit Euterpe precatoria in both environments and Brazil nuts Bertholletia excelsa in the terra firme), handicrafts, and community-based tourism (Peralta & de Lima, 2013).Communities throughout the reserves are also involved in natural resource management, most notably of the pirarucu fish (Arapaima gigas), which has increased local fish productivity and household incomes (Gonçalves et al., 2018).Average household income within Fig. 1 Map of the Mamirauá and Amanã Sustainable Development Reserves, showing the extent of várzea (light blue) and terra firme (green) environments and the locations of the study communities (red triangles).See Table S1 for details on the location of each study community within reserves and environments the reserves (rural) is greater than that of households located in the municipalities and region overall (rural and urban), although they are still low at around US$5,000 per year.This income is largely derived from government benefits (45%), household production (37%), and wages and services (17%) (Peralta & de Lima, 2013;Moura et al., 2016).

Seasonal Changes in Water Level
The Mamirauá Reserve is composed of várzea habitat, meaning that the entire 1.1-million-hectare area is submerged by water during the high-water season.The Amanã Reserve covers an area of 2.3 million hectares and includes várzea and terra firme zones, such that much of the land remains unflooded throughout the year.
Although there is some interannual variation in water levels, annual flooding is cyclical, with the high-water period occurring from May until mid-July and the low-water period being between September and November (Ramalho et al., 2009;Fig. 2).Based on water level data collected between 1990-2020, there is a mean change of 10.9 m (± 1.87) between the high and low-water levels ( Fig. 3).
Despite increases in hydrological extremes (exceptional high-and low-water levels) over the past couple of decades, water level data (Fig. S2) show that our study period can be considered representative of an average year.Daily water level data (meters above sea level -masl) were obtained from long-term monitoring in the Mamirauá Sustainable Development Reserve (Instituto de Desenvolvimento Sustentável Mamirauá, 2021).In linear mixed models, we treated water level as a continuous variable (masl).However, when discrete categories were used in descriptive statistics, SIMPER analysis, and bar charts, we used the term "high-water" when levels were ≥ 33 masl.The rest of the time was referred to as "low-water" (Ferreira-Ferreira et al., 2015).

Dietary Data Collection
Dietary data were collected between 2005 and 2008 by residents of the studied communities, who received prior training.Visits were made to individual households during mealtimes: breakfast, morning snack, lunch, afternoon snack, and dinner.Over this period, data were collected three times a week from at least three households in each of the nine study communities: two of the households were randomly selected and the third was always the data collector's household.In other words, the collector's house was subject to continuous collection (three times a week for 12 months), while the other two houses could change every week in effort to include as many households as possible per study community.The collector returned to the same household once all community households willing to participate had been sampled.Ideally, once a household was selected, the daily visits occurred on three consecutive days, however, this data collection protocol was not always feasible.Data collectors avoided recording data on holidays and during other festivities, when changes to regular dietary patterns are likely.In addition to gathering dietary data, data collectors recorded the total number of household members, and the number of people present during each meal as not all household members consumed all meals together.Dietary data were collected using one of three methods.The first, and preferred, method was when a data collector was in the household at the time of food preparation/consumption and could weigh all foods using a balance (Pesola, 3 kg capacity, 10 g precision).The second method entailed residents estimating quantities and weights of items that were consumed prior to the arrival of the data collector.For example, a resident would report consumption of two fish with an estimated weight of 1.0 kg each at lunch, or state the number of a certain fruit, packet of crackers, or canned good.To avoid overestimation, for whole unprepared fish (5% of fish consumption records) we adjusted the weight measured/reported using a reduction factor of 0.86, based on field data collected by the lead author, to account for weight loss in cleaning/preparing (i.e., yield minus scales and guts).The third method, also employed in cases where the data collector was not in the home at the time of consumption, was used for items whose weights could not be easily estimated by residents (e.g., sugar, coffee powder, or liquids in general).In these cases, quantities were reported by residents using measures such as liters, cups, spoonfuls, etc.To convert these estimates into weights, we re-measured the reported quantities of these items in the laboratory using similar utensils and containers, calculated an average weight, and then used this standardized value in all analyses.A similar approach was used for items such as fruits, whereby we purchased the items in the nearby town, weighed them, generated an average and used the standardized value in all subsequent analyses.For processed foods such as crackers and canned meats, we used weight data from packaging to estimate intakes.The data collected using the three methods were collated within a single database.
To convert the dietary data into energy and macronutrient intakes we used two food composition tables: the Brazilian Food Composition Table (Tabela Brasileira de Composição de Alimentos; TACO) (NEPA, 2011) and the Amazonian Food Composition Table (Tabela de Composição de Alimentos Amazônicos).The latter is specific to Amazonian foods (Aguiar, 2019).In the few cases where an item was absent from both tables, we used nutritional information from the packaging of a brand commonly consumed in the region.See online shared data for the nutritional values used.

