Study area

The research was carried out in a rural community within the coastal “Restinga” vegetation of Piaçabuçu, situated on the southern coast of Alagoas state. Piaçabuçu spans an area of 243.686 km2, housing a population of 15,908 individuals [28]. It features a tropical ‘As’ climate in the Köppen and Geiger classification, with an average annual temperature of 25.3 °C and an annual rainfall average of 1283 mm [29]. Notably, the municipality is designated with two sustainable use Conservation Units: the federally instituted Piaçabuçu Environmental Protection Area, established in 1983, and the state-sanctioned Marituba do Peixe Environmental Protection Area, created in 1988.

The Marituba do Peixe Environmental Protection Area spans 18,556 hectares and extends over portions of the Alagoan municipalities of Piaçabuçu (45%), Feliz Deserto (43%), and Penedo (6%) [30]. This area boasts diverse vegetation, including native “Restinga”, “Várzea”, and other forest formations [30]. Within the Indirect Influence Area of Marituba do Peixe Environmental Protection Area lies the village of Retiro (depicted in Fig. 1), which was the focal point for the ethnobiological segment of this study.

Fig. 1

Geographic Location of the Retiro Community in the Municipality of Piaçabuçu-Alagoas, Brazil

The Retiro community is structured with a residents’ association and a family farmers’ association. It is equipped with a primary healthcare unit and a municipal elementary school. The predominant faith among residents is Christianity, represented by two Catholic and two evangelical churches. Currently, the community comprises approximately 288 families, a decrease of 81 families since before the COVID-19 pandemic, as reported by Gomes et al. [31]. This discrepancy may be partly due to some families not being documented, a requirement for health unit registration.

Retiro was selected for this study due to the local reliance on plant resources for both food and wood. The community’s economy is significantly driven by the extraction and commercialization of wild food plant fruits [31], along with shrimp and fish [32]. Wood resource extraction for personal use and commerce, particularly firewood, charcoal, and materials for fencing, is also prevalent. These resources are marketed through open markets or direct orders in Piaçabuçu and Penedo, whereas wood products are solely distributed by order.

Firewood is the primary cooking fuel in the community, though some households use both cooking gas and firewood. Meals are typically prepared on traditional clay or makeshift brick stoves. Firewood also serves in roasting shrimp and baking cakes from rice straw, a common bait for shrimp in local fishing gear known as “cóvu.”

Architecturally, many “taipa” houses (rammed earth) are present within the community, often serving as dwellings for individuals from other regions staying temporarily in the area.

Ethical and legal aspects of the research

This research project received approval from the Research Ethics Committee by Federal University of Alagoas (UFAL), No. 1998673, securing authorization for studies involving human participants as per the stipulations of National Health Council Resolution 466/2012. Additionally, scientific activities involving the collection and transport of botanical specimens within the Marituba do Peixe Environmental Protection Area were duly registered with Chico Mendes Institute for Biodiversity Conservation/Biodiversity Authorization and Information System (ICMBio/SISBIO), No. 87,112-1.

To ensure ethical compliance, all community members aged 18 or over—to whom the objectives of the research were explained—and who consented to participate, were asked to provide a signature or thumbprint on the Informed Consent Form (ICF), as well as on the image use authorization form.

Data collection

Data collection was carried out in three distinct phases: a participatory workshop, a forest inventory, and checklist-interviews.

1st data collection stage: participatory workshops

Participatory workshops with the residents of the Retiro community aimed to identify significant wild food plant species for consumption and commercial use, as well as their harvesting locations. These workshops were facilitated by local leaders and a researcher from the Laboratory of Biocultural Ecology, Conservation, and Evolution (LECEB), who had previously interviewed community members. The residents were recruited through door-to-door invitations on the day before the workshop was held.

In the inaugural participatory workshop, participants were asked to list the wild plants they harvested for sale or consumption. They recorded the common names on a piece of cardboard, selecting eleven for further discussion. We then asked which of those species were most important for sale and consumption within the community, and they ranked the top five in order of importance. Additionally, the workshop served to note wood resources tied to food plants and their utilization for consumption and commerce within the community.

