Evaluation of fisheries management strategies using a biotic integrity index in floodplain lakes in the lower Solimões River, Amazonas, Brazil

Fisheries is an important activity in the Amazon basin and potentially has a high impact on ichthyofauna. With the aim of achieving sustainability in fisheries of target species, protected areas were established within a fisheries management framework known as the Amazon Lakes Management System (ALMS). With the aim to evaluate the effect of ALMS on fish assemblages as a whole, we compared floodplain lakes with different levels of fish management in the lower Solimões River, in the Brazilian Amazon, developing an index of biotic integrity, named ALMS-IBI. We sampled fish in three lake categories: protected (PR), managed (MN), and commercially fished (CM). The ALMS-IBI was developed by selecting and testing fish assemblage metrics based on samplings carried out in 2017 and 2018, using gillnets with different mesh sizes. We captured 4565 fish of 113 species and seven trophic categories. The final index is composed of seven metrics related to species richness, trophic structure, and fish abundance, and characterized the fish assemblage in the PR and MN lakes as acceptable, and in the CM lakes as poor. Our results indicate that the ALMS-IBI can be an efficient method to monitor the whole fish assemblage in Amazonian floodplain lakes, and can be used as a complementary tool in the ALMS to assess environmental sustainability.


INTRODUCTION
Aquatic environments suffer many anthropic impacts (e.g., invasion of alien species, habitat destruction and fragmentation caused by hydromorphological modifications, eutrophication, climate change, fish overexploitation) that generate multiple pressures on the biota and on the environment as a whole (Poikane et al. 2017).The effects of environmental stressors can result in imbalance in ecosystem services of aquatic environments, such as habitat loss, species decline, drop in water level, and degradation of water quality (Bellwood et al. 2003;Hanna et al. 2018;Vári et al. 2021), challenging the stability and integrity of aquatic ecosystems (Fausch et al. 1990;Karr and Chu 1999).
One strategy used to protect aquatic ecosystems and their services (e.g., water quality control, habitat provision, erosion prevention, provision of fertile soils, food, flood regulation, microclimate regulation) is the establishment of protected areas (Saunders et al. 2002).Protected areas have effectively prevented habitat degradation and biodiversity decline, and have also aided the recovery of fish stocks (Schram et al. 1995;Reid et al. 2001;Suski and Cook 2007).
In Amazonian aquatic environments, anthropic impacts are mainly related to overexploitation of commercial fish species (Barthem and Goulding 2007;Castello et al. 2011), posing a threat to both commercial and non-commercial species, and to the integrity of the aquatic environment.The effects on fish include decreased density of commercial species, reduced fish stocks, and changes in assemblage structure, composition and functionality (Barthem and Goulding 2007;Silvano et al. 2009;Castello et al. 2011;Silvano et al. 2017).
In the Brazilian Amazon, protected areas have been established as a strategy to achieve sustainable fishing management, in a framework known as the Amazon Lakes Management System (ALMS) (Sistema de Manejo dos Lagos da Amazônia, in Portuguese) (McGrath et al. 1993;Castro and McGrath 2001;Benatti et al. 2003).The ALMS is specifically focused on floodplain lakes (Aquino et al. 2007) and was implemented in the 1980s by local fishermen and researchers who were concerned about the overexploitation of fish species.In general, the ALMS is based on the assignment of fishing restriction categories to limit the irrestricted access of fishers to floodplain lakes, thus establishing a more controlled and sustainable use of fishery resources and the environment (Nolan et al. 2009).It is a management instrument regulated by the federal environmental protection agency, Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis (IBAMA) and the Amazonas state agency for the environment, Secretaria do Estado do Meio Ambiente (SEMA) through fisheries agreements to promote sustainable fisheries by riverine and commercial fishers, and consists of defining the rules for the access to and use of local fishing resources and/or fishing gear and techniques, applicable in the state of Amazonas, Brazil (SDS 2011).
The ALMS defines lake categories based on levels of fishing restrictions in the floodplains of the Solimões-Amazonas River and its main Andean and Pre-Andean tributaries, and respects the existing laws (McGrath 1998;Castro and McGrath 2001;SDS 2011;Campos-Silva and Peres 2016).Another characteristic of the ALMS is the rotation of fishing access to lakes, which is based on the population dynamics of the target species, as a management strategy to improve ecosystem integrity in the floodplain lakes.
The success of this management model has been attested by the recovery of stocks of the main target species, such as Arapaima spp., in several areas of the Brazilian Amazon (Almeida et al. 2009;Campos-Silva and Peres 2016;Campos-Silva et al. 2017).For example, protected lakes and lakes with some degree of protection showed larger populations of the target species than unprotected lakes (Campos-Silva and Peres, 2016;Campos-Silva et al. 2017).
In addition to the main effect on a single target species or group of target species, the co-management of lakes through ALMS can promote the conservation of fish diversity as a whole in aquatic environments (Medeiros-Leal et al. 2021).For example, in the lakes of Paciência Island, in the region of the lower Solimões River, the implementation of ALMS had a positive effect on the composition and structure of fish assemblages, increasing fish abundance, biomass, fish size and species richness (Medeiros-Leal et al. 2021).
In this context, the aim of this study was to develop an ecological index of biotic integrity (IBI) for the lakes included in the ALMS by using metrics of richness, abundance and trophic category, according Petesse et al. (2016), to compare fish assemblages among lakes of different management categories.Since the development of IBI by Karr (1981), this tool has been used for the assessment of ecosystems exposed to distinct levels of environmental impacts (Carvalho et al. 2017;Carter et al. 2019), but few studies have used metrics of ecological integrity to assess the responses and effectiveness of management actions (Parrish et al. 2003;Tierney et al. 2009;Carter et al. 2019), and their application in sustainable fishing management programs is novel.
We evaluated fish richness metrics (e.g., total number of species, number of species with moderate/high vulnerability), abundance (e.g., total number of individuals, abundance of piranhas, equitability) and proportion of specialist trophic groups (e.g., percentage of piscivorous individuals), and selected those metrics that best represented the variance among lake categories.We propose to use this index of biotic integrity (IBI) as a complementary aspect to ALMS, to measure environmental quality for fish assemblages as a whole, and name it ALMS-IBI.We evaluated whether ALMS-IBI can adequately measure the effect of fisheries management ACTA AMAZONICA on Amazonian lakes, and hypothesized that ALMS-IBI scores are significantly higher in protected (no fishing) and managed lakes (restricted fishing) than in lakes where fishing is permitted without restriction.

