<b>Community-based management expands ecosystem protection footprint in Amazonian forests</b>

This study complies with all relevant ethical regulations, including ethical requirements (CAAE research ethics permit 52148721.6.0000.5013) and the Brazilian Biodiversity Authorization and Information System – SISBIO (84412-3). Study Area This study was primarily conducted within the ~2.58-million-hectare municipal county of Carauari (4° 52′ 58″ S, 66° 53′ 45″ W), State of Amazonas, Brazil, along the Juruá River, a major tributary of the Solimões (=Amazon) River. This region is strongly influenced by commercial and subsistence activities involving fishing, agriculture, and Euterpe (açaí) fruit and oilseed extraction (Newton et al., 2012). This region contains two contiguous sustainable-use protected areas: the 632,949-ha Uacari Sustainable Development Reserve (RDS Uacari, 5º43'58"S, 67º46'53"W; Decree No. 25,039 of 1st June 2005), and the 253,227 ha Extractive Reserve Médio Juruá (ResEx Médio Juruá, 5º33'54"S, 67º42'47"W; Decree of 4th March 1997). These reserves were decreed in 1997 and 2005, respectively, and currently contain ~4,000 inhabitants distributed across 74 communities, most of which near the river channel, along a fluvial distance of 800 km, in addition to communities located along the banks of oxbow lakes and perennial streams (Figure 1B). Resource governance resulting from arapaima co-management To ensure both economic and food security for rural communities, Fishing Accords (i.e. formal agreements) were widely negotiated in the mid-Juruá region during the 2010s. These accords involved local communities, including those outside protected areas, as well as the Fishers Cooperative of Carauari, the nearest urban centre. The agreements created three different categories of access to lake resources during the dry season, when lakes become clearly discrete geographic features where fish concentrate: (1) Subsistence-use lakes, which are intended to supply local subsistence needs, and which are restricted to artisanal fishers from the resident community who are responsible for guarding that lake; (2) Protected lakes, which are managed by local communities primarily as arapaima stock recovery sites, and exclude both commercial and subsistence fishing boats, except for a brief community-led offtake season based on a strict harvest quota predetermined by IBAMA, the Brazilian Natural Resources Agency (Campos-Silva and Peres, 2016); and (3) Production lakes, which are open-access to both commercial and subsistence fishers. A floating wooden watchtower is typically erected at the main strategic entrance to the lake. Equipped with makeshift hunting gear and subsistence supplies, these stationary posts, occupied by a small patrol unit and managed by the resident community, conduct year-round, 24/7 surveillance, often armed with shotguns. During the arapaima management season, some of the protected lakes are harvested by the resident community for a brief period of up to 5 days per year, according to a previously determined proportional harvest quota based on a stock assessment defined as the number of adult and juvenile arapaima counted at that lake in the previous year (see Campos-Silva and Peres, 2016). Annual arapaima counts began at several lakes along the mid-Juruá in 2005, and lake management was implemented in 2010 by a partnership between local communities, local associations, and federal and state agencies. Arapaima counts take place during the low-water season at each monitored lake each year, and the census data are forwarded to IBAMA. IBAMA then authorizes a lake-specific harvest quota of up to 30% of all adults (>1.5m in length) counted, depending on the fish processing requirements of the resident community and other extenuating factors. Data analysis Quantifying territorial protection We conducted participatory community mapping through semi-structured interviews with lake guards, community leaders, and community residents. First, we asked general questions to describe the surveillance dynamics, including the main actors, surveillance alternation dynamics, impact of seasonality on surveillance dynamics, surveillance pathways, conflict resolution strategies, and associated costs. Participatory community mapping occurred interactively using A3-sized hardcopy cartographic maps showing LANDSAT-8 satellite images in RGB (5,4,3) colour composition, with a scale of 1:100,000 for location and identification of each lake where territorial surveillance had been deployed by each community. Each lake management category was identified by outlining locations on the map using colour markers (Wartmann and Purves, 2017). Participatory community mapping was carried out with community residents who had extensive previous experience with both spatial landmarks across the waterscape, which is the main form of transport in this region, the overall landscape, and in-depth knowledge of arapaima co-management activities (Patton, 2015; Silvano et al., 2023). Floodplain mapping was carried out within the scope of either community meetings or visits to resident households (Saija et al., 2017). Experienced individuals were identified by community leaders. Arapaima co-management activities exert varying impacts at different spatial scales of influence (Figure 3A). First, there is a (1) direct scale of protection, represented by the immediate lake area where actual surveillance takes place. Second, there