Thirteen women and three men, ranging in age from 31 to 82, contributed to this first workshop. While all were identified as gatherers, some also engaged in agriculture and fishing. A follow-up workshop sought to enrich this data with contributions from another set of gatherers (n = 17), including eight newcomers. This session, comprising thirteen women and four men from the same age bracket, validated the initial findings regarding species and harvesting sites.

Participants utilized a detailed satellite image from Google Earth to denote areas frequented for food and wood collection. An overlay of transparent acetate allowed them to make corrections directly on the map.

After pinpointing these areas, we selected those most frequented for the harvesting of both food plants and wood, prioritizing locations where the ranked key species were prevalent. Among the listed key species—Myrciaria floribunda (H. West ex Willd.) O. Berg (“cambuí”), Genipa americana L. (“jenipapo”), Psidium guineense Sw. (“araçá”), Spondias mombin L.(“cajá”), and Tamarindus indica L. (“tamarino”)—only the first three were represented in the forest surveys due to their presence in forests. Although S. mombin and T. indica are also key species in the region, the former occurred at the edges of roads or in backyards, while the last is found in fenced area with wire, with reports of increasing cultivation for pulp production by large landowners. Therefore, our study only included three out of the five key species, as well as the species that co-occur with them.

Three sites were thus chosen for the forest inventory: two with a natural predominance of key species and one characterized by a more generalized distribution of various plant species, including those bearing edible fruits.

Gomes et al. [31], the research design we adopted gave precedence to examining species that occurred alongside the key species; consequently, not all food and wood plants were included in our scope. Notably, Schinus terebinthifolia Raddi, while not a key species within Retiro, is a significant commercial species in the community and the most important commercial plant in neighboring areas, such as the Fazenda Paraíso settlement [31].

2nd data collection stage: forest inventory and field herbarium

The research included a forest inventory as part of a larger investigation of our research group into plant resource utilization within the region, although ecological data is not part of this study. For the purposes of this study, the forest inventory was only used to identify and collect species that co-occur with the key species. Without the forest inventory, we would have no baseline for the field herbarium (notebook with exsiccates of the species used as a visual stimulus during the application of the checklist-interview technique), since we would not know which woody plants co-occur with our key species. Therefore, although it is not our purpose to present results on forest structure and composition, the inventory was fundamental for the research. Exsiccates of these species were included in the field herbarium based on their abundance, as detailed below.

The sites selected for the inventory were privately owned yet accessible to local gatherers. Two of these sites fell within the Marituba do Peixe Environmental Protection Area boundaries in Piaçabuçu, while the third was in the municipality of Penedo, not included in this protection area but still proximal to the community.

We established five permanent plots, each measuring 50 × 20 m, and further divided these into 50 smaller subplots of 10 × 10 m situated within the primary native vegetation gathering sites designated during the workshops. This amounted to 0.5 hectares per area, with a total of 1.5 hectares surveyed across all areas.

During the inventory, we collected at least three reproductive samples of each plant species within the plots for identification and to assemble a field herbarium for use in subsequent interviews. Certain species, commonly referred to as “ingá” and “pau d’arco”, lacked fertile material at the time of collection, leading us to categorize them as ethnospecies for the purposes of this study. Consequently, in our identification records, we referred to these simply as “ingá” and “pau d’arco”, acknowledging that these common names might represent multiple botanical species. Furthermore, the ethnospecies “cambuí”, although biologically uniform—belonging to the species Myrciaria floribunda (H. West ex Willd.) O. Berg—was recognized by some residents as having different ethnovarieties—a distinction not universally acknowledged. In our analysis, we accounted for each mention of “cambuí” by participants, even though the general data summary did not differentiate between ethnovarieties. For instance, if interviewee A identified two types of “cambuí” (Yellow and Red) and Interviewee B referred to one (a general “cambuí”), we recorded two entries for A and one for B in our database.