Study area
The study was carried out in six floodplain lakes located on Paciência Island in the lower stretch of the Solimões River, in Amazonas state, Brazil (Figure 1).The islands of the Amazonian floodplains support complex systems of relatively small and shallow lakes located close to each other, and that are often connected during the high water season (Freitas et al. 2010).Two of the six study lakes were assigned to each of three management categories, as defined by ALMS: 1) Preserved lakes (PR): Cacau Lake (3°18'32.9"S,60°12'54.1"W)and Baixo Lake (3°18'09"S, 60°13'35.4"W).Fishing is prohibited in these lakes, to allow for the recovery of fish stocks; 2) Management lakes (MN): Sacambú Lake (3°18'46"S, 60°13'19"W) and Preto Lake (3°18'33.0"S,60°13'09.5"W).In these lakes, the fishing of Arapaima spp. is regulated by annually pre-defined quotas of fish for each fisherman or fishing sector; and 3) Commercial use lakes (CM): Piranha Lake (3°16'57.3"S,60°13'20.0"W)and Poção Lake (3°18'53.1"S,60°11'10.5"W).These lakes are open to unrestricted fisheries, including commercial fisheries.The distance between lakes varies from 0.5 to 6 km and the PR and MN lakes are closer to each other than to the CM lakes (Figure 1).
The lakes have been assigned to these categories in 2011, when ALMS was implemented on Paciência Island.The ALMS established that lakes rotate periodically in category assignment, but the rotation has not yet occurred on Paciência Island.The ALMS co-management protocol determines that PR lakes are located further away from the Solimões River channel in order to make access difficult for commercial fishers, and that MN lakes are located closer to fisher communities, to facilitate the communitary monitoring of fishing.In the MN lakes, commercial fishing is prohibited, but subsistence fishing by the community and annual fishing (determined by quotas) of the target species of fisheries management (Arapaima spp.) are allowed.