Datasets for Hypothesis Testing
Systematic collection of dietary data in rural communities across the hydrological cycle was challenging.Thus, our database had limitations including missing information on date of collection, the weight/quantity of reported food items, and/or the number of individuals who partook in specific meals.In an initial cleaning, we deleted the food items (1.7% of total) that lacked household ID and/or consumption date.To maximize data usage to address our research hypotheses we then created three datasets from the resulting cleaned database.
The first and largest dataset was used to characterize the diets of ribeirinhos in frequency of consumption of food items.The second dataset, which was used to characterize the diet in biomass consumed and to test H1, was restricted to days when we had complete dietary data, meaning that all meals were accounted for, and food items had a weight.To achieve this, we deleted any day in which a household reported having consumed fish or farinha (the most frequently food items), but where these records lacked weight data (23% of household days deleted).
To investigate spatiotemporal differences (H2), we created a third dataset, which was limited to days where we had both weight data on all foods consumed (i.e., starting with the second dataset) in addition to data on the number of individuals present at each meal.Using these restrictions, a further 18% of meals were excluded.Individual daily dietary adequacy was calculated by dividing the total intake of energy, carbohydrate, fat, or protein in the home over the course of the day by the number of people who participated in each meal.

Food Groups
Food items were categorized into the Food and Agricultural Organization's (FAO) 12 aggregated food groups: cereals; white tubers and roots; vegetables; fruits; meats; eggs; fish and other seafood; legumes, nuts and seeds; milk and milk products; fats and oils; sweets; and spices, condiments, and beverages (FAO, 2010).These data were also used to characterize the diet and investigate the principal sources of energy and macronutrients, and how they varied seasonally.
To investigate the impact of seasonality on macronutrient intake, we focused on five food categories (fish, farinha, fruit, wildmeat, and domestic and processed meat), which consisted of the four most nutritionally important FAO food groups according to our calculations of macronutrients (Fish, White tubers and roots, Fruit, and Meats), with meat divided into two groups: wildmeat or domestic and/or processed meat.Note that in this paper we follow the definition of wildmeat used by Ingram et al. (2021): "the meat and other body parts of wild terrestrial and aquatic animals (excluding fish) used for food."Since farinha is the dietary staple, we considered it on its own rather than group it with all white tubers and roots.

Statistical Analysis
Statistical analyses were performed in R statistical software version 4.0.2.(R Core Team, 2020).LMMs (linear mixed models; LME4 R package) were used to test H1 (trends in food and macronutrient intake with water level), H2 (differences between várzea or terra firme environments), and the interaction between water level and environment.In addition to the fixed effects for these variables, LMMs include random effects to account for variation associated with repeated measures, which in this case we included as community and household.The sampling unit used for statistical analysis was a household day-i.e., one household on one day.As meat consumption was absent from most daily household records (10% contained wildmeat and 7% domestic and processed meat), zero-inflated GLMMs (glmmTMB package) were used to analyze their consumption.Mean values per person per day were calculated as the mean household values, divided by the maximum number of people present at a meal that day, and standard deviations presented in the text.Similarity percentage (SIMPER) analysis was used to identify the FAO food groups that most contributed to the dissimilarity in consumption between the high-and low-water seasons.