For the field herbarium, we mounted exsiccates from species with more than 15 individuals in the surveyed areas onto duplex paper of dimensions 42 × 29.7 cm and stored them in folders of matching size. The herbarium included 24 species in total and 2 taxa that were treated as ethnospecies.

Photography of each species was conducted in situ, capturing images that emphasized the plants’ distinguishing features: overall appearance, flowers and/or fruits, branches, and stems. These photographs were compiled into folders on a tablet, which was employed to display the images during interviews. Both the exsiccates and the photo folders were numerically coded to correspond with the identifiers on the interview forms, ensuring that interviewees were unaware of the plant names and assisting the interviewer.

The botanical collection phase commenced in November 2021 and concluded in April 2023, an extended period due to intermittent interruptions from COVID-19 peaks and flooding that hindered fieldwork.

A local guide with extensive knowledge of the vegetation provided assistance for all fieldwork involving local vegetation access. We adhered to standard botanical collection protocols, and the exsiccate samples were deposited at the Dárdano de Andrade Lima herbarium of the Agronomic Institute of Pernambuco.

3rd data collection stage: checklist interview

Before commencing the interviews (third stage), we mapped all Retiro households in May 2023. This mapping was imperative for sample size calculation due to the absence of a census record; the health unit’s data was limited to registered families. We determined that household heads (one per household) aged 18 or older present during our visit would be interviewed. Considering that some individuals reside in the community only for short periods, we established an inclusion criterion that only families living in the area for more than one year would be eligible for the study.

We ascertained the number of residences, including both occupied and vacant, to be 361, initially yielding a sample size of 187 residences based on a 95% confidence level and a 5% margin of error. Subsequently, we conducted a simple random selection.

As every house in the community was recorded, including unoccupied ones, some selected residences were vacant. Additionally, given the research’s focus on potentially harmful wood resource use within the Environmental Protection Area, some families were reluctant to participate. Therefore, from the 187 chosen residences, we could only conduct interviews in 81 interviews of them. To overcome refusals, flood-affected houses, and temporary residents, additional draws were made.

After all draws, we excluded unoccupied houses (n = 82), residences on flood-impacted streets (n = 12), households temporary inhabitants (n = 18) and refusals or unavailability (n = 74) from the sample. After three unsuccessful attempts to locate a household head, we inferred their non-participation.

A notable number of individuals opted out of the study, a figure aligned with expectations for wood use research in protected areas, mirroring findings from Medeiros et al. [3]. The considerable number of unoccupied houses in the community can be primarily attributed to their use as summer residences by individuals from nearby municipalities, taking advantage of the community’s closeness to the beach. Additionally, a number of these houses are situated in areas susceptible to flooding during the rainy season, which also contributes to their vacancy.

The final sample consisted of 115 individuals—81 women and 34 men. Interviews were conducted from May to July 2023. During interviews, we applied the checklist-interview technique [27] to ensure uniform visual stimuli across all informants, enhancing recall of all plant-associated uses.

Interviewees were shown photos of each species and queried on whether they recognized the species. Affirmative responses led to further questions on the plant’s name, its uses (food and wood), whether the interviewee actually used the species, parts utilized, commercial harvesting, and collection and sale sites. For recognized plants, a Likert scale rated: perceived availability (only for those interviewees that often frequent vegetation areas), wood quality by use category (fuel, construction, technology), domestic use for wood and food, and commercial use.

In the fuel category, wood is used as firewood or charcoal for generating energy, cooking food, and heating water or spaces. The construction category encompasses the use of wood in structures for territorial demarcation, building homes, shelters for animals, and storage of items (e.g., fences, posts, house lines, rafters, battens, doors, windows). Technology refers to the use of wood in manipulated items that are not intended for demarcating spaces, such as tool handles, benches, tables, chairs, canoes, and oars, among others [33].

The ratings and responses in Likert scale are presented in Fig. 2.