Fish sampling
Fish were collected in five sampling events during the hydrological periods of rising water (April), high water (June), receding water (August), and low water (December) in 2017, and rising water (April) in 2018, with a total of five sampling in each lake, totaling 30 sampling events overall.In each sampling event, we used ten gillnets, divided into two gillnets in each lake covering different points (aquatic herbaceous stands and open water).The gillnets had standardized dimensions of 15 m long and 2 m high, and mesh sizes varying from 30 to 120 ACTA AMAZONICA mm between opposite knots.The gillnets were placed in the water for eight hours, divided into two crepuscular phases: 5:00 until 9:00 a.m. and 5:00 until 9:00 p.m..The sampled fish were identified using reference keys (Soares et al. 2008).The sampling was conducted under license # 62867-1 granted by IBAMA through the Brazilian Biodiversity Authorization and Information System (SISBIO) and was authorized by the ethics committee in the use of animals of Universidade Federal do Amazonas (UFAM) (protocol # 037/2017 CEUA/ UFAM).The fishes were deposited in the ichthyological collection of Universidade Federal do Amazonas (UFAM), Manaus, Amazonas state, Brazil.

Environmental variables
We measured the following environmental variables in situ during fish sampling using a portable digital multiparameter (Hanna HI9829) to record abiotic environmental information: temperature (ºC), pH, dissolved oxygen (mg•L −1 ), depth (m), conductivity (s/m) and transparency (m).The measurements were made on the five sampling events, in the early morning and late afternoon, near the surface and bottom of the lake.

Reference lakes
In our study, the protected lakes (PR), Cacau and Baixo, were used as reference lakes for the biological integrity of the lake biota.Due to the absence of pristine sites in the region, these lakes were considered as reference condition (Whittier et al. 2007), i.e., the original state of biological integrity before any human disturbance.As these lakes are no-fishing lakes by the ALMS protocol, the fish assemblages are expected to exhibit natural cycles, unaffected by the fishing pressure elsewhere in the region.They are located in the interior of the island, far away from the main channel of the Solimões River.

Candidate metrics and selection procedure
Candidate metrics for the ALMS-IBI were organized in three groups: species richness, trophic categories and species abundance.We used the set of metrics validated by Petesse et al. (2016), excluding "number of non-native species", as there is no record of exotic fish species in the Solimões River.We added four metrics, namely, "number of Cichlidae species"; "abundance of piranhas" (Serrasalmidae), referring to Pygocentrus nattereri, Serrasalmus maculatus and S. altispinis; "abundance of branquinhas" (Curimatidae), referring to Psectogaster rutiloides, Potamorhina latior and P. altamazonica; and "abundance of sardines" (Triportheidae), referring to Triportheus albus, T. angulatus and T. auritus.These metrics were added because these species are the most representative (in abundance) of their trophic groups in the floodplain environment and because we would like to assess how fisheries impact the abundance of these species.In total, we used 23 candidate metrics (Supplementary Material, Table S1).
The trophic categories used to develop the ALMS-IBI were based on a literature survey (Mérona and Rankin-de-Mérona 2004;Santos et al. 2009;Soares et al. 2008) and on information from the FishBase website (Froese and Pauly 2018).The following trophic groups were considered: omnivores, herbivores, detritivores, insectivores, invertivores, planktivores and piscivores.The ecological indices of richness (S), abundance (N), Shannon diversity index (H'), equitability of Pielou (J') and dominance of Berger-Parker (d), as described by Magurran (2004), were estimated to compose complementary biotic data on the fish assemblages.The indices were estimated for each sampling event in each lake, i.e., five sampling events in each lake.
We used the dataset of the references lakes (Baixo and Cacau lakes) to score metrics and we randomly sub-sampled the dataset to validate the ALMS-IBI, according to the criterion used by Petesse et al. (2016).The metrics were submitted to three tests in sequence: 1) metric range test; 2) metric sensitivity test; and 3) metric redundancy test (Petesse et al. 2016).The range test evaluated the distribution of values for each metric and the difference between the maximum and minimum values of the range.When the difference was < 4, the metric had no variability and was excluded in the initial phase (Petesse et al. 2016).For metrics with sequences of equal values (e.g., sequence of zero values) the criterion of 75% of the values was used, which corresponds to the percentage calculation of equal values, according to the equation: n/N total x 100; where n = number of equal values in the metric; N total = total number of samples.Percentages over 75% meant that the metric had no variability and would probably not distinguish between reference sites and test sites, according to the criterion of Petesse et al. (2016).For the metrics "percentage of insectivorous individuals", "percentage of planktivorous individuals" and "abundance of branquinhas", the criterion of 75% of equal values was used.The "equitability" and "dominance" metrics were converted into percentages to apply the same criteria as to the other metrics.
The sensitivity test was carried out using box-plot graphical analysis to evaluated the ability of the metric to discriminate between reference lakes and test lakes by assessing the degree of overlap of quartiles and medians.The greater the overlap, the lower the sensitivity of the metric to distinguish between reference areas and impacted areas (Barbour et al. 1996;Hughes et al. 1998).The redundancy test evaluated the correlation between metrics, since highly correlated metrics do not respond to the assessment of environmental integrity (Seegert 2000).This analysis used Spearman's correlation coefficient (Petesse et al. 2016).