is an (2) effective scale of protection, which is represented by the total area within the community surveillance boundaries. Third, there is a (3) functional scale of protection, represented by the functional impact zone exerted by spatial exclusion, particularly related to the vagrancy and movement capacity of the target species protected at each lake. Finally, there is an (4) incidental scale of protection at which local communities indirectly protect large portions of upland (terra firme) forests farther inland by simply restricting entry to strategic access points within the more accessible adjacent floodplains. We emphasize that the collective surveillance dynamics are structured by the internal regulations of each community. These regulations establish the local rules that must be followed by both local residents and outsiders who may use the areas protected by the communities. The internal regulations outline the local norms, clearly stating that illegal activities and unregulated resource exploitation are prohibited in the lakes and adjacent forests, including illegal hunting, illegal logging, and gold mining, for example. Direct scale of protection During the mapping sessions, all lakes protected through surveillance that are managed by any given community were identified and further classed as direct surveillance areas, as they are the focus of management activities, and their total area was measured using the MapBiomas Água Project collection 1 dataset (MapBiomas 2021), which mapped all open water bodies across Brazil. Effective scale of protection Territorial surveillance for lake protection is a set of actions and adaptive strategies that occur on a full-time basis, but intensified in the dry season, to protect areas of management interest. These areas include subsistence-use, protected, and production lakes that are harvested for local subsistence. Surveillance aims to protect lakes from illegal harvesting by either local or external fishers, and any other exploitation activities that can disturb the lake and the surrounding forest, such as hunting and timber extraction. Surveillance strategies are continuously adapted according to the needs of each community and depend on the number of managed lakes, number of people available for surveillance, landscape context, and geographic accessibility of each lake. Surveillance is conducted by travelling around the perimeter of each lake by canoe or on foot, depending on the season, searching for any presence or signs of intruders. In several communities, floating wooden houses are placed at strategic entry points of access to lakes to optimize surveillance. During surveillance, lake guards cover a floodplain area much larger than the size of individual lakes, which we refer to as “effective scale of protection”, where illegal activities, including poaching, fishing, and logging, are excluded. To estimate the effective protection of each lake, we combined GPS tracks and spatial data recovered from interviews to map the daily paths that community guards frequently travelled to protect each lake. The area effectively protected, including seasonally-flooded várzea forest and open-water bodies, was estimated, including all reported paths on foot and/or canoes between lakes, and all strategic surveillance points that were frequently accessed by outside users attempting illegal fishing. Polygons drawn during participatory mapping were reproduced in QGIS 3.14 (QGIS, 2023) at the same scale using the corresponding satellite image to fine-tune estimates of the effective scale of protection. Functional scale of protection We also estimated the functional protection area of each lake based on the ranging ecology of giant arapaima (Arapaima gigas), the conservation target species in this arrangement. We therefore considered arapaima movement patterns, which had been quantified during a previous telemetry study (Campos-Silva et al., 2019), to estimate the capacity of each lake to function as a source area of individuals moving into depleted lakes and the spatial configuration of landscape-scale population gene flow, both of which can sustain ecological interactions and top-down control of food webs mediated by an apex predator (Campos-Silva et al., 2021). This was estimated using a 1,730-m buffer area around the dry-season perimeter of each lake (i.e. the direct scale of protection). This threshold value corresponds to the radius of an average circular Arapaima home range area, defined by the Minimum Convex Polygon formed by positional fixes obtained for 12 juveniles and adults. Six of these individuals were tracked in our study area in 2014 and seven in 2015 using conventional VHF telemetry, amounting to 309 locations, 125 and 184 of which during the dry and wet seasons, respectively (see Campos-Silva et al., 2019). Individual estimates are available in Table S.5. Incidental scale protection In addition to these three scales of protection, oxbow lake surveillance also incidentally protects all the rear areas of upland forests by closing off the physical accessibility to unauthorized users of the várzea floodplain. This strategy prevents non-resident loggers, hunters, and fishers from accessing upland areas, typically to stealthily exploit sub-canopy natural resources without the explicit consent of the local community. This scale of protection was estimated by multiplying the total width of várzea floodplains protected at the effective