Characteristics of the Local Diet
A dataset complete with all daily meals and biomass data for all food items (dataset 2) was used to characterize the dietary patterns of the ribeirinho people in the nine study communities.This dataset included 20,215 kg of food consumed over 8,119 meals by 177 families over a 45-month period.We identified 237 distinct foods, which we grouped into the 12 FAO food categories (Table 1; Fig. S1; online shared data).
The diet was dominated by fish and farinha, making up 44% and 26% of the total consumed biomass respectively (Fig. 4).Note that farinha made up 91% of the FAO food group "white tubers and roots."In terms of mass, other important food groups included spices, condiments, and beverages (87% coffee and 12% soda), fruits, meat (76% wildmeat and 24% domestic/processed), and cereals and their products (41% crackers, 25% rice, 18% wheat bread and 5% pasta).The most common breakfast foods were sugared coffee, fruit, crackers, powdered milk, wheat bread, farinha and tapioca/beiju, a pancake made from manioc starch.On average, households consumed both fish and farinha twice a day (modal average), typically during the lunch and dinner meals.Other common items consumed at both lunch and dinner were wildmeat, domestic and processed meat, rice, beans, and fruit.Daily snacks most often consisted of sugared coffee, fruit, crackers, soda, farinha, powdered milk, and wheat bread.

Sources of Dietary Energy and Macronutrients
Per capita consumption was calculated from a trimmed dataset, limited to meals where we had data on the number of individuals present (dataset 3).This dataset included 16,489 kg of food consumed over 6,573 meals.Average food consumption was 1.59 (± 1.26) kg per person per day, amounting to 3,118 (± 2,459) kcal, 444 (± 377) g of carbohydrate, 181 (± 153) g of protein, and 68 (± 68) g of fat (Table S2).Fish was the main source of protein (78%) and fat (66%) but, also, an important energy source (31%).Farinha was the main source of dietary energy (48%) and carbohydrate (80%).Meat (wild and domestic) provided 10% of protein and 9% of fat intake.Fruits were also an important source of fat (10%) (Fig. 5).

Spatiotemporal Variation in Food Consumption Patterns
There was no significance difference in the per capita daily consumption (weight) of fish (p = 0.45), fruit (p = 0.72) or domestic and processed meats (p = 0.98) between várzea and terra firme communities.More farinha was consumed in várzea communities (p < 0.001), while more wildmeat was consumed in terra firme communities (p < 0.001), Daily per capita fish (683 ± 687 g), farinha (399 ± 379 g), fruit (110 ± 360 g), wildmeat (58 ± 221 g) and domestic and processed meat (22 ± 112 g) consumption varied with water level (Fig. 6).Fish consumption declined (p < 0.001) as water levels rose, and no interaction was found between water level and environment (p = 0.08).Mean per capita daily fish consumption fell from of 756 (± 677) g during the low-water (< 33 m above sea level) to 569 g (± 687) in the high-water (≥33 m above sea level) season.
A significant interaction term for wildmeat consumption (p < 0.001) demonstrates different seasonal trends between environments.Terra firme communities ate more wildmeat as water levels rose (p < 0.001).However, no seasonal trend  Rising water levels were associated with increased domestic meat consumption (p < 0.001), however, we found no interactions between water level and environment (p = 0.61).The per person per day consumption rate of domestic and processed meat was 18 (± 86)g in the low-water season, but rose to 29 (± 143)g in the high-water season.

Spatiotemporal Variation in Energy and Macronutrient Ingestion Patterns
In the várzea, there was a significant decline in energy (p < 0.001), carbohydrate (p < 0.001), protein (p < 0.001), and fat (p < 0.001) consumption as water levels rose.In the terra firme, we found no change in protein consumption across the hydrological cycle (p = 0.056).However, in contrast to várzea households, we found a significant increase in energy (p < 0.001), carbohydrate (p < 0.001), and fat (p < 0.001) consumption with rising water levels in the terra firme (Fig. 7; see Table S3 for full statistical results).
Mean energy (3,118 ± 2,459 kcal) and protein (181 ± 153 g) ingestion exceeded the recommended intake of 2,000 kcal and 80 g of protein per person per day (FAO/ WHO/UNU, 1985), even within seasons and environments (Table S2).However, we documented considerable variation, as expressed by the large standard deviations, providing evidence of high interhousehold variation.Despite high average ingestion, per capita energy and protein ingestion fell below recommended levels on a third (35%) and a fifth (22%) of household days (i.e., one household on one day).