Procedure for scoring the metrics
For each metric selected for the reference lakes, the 65 th percentile was used as the upper threshold, and the 32 nd ACTA AMAZONICA percentile was used as the lower threshold (Petesse et al. 2007).These thresholds were chosen based on a conservative approach and assumed that some metric values from the reference lakes could have relatively low integrity (i.e., due to natural variability in the hydrologic cycle or to an anthropogenic effect that occurred in the past).A continuous score scale from zero (worst situation) to 10 (best situation) was used.Values above or equivalent to the 65 th percentile received a score of 10, representing metrics of high integrity, while values equal to or below the 32 nd percentile received a score of zero.For the values in between the percentiles, the following equation was used: (observed value -value of the 32 nd percentile)/(value of the 65 th percentile -value of the 32 nd percentile) x 10.For low-integrity metrics, the score was inverted, and the values between the percentiles received scores according to the following equation: (value of the 65 th percentile -observed value)/(value of the 65 th percentilevalue of the 32 nd percentile) x 10 ( Petesse et al. 2016).
To standardize the final index, the sum of the partial scores for each lake category was divided by the total number of metrics selected and multiplied by 10 in order to obtain the ALMS-IBI, which varied between 0 and 100.This interval was divided into four classes, where 0-25 (poor condition), i.e., human pressure is dominant and the negative effects are evident; 26-50 (regular condition), i.e., signs of human pressure are evident and negative effects begin to appear; 51-75 (good condition), i.e., signs of apparent human interference, but the use of resources is environmentally sustainable; and 76-100 (excellent condition), i.e., condition minimally impacted or without significant anthropogenic interference.

Statistical analysis
The non-parametric Kruskal-Wallis test was used to compare the ALMS-IBI among the four hydrological periods.Data refer to two samplings per lake category for the high, receding and low water periods, and four samplings per lake category for the rising water period.
Each ALMS-IBI candidate metric was compared among lake categories using the non-parametric Kruskal-Wallis test, using 10 samples per lake category, i.e., five samples for each lake.We also compared the final ALMS-IBI scores among lake categories, using the Kruskal-Wallis test, based on ten sampling values per category (five for each lake).The pairwise difference between lake categories was tested by Wilcoxon test with the adjustment method "hommel", a method considered robust and valid for independent samples.A significance level of 5% (p ≤ 0.05) was used in all tests.
Non-metric multidimensional scaling (NMDS) was applied based on Manhattan distance, using lake categories as objects and the metrics of the biotic integrity index as attributes.The analyses were performed in R software using the Vegan package (R Core Team 2020).The metrics were standardized using the Z scores method.