scale by a conservative 10-km length of upland forests that could be potentially affected by illegal extractive activities (Benítez-López et al., 2019; Peres et al., 2016). To assess differences in spatial extent between different scales of protection, an Analysis of Variance (ANOVA) was performed with the response variable on a logarithmic scale. Assumptions of normality of residuals and homogeneity of variances were evaluated using the Shapiro-Wilk and Levene tests, respectively. Assessing protection dynamics and costs To better understand local surveillance priorities according to the flood pulse dynamics, we organized focal group interviews at each community with 45 experienced fishers who had conducted local lake surveillance for at least 15 years. These focal groups were adept at mapping the seasonality of surveillance because of previous experience and fluctuations in water level change the accessibility to water-bodies and their vulnerability. Monetary costs of surveillance were acquired during interviews and encompassed general operational expenditure including fuel, food, and butane gas used as fuel to power outboard motors during surveillance routes, according to the unique ways in which each community carried them out. This excludes labour input and expenditure related to purchase and maintenance of wooden or aluminium boats, outboard motors, paddles, and infrastructure such as strategically positioned floating houses, which served to accommodate lake guards during surveillance shifts. To supplement our field data, we compiled the annual reports of arapaima management fisheries provided by the Association of Rural Producers from Carauari (ASPROC) produced in 2022. ASPROC is a grassroots smallholder and fisher-led organization leading the arapaima management along the Juruá River. We computed the surveillance expenses associated with all four scales of surveillance and subsequently compared costs under three different scenarios: 1) current expenditure covered by local communities or guards who were community members lacking any labour wage payments, 2) general expenditure and costs considering local daily wages of US$14.30 for two people working all year-round; 3) costs incurred by hiring two individuals receiving a minimum wage of US$442.24 (US$247.30 in wages plus US$194.90 in labour taxes) to conduct surveillance in compliance with Brazilian labour regulations, and 4) potential expenditure of US$852 (US$510.20 in wages plus US$341.80 in welfare taxes) covered by the Brazilian Environmental Agency for two additional environmental agents, according to the hiring notice SEI/ICMBio 15343964 and Law 7.957/1989. We also performed Generalized Linear Models (GLM) using a Gaussian distribution for continuous data to investigate the community-scale variation in protection costs (response variable) as a function of fluvial distance to the nearest town, number of lakes requiring protection, distance to the farthest lake, and the locally authorized harvest quota. We mitigated for collinearity between predictors using the Variance Inflation Factor (VIF < 3), excluding variables above this threshold (Zuur et al., 2010). We further combined all possible models, from the constant to the full model, using the dredge function of the MuMIn package. Models were selected based on the lowest Akaike information criterion (AIC) corrected for small sample sizes (AICc). The ΔAICc value represents the difference between the AICc of a given model and the lowest AICc, whereas ΔAICc < 2 represent the most likely set of parsimonious models (Burnham and Anderson, 2002). Finally, we applied a model-averaging approach, which represented the beta average of all predictors included in the set of most parsimonious models, and determined the relative importance of each explanatory variable given their model frequency and cumulative Akaike weight. All analyses were conducted in R 4.3.1. All monetary costs were standardized and corrected for inflation from October 2021 to July 2023 and converted into USD using a 4.91 BRL exchange rate. Lastly, we estimated the financial imperative of meeting the overall costs of territorial protection through a PES mechanism. In an attempt to estimate values that could cover the costs of territorial protection, we built three alternative scenarios (i) considering all operational costs, including fuel and food requirements and, at least two people hired through daily wages, (ii) operational costs and minimum wages following Brazilian labour regulations, and (iii) operational costs and human resources hired through the standard practices followed by the Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), Brazil’s environmental agency responsible for Protected Areas and environmental management. These compensation mechanisms provide valuable insights into ways of rewarding local dwellers for their role in territorial protection. However, it is essential to recognize that community protection of their environments is a collective effort. Therefore, involving local leaders in program design is crucial from the outset to identify the most effective ways of rewarding those engaged in territorial protection. Finally, we divided these values by the potential fish catch of each community to calculate monetary expenditure per unit of fish biomass harvested, which can facilitate the rational implementation of a PES program based on territorial protection and official catch statistics.