Fig. 2

Information collected using a Likert scale on the variables considered in this study

This classification facilitated the synthesis of scoring for perceived wood quality, allowing individuals to assign ratings by category rather than for each specific use. If a participant identified a plant as useful for wood but did not personally use it, we probed for the reasons behind this choice. We also asked if there were any of the mentioned plants that, despite being good for wood uses, the interviewee did not harvested. These questions were included to gather information on self-conscious protective behaviors associated with the food use of woody species.

Only for the ethnospecies “ingá” and “pau d’arco”, instead of showing the photos and exsiccate, we asked directly if the person knew them for food or wood uses. In case of a positive answer, we asked the same cycle of questions conducted for the other species. This was done because we did not obtain sufficient fertile material for the taxonomic identification of all species of “ingá” and “pau d’arco” during the various collection events.

Additionally, we gathered socio-economic data from all informants through structured interviews, including gender, age, occupation, income, place of origin, education, and length of residence. This information enabled the characterization of the socio-economic profile of the interviewees.

In this sense, the primary livelihoods include gathering, particularly collecting edible fruits, as well as pension, fishing, and agriculture, with some engaging in multiple occupations. A variety of other professions are represented to a lesser extent. The age of interviewees spans ages 18 to 82, with an average age of 48.14 years.

Most interviewees are literate (76.65%) are literate, of whom 73.91% have completed or partially completed basic education, and 1.74% have higher education qualifications.

The number of people occupying the residences ranges from one to seven residents. However, the majority of houses are occupied by: two or three residents (29.57%), followed by one or four resident(s) (15.65%).

Household incomes show substantial variation: (a) under one minimum wage (28.70%), (b) exactly one minimum wage (14.78%), (c) one and a half to two minimum wages (41.74%), (d) up to three minimum wages (13.04%), with a minority exceeding five minimum wages (1.74%).

Data analysis

For statistical analyses, we removed from the database any instances where species were identified for food purposes but not for wood purposes, as the focus of the research was on criteria for selecting wood plants. Consequently, non-wood plants were disregarded. Similarly, we excluded data from individuals who did not frequent forest environments to ensure that our information on species availability came from realistic assessments.

Our response variable, domestic wood use, was ordinal, as depicted in Fig. 3. Therefore, we utilized Cumulative Link Mixed Models (CLMMs), incorporating the interviewee as a random effect to account for the non-independence of information from the same individual. The CLMMs were executed using the clmm function from the R package ordinal.

Fig. 3

Widespread species model and key species-based model with their variables and respective measures

To evaluate the stability of our models and check for multicollinearity, we used the omcdiag function from the mctest package in R. We determined an absence of multicollinearity if none or at most one of the six indicators were positive. To circumvent multicollinearity, we constructed two models. The first model, termed the widespread protection model, assigned domestic and commercial food use values on a 5-point Likert scale based on reported usage intensity. For the key species-based protection model, food use was a binary variable: it took the value of 1 if the mention included the use of a key species, and 0 if the mention involved a key species only known but not used, or non-key species, regardless of usage.

Model selection was based on the most parsimonious option, as indicated by the lowest Akaike Information Criterion corrected for small sample sizes (AICc). We interpreted a ΔAICc (difference from the lowest AICc) of less than 2 as substantial support for the model’s inclusion among the best set of models, following Burnham and Anderson [34]. Following model selection, we computed a model average, which considered the average beta of all variables within the parsimonious models. Since the variables were standardized via z-standardization, we compared the relative effect sizes of all variables.

The variable ‘commercial wood use’ was not included in the models due to its limited mentions (n = 5) within the community and only six citations of species that are commercially traded for wood, exclusive of domestic use.

In addition to the explanatory variables related to food use, both models incorporated control variables for availability and quality, as previously identified in the literature as predictors of wood use [20, 24, 25]. Our quality indicator was the maximum perceived quality. It was determined by the highest Likert scale quality rating given by an interviewee for a species across the three categories of wood use. For example, if, for a given species, values of 3, 4, and 5 were assigned by an interviewee to the categories of construction, technology, and fuelwood, respectively, the maximum perceived quality would be recorded as 5.

To analyze the qualitative data on protection behaviors, we examined the responses, categorized them, and used descriptive statistics.