Structure of the fish assemblages
A total of 4,565 fish specimens were collected, distributed in six orders, 25 families and 113 species (Supplementary Material, Table S2).Characiformes was the order with the highest proportion of species in MN (88.4%) and PR (75.4%) lakes, while Siluriformes had the highest proportion in CM lakes (36.9%).Most species belonged to Curimatidae and Serrasalmidae in the MN (52.8%) and PR (43.3%) lakes, and to Triportheidae and Pimelodidae in the CM lakes (50.2%).The Sacambú Lake (MN) had the highest species richness estimates and abundance of individuals, while Preto Lake (MN) had the highest species diversity.The PR lakes had the second highest species diversity and highest equitability.The Shannon diversity, Pielou equitability and Berger-Parker dominance indices were similar among the lakes (Table 1).
In the CM lakes, omnivores made up 51.9% of the fish sampled overall.In the MN lakes, omnivores (39.6%) and detritivores (36.3%) were similarly predominant.In the PR lakes, omnivores, carnivores and detritivores were equally abundant (approximately 29% each) (Figure 2), but the highest richness was of omnivores (S = 20).

Environmental variables
Cacau Lake (PR) had a lower mean temperature (25.89 ºC) and higher mean transparency (0.91 m) (Table 2).Baixo Lake (PR) showed a mean temperature of 28.47 ºC and had the highest pH value (8.06).PR lakes showed highest transparency, highest pH and lowest temperatures.CM lakes had lowest pH, highest conductivity, lowest transparency and temperature above 29 °C.MN lakes, which are the deepest lakes (Table 2), had highest dissolved oxygen.

Metric selection for the ALMS-IBI
All 23 metrics showed sufficient variability in the range test (Supplementary Material, Table S3) and thus proceeded to the next test.According to the sensitivity test (Supplementary Material, Figure S1 and S2), only seven metrics differentiated among lake categories (number of Siluriformes species, number of Cichlidae, percentage of omnivorous individuals, percentage of piscivorous individuals, equitability, abundance of sardines and abundance of piranhas) and were used in the redundancy test (Supplementary Material, Table S4) and included in the ALMS-IBI (Supplementary Material, Table S5).
The ALMS-IBI presented higher scores for the PR and MN lakes.The metric "percentage of piscivorous individuals" scored high in the PR lakes (Baixo = 8.87; Cacau = 8.09) and the metric "abundance of piranhas" scored high in the PR lakes (Baixo = 10; Cacau = 8.98) and MN lakes (Sacambú = 10; Preto = 9.79).The metric "number of Cichlid species" achieved mixed scores in the PR lakes (Cacau = 10; Baixo = 4) and scored high in MN lakes (Sacambú = 10; Preto = 10).The metrics "Percentage of omnivorous individuals" and "Abundance of sardines" were higher in CM lakes and consequently presented zero scores.The PR and MN lakes were classified as having regular integrity, while the CM lakes were classified as having poor integrity (Figure 3).
The NMDS (stress value = 0.0210) separated the three lake categories in the multivariate space (Figure 4).The metrics "percentage of piscivorous individuals" and "abundance of piranhas" tended to be associated with PR lakes, while the metrics "abundance of sardines" and "number of Cichlidae" were associated with CM and MN lakes.Equitability was associated with MN lakes, and the metric "number of Siluriformes" was associated with MN and CM lakes.