Spatiotemporal Variation in Sources of Protein and Fat
Fish remained the main source of protein (78.4%) and fat (66.2%) in both high and low-water seasons, and across environments.However, the contribution of other food groups to total protein and fat ingestion increased during the highwater season (when less fish was consumed), particularly in the terra firme.Of the FAO food groups, fish and meat were identified by SIMPER analysis to contribute the most to seasonal dissimilarity in protein intake in the terra firme (Table S4), with the contribution of meat to total protein consumption doubling from 9.6% (18 ± 50 g protein per person per day) to 22.6% (34 ± 86 g protein per person per day) between the low and high-water seasons (Fig. 8).This increase is largely attributed to wildmeat (93% of protein from meat).Of the FAO food groups, fish and fruit and meat were identified by SIMPER analysis to contribute the most to seasonal dissimilarity in fat intake in the terra firme, with the contribution of fruit to fat ingestion also doubling from 12.6% to 27.1% (Fig. 9) between the low and high-water seasons.This was due mainly to increased consumption of the palm fruit pupunha (13.1% of total fat consumption and 73.7% of fat from fruit).Although less significant, the contribution of meat to fat ingestion also increased from the low to high-water season in the terra firme from 7.4% (5 ± 13 g fat per person per day) to 12.8% (9 ± 29 g fat per person per day) and in the várzea from 6.0% (4 ± 15 g fat per person per day) 12.3% (8 ± 27 g fat per person per day).

Discussion
At the time of data collection, the diet of ribeirinhos from the Mamirauá and Amanã Sustainable Development Reserves was dominated by fish and farinha, making up 70% of the total food mass consumed.These dietary staples were complemented by fruit, wildmeat, and market goods.Farinha was the principal source of dietary carbohydrates and energy.Fish, by far the most important source of protein and fat, was also the second most important source of energy.In this regard, our results are consistent with previous studies among ribeirinho populations (Murrieta & Dufour, 2004;Piperata et al., 2011Piperata et al., , 2013)).
Our data reveal a per capita fish consumption of 683 g per day (249 kg per year), which to our knowledge is the second highest fish consumption rate ever published globally, after 805 g per day, also in the Brazilian Amazon (Rio Japurá) (Fabré & Alonso, 1998).Research showing that residents of the nearby town of Tefé are the highest urban fish consumers in Brazil (Ferraz & Barthem, 2016), eating four times the national average, provides further evidence of high fish