DISCUSSION
Our results provide evidence that the ALMS comanagement rules improve biotic integrity at the fish assemblage level.The ALMS has been implemented Table 2. Limnological and physical-chemical parameters for six floodplain lakes sampled on Paciência Island (lower Solimões River, Amazonas, Brazil) between April 2017 and April 2018.Category (according to ALMS): PR = protected; MN = management; CM = commercial use.Values are the mean of five sampling events in each lake to bottom and surface.ACTA AMAZONICA throughout Amazonas state, notably so in two sustainable development reserves in the middle Solimões River region that combine the preservation of the environment and the sustainable exploitation of natural resources (RDS Piagaçu-Purus on the lower Purus River, and RDS Mamurauá and Amanã).In both reserves, the management of Arapaima spp. is well established and shows positive results in the recovery of populations (Amaral et al. 2011).A study conducted in the lower Solimões River (Medeiros-Leal et al. 2021) showed that fisheries management has a positive effect on the structure and composition of the fish assemblage as a whole, increasing fish abundance and richness in the whole lake ecosystem.Likewise, the environmental quality of Amazonian Forest streams in areas of reduced impact logging was higher than in unmanaged logging exploitations (Prudente et al. 2018).In the Pantanal biome, a multi-metric index obtained high values of integrity (e.g., good and excellent) in a protected area, confirming the effectiveness of the environmental protection (Polaz et al. 2017).Studies in other biomes also provided evidence for high biotic integrity in protected aquatic environments (Mancini et al. 2005;Reza et al. 2011;Sobczak et al. 2013;Fitzpatrick et al. 2015).In our study, the protected lakes (PR), where fishing is prohibited, may represent a reliable reference site for the assessment of fishing pressure on the biotic integrity of the ichthyofauna for this region, since they have better preserved fish assemblages in terms of species composition and trophic structure in comparison to what was found in the CM lakes.

Environmental variable
The ALMS-IBI revealed that piscivorous fish, particularly piranhas, were predominant in the PR lakes, supporting other reports on a relatively higher proportion of fish of higher trophic levels in protected environments (Paine 1966;Araújo 1998;Ganasan and Hughes 1998).In the study by Petesse et al. (2016), the lakes with excellent biotic integrity presented the highest percentages of predatory (i.e., carnivorous) fish.Carnivorous fish play a role in structuring the lower trophic levels (Jia et al. 2021), therefore changes in the proportion of fish belonging to high trophic levels can alter the fish assemblage to a higher dominance of lower trophic levels.Commercial fishing frequently focuses on large species of high trophic level (e.g., Arapaima spp.), which consequentially alters the fish assemblage (Jia et al. 2021) and explains the importance of this metric to assess biotic integrity in ALMS-IBI.In the case of piranhas, we attribute their higher abundance in the PR lakes relative to the CM lakes to the sedentary behavior of these species (Soares et al. 2008), which characterizes them as residents in the PR lakes, at a longer distance from the CM lakes and isolated in the low water period.In addition, the association of PR lakes with piscivorous fish was probably influenced by the tendency to higher transparency of the water column in these lakes, as had been observed in a previous study (Medeiros-Leal et al. 2021).Further studies should elucidate how much of the variation in limnological parameters in ALMS lakes is explained by the co-management as opposed to natural variability.
Species richness and abundance were lower in the PR lakes compared to the MN lakes, which may be associated with the greater abundance of predators in PR lakes, as these exert strong pressure on the assemblages (Both et al. 2009;Freitas et al. 2010), which could result in the lower richness in these lakes.In the presence of many predators, lateral displacement of prey may occur and result in decreased richness of fish, mainly in the high water period (Melo et al. 2007).The MN and PR lakes showed equitability in species abundance and higher Shannon diversity compared to the CM lakes, corroborating our results that PR and MN lakes have better biotic integrity.The proximity of the PR and MN lakes probably influences the structure of the fish assemblages in these lakes, as they were similar in diversity and evenness, which reflected in similar ALMS-IBI scores of regular biotic integrity.Similarity in fish assemblages of island lakes in the same region during the high water period was also observed by Freitas et al. (2010).
Regarding CM lakes, the lower biotic integrity scores for these lakes may be related to their environmental conditions (higher temperatures, less transparency and high conductivity), which are probably influenced by the influx of sediments and suspended material from the main channel of the Solimões River, to which these lakes are directly connected in the high water period.In addition, the current anthropic impacts in the CM lakes, such as pollution and habitat degradation by human presence and commercial fishing, probably also contributed to the lower ALMS-IBI scores for these lakes.