Dietary Resilience as a Response to Seasonal Flooding
Reductions in farinha consumption during the high-water period were only found in várzea habitats, which may be explained by the seasonal inundation of várzea croplands, distinct from the year-round production in the terra firme.Although the relative ease of farinha (a dry flour) storage may buffer seasonal trends in its consumption, it has been observed that household stores commonly run out in várzea communities (Lima-Ayres, 1992).Increased farinha consumption in terra firme communities helps to explain the increase in energy and carbohydrate intake as water levels rise.
A key finding of this study is that fish consumption rates plummet as water levels rise.This result was expected (H1), given that catch rates are known to fall in the high-water season as fish habitats expand (Endo et al., 2016;Tregidgo et al., 2020).Prior to the most significant contemporary dietary changes associated with the nutrition transition, reduced fish catch rates were associated with decreased human body weights in the study region during the high-water season (Lima-Ayres, 1992).Moreover, since the recognized period of nutrition transition, this seasonal reduction in fish catch has been associated with severe food insecurity in other Central Amazonian ribeirinho floodplain communities (Tregidgo et al., 2020).
Despite the negative impacts of seasonal flooding on fish consumption rates, our data suggest that the average household maintained high fish consumption across the hydrological cycle, resulting in average daily protein consumption well above the recommended level of 80 g per capita (FAO/ WHO/UNU, 1985).However, it was not uncommon for a household to lack fish during the high-water season, and the fact that protein consumption was found to be below recommended levels on a fifth of household days demonstrates food instability -specifically, the lack of adequate food at all times.
Our findings indicate that while ribeirinho families are vulnerable to food insecurity, i.e., seasonal instability in availability and access due to the flood pulse, they display high dietary flexibility (and hence resilience) across the hydrological cycle, shifting both between and within food groups.For example, the ability of the average household to maintain high, albeit reduced, fish consumption during the high-water season appears related to seasonal shifts in the fish species consumed (Fig. S3).These data demonstrate the versatility of local fishers to change fishing grounds and techniques to exploit local biodiversity (Tregidgo et al., 2021).As well as being a likely coping strategy to seasonal shortfalls, high dietary species richness is also associated with high diet quality (Lachat et al., 2018).For example, Heilpern et al. (2021a) show that greater fish diversity in the diet is associated with nutritional adequacy (focusing on micronutrients).While the ribeirinho diets in this study were dominated by two general food groups (fish and farinha), our data suggest that the high species diversity of fish, wildmeat, and fruits consumed may permit a nutritious diet with macronutrient adequacy.
In addition to shifting to different fish species, increasing meat consumption appears to be another coping strategy employed by these ribeirinho households during the high-water season.However, unlike other recent studies (Piperata et al., 2016;van Vliet et al., 2015), the increase in meat consumption we document is largely from wildmeat, rather than domestic and processed meats (1.3% of food consumed).Given the increase in domestic and processed meat consumption in Amazonian communities in recent years (Piperata et al., 2016), we would expect greater consumption if the study were to be repeated today.
Wildmeat was the second most important protein source in these ribeirinho communities.Wildmeat has been identified as an important alternative to fish in subsistence communities worldwide during periods of fish scarcity (Brashares et al., 2004), including during the high-water season in Central Amazonia (Endo et al., 2016;Tregidgo et al., 2020).As wildmeat and fish species have approximately the same protein content (~ 20% of biomass) (Aguiar, 2019;NEPA, 2011), wildmeat may act as a protein safety net during the seasonal flood pulse.Although nutritional data are unavailable, wildmeat also appears to be an important source of micronutrients, demonstrated by evidence that ribeirinho children who consume more wildmeat have lower rates of anemia (Torres et al., 2022).In fact, the high rate of anemia among ribeirinho children may be attributed to their heavy Fig. 8 Changes in dietary sources of protein between seasons in terra firme and várzea households reliance on fish, which (based on the only available evidence from European wild and domestic meats) is likely to be much lower in iron than wildmeat (Heilpern et al., 2021a;Strazdiņa et al., 2013).
Seasonal changes in fruit consumption may also be indicative of dietary resilience in these ribeirinho communities.Throughout the Amazon region, mature fruits of many species are only seasonally available (Haugaasen & Peres, 2007;Phillips, 1993;Shanley et al., 2011).One seasonally available species, the palm fruit pupunha, made a surprisingly large contribution to fat consumption during the highwater season in terra firme households.The contribution was so large that despite the reduction in fat intake from fish, total fat ingestion in the high-water season increased in these households.Pupunha, which has a fat content of 27%, is a culturally important fruit, often consumed in large quantities as part of meals or snacks.While we appreciate the considerable variation in nutrient composition of wild fruit species in different environments (Burlingame et al., 2009), we remain confident that our observations reflect the important contribution that fatty Amazonia palm fruits can make to local diets when mature fruits are available.

Variation in Diet and Food Stability Between várzea and terra firme
Significant increases in wildmeat consumption during the rising waters were observed only in the terra firme.This may be explained by the greater abundance of large terrestrial mammals such as tapir, peccary, and paca (Haugaasen & Peres, 2005), and the improved aquatic access to some terra firme hunting grounds through navigable floodwaters.Many of these species are absent from the várzea either permanently or during the high-water season due to their requirement for dry land (Valsecchi, 2012;Fig. S4).
As was the case with wildmeat, an increased reliance on fruit was only seen in terra firme communities as water levels rose.The negative trend between fruit consumption and water level in the várzea is likely partly due to the inability to maintain many fruit crops on flooded land-a Fig. 9 Changes in dietary sources of fat between seasons in terra firme and várzea households key exception being the açaí palm.This trend is exemplified by watermelon, which is planted in várzea floodplain areas (often beaches) and thus must be harvested prior to the flood.Terra firme communities may be better positioned to exploit upper-land habitats where perennial fruit trees, including pupunha and other palms with fatty fruits, grow and can be managed.
Dietary studies of traditional Amazonian communities tend to focus on protein, yet fat is thought to be scarcer than protein in tropical forests (Sirén & Machoa, 2008).Hence, fatty palm fruits may be an important and neglected contributor to nutritional security in the region (Kahn, 1991).These data, overall, support H2, that seasonal declines in energy and macronutrient intake will be more pronounced in várzea than in terra firme households.This appears to be due to a greater ability among terra firme households to exploit alternative species and foods when the flood waters rise.We note that not all Amazonian terra firme communities are equal, and that the seasonal food stability demonstrated in this study of a protected area near nutrient-rich white-water várzea may not transpire in other terra firme environments, for example in overhunted or unprotected areas, or hunting grounds associated with nutrient-poorer black waters, or far from aquatic ecosystems.

Conclusion and Future Research
The seasonal fall in macronutrient intake evidenced in this study demonstrates food and nutritional instability in ribeirinho communities, which is more pronounced in várzea than terra firme communities, largely as a function of access to nutritious foods other than fish in the terra firme.In demonstrating variation in macronutrient intake along the hydrological cycle, this study stimulates further research questions concerning micronutrient intakes and climate change impacts.Global food and nutritional security concerns are shifting from macronutrient to micronutrient malnutrition (Haddad et al., 2016), known to be highly prevalent in traditional Amazonian communities (Kempton et al., 2021;Torres et al., 2022).Micronutrient composition differs dramatically among wild species, as evidenced with marine fish (Hicks et al., 2019).Moreover, the nutritional role of wildmeat (Torres et al., 2022) and fruit (e.g., Aguiar, 2019;Rodrigues et al., 2001) may be disproportional to their consumed biomass.Hence, the nutritional (micronutrient) composition of the most consumed species of fish, wildlife, and fruit must be revealed and factored into future dietary studies.
Climate change is causing more prolonged flooding, higher peak water levels, drier dry seasons, and a less predictable hydrological regime in Central Amazonia (Marengo et al., 2013;Fleischmann et al., 2023) Hence, while ribeirinhos have adapted remarkably to the extremes of the flood pulse, their vulnerability to food insecurity is likely to intensify as the changing hydrological cycle impacts fish populations and catch rates, shortens and reduces the predictability of crop planting periods (namely the staple manioc), and dries up aquatic transport networks (Parry et al., 2018).Interdisciplinary research is essential for understanding these impacts and facilitating adaptations to an increasingly extreme environment.

Fig. 4
Fig.4Foods consumed based on total mass using the 12 FAO food groups(FAO, 2021a)

Fig. 5
Fig. 5 Principal sources of energy and macronutrients, organized according to the 12 FAO food groups (FAO, 2010).Percentage contribution of food groups to A energy, B carbohydrate, C protein and D fat consumption

Fig. 6
Fig. 6 Changes in consumption of major food categories in A várzea and B terra firme environments across the hydrological cycle.Grey areas show confidence intervals.Sample unit is a household day (one household on one day)

Fig. 7
Fig. 7 Changes in daily per capita A energy, B carbohydrate, C protein, and D fat ingestion, in várzea (blue) and terra firme (red) environments across the hydrological cycle.Grey areas show confidence intervals

Table 1
Categorization of local foods within the 12 FAO food groups(FAO, 2010)a The vegetable group is a combination of vitamin A rich vegetables and tubers, dark green leafy vegetables, and other vegetables b The fruit group is a combination of vitamin A rich fruits and other fruits c The meat group is a combination of organ and flesh meat , corn seeds (milho de mungunzá), corn bread, popcorn, corn snack (salgadinho de milho), wheat flour, spaghetti, instant noodles, bread, sweet bread, cream crackers, fried pastries 2. White tubers and roots Farinha (manioc flour, Manihot esculenta), tapioca, manioc, tucupi (manioc derived condiment), potato, Spices, condiments, and beverages Salt, colorau (spice from urucum Bixa orellana), cumin, coriander, black pepper, pimento de cheiro pepper, chili pepper (Capsicum chinense), coffee, barley coffee, soda, powdered fruit drink, sugar cane juice, artificial sweetener