ACTA AMAZONICA
Changes in biotic integrity reflect the trophic structure of species (Bozzetti and Schulz 2004;Costa and Schulz 2010), and the dominance of one or a few species over the others (Casatti et al. 2009).A greater abundance and dominance of fish with omnivorous feeding habits and migratory behavior was recorded in the CM lakes, particularly sardines, which normally live in shoals (Ponte et al. 2016).

CONCLUSIONS
Defining metrics that represent the effects of fisheries (particularly commercial fisheries) in the Amazon has been a challenge.Our results suggest that the ALMS-IBI was successful in characterizing fish assemblages in lakes in the lower Solimões River using a few selected biotic parameters, and may be a useful complementary tool for the evaluation of biotic integrity in different categories of ALMS lakes.Future studies should consider samples from a wider geographic range and the relationship between biotic and abiotic metrics in different lake categories to improve the ALMS-IBI.Despite the limitations of our study, our results show the benefits of the ALMS regarding the evaluation of management strategies for the sustainable use and conservation of biodiversity.
Table S1.Candidate metrics for the development of an index of biological integrity for the Amazon Lakes Management System (ALMS-IBI).ER = metric response to disturbance: decrease (↓) or increase (↑) in presence of the environmental disturbance, according to Petesse et al. (2016) and study present.

Species richness
Total number of species ↓ Measure of diversity and generally decreases with environmental degradation.
Number of species with moderate/high vulnerability ↓ Assesses the importance of species with moderate and high vulnerability in relation to fishing, according to Fishbase (2018).

Number of orders ↓
Higher richness of orders is expected in an environment with good quality.

Number of families ↓
Higher richness of families is expected in an environment with good quality.

Number of Characiformes species ↑
Water column Characiforms are ecological equivalents of the "sunfish" group originally proposed by Karr (1981).Used by Petesse et al. (2016) to indicate poor quality conditions.Characiformes are widely distributed in tropical environments and are representative of the floodplain fish assemblage.

Number of Siluriformes species ↓
Siluriformes can be used as an ecological equivalent of the "darter" group used by Karr (1981).Assesses the conditions of the bentho-pelagic habitat and migratory fish.

Number of Perciformes species ↑
Used as the ecological equivalent of the "sunfish" group in tropical and subtropical environments (Hughes et al. 1999).Higher values represents environments with poor conditions.

Number of Cichlidae species ↓
Suggested by Hocutt et al. (1994) andHugueny et al. (1996) as indicators of margin conditions, since these fish use this environment for spawning and nest building (Hughes et al. 1999) Table S3.Metrics range test for the development of an index of biological integrity for the Amazon Lakes Management System (ALMS-IBI), according to Petesse et al. (2016) and the present study.

Figure 1 .
Figure 1.Location of the study area in northern Brazil, and of the studied lakes (red dot) on Paciência Island in the lower stretch of the Solimões River.The close up satelite image of Paciência Island shows the six sampled floodplain lakes.Green points = preserved lakes; blue points = managed lakes; red points = commercial fishing lakes.This figure is in color in the electronic version.

Table 1 .Figure 2 .
Figure 2. Relative abundance (%) of the trophic groups of fish sampled in six lakes on Paciência Island in the lower Solimões River (Brazil).Lakes are categorized according to the Amazon Lakes Management System (ALMS): PR = protected; MN = managed; CM = commercial use.Values refer to grouped data of five sampling events in each of two lakes per category.

Figure 3 .
Figure 3. Score distribution of the index of biotic integrity of the Amazon Lakes Management System (ALMS-IBI) for six lakes on Paciência Island in the lower Solimões River (Brazil).The central line in bold represents the median, the box the 25-75% quartiles, and the bar represents the range.Values correspond to five samplings in each of two lakes per category.

Figure 4 .
Figure 4. Non-metric multidimensional scaling (NMDS) showing the relation of lakes on Paciência Island in the lower Solimões River (Brazil) categorized according to the Amazon Lakes Management System (ALMS) with the metrics on fish assemblages selected for the index of biotic integrity (IBI).Green points = protected lakes; blue points = management lakes; red points = commercial fishing lakes.Stress-value = 0.0210.This figure is in color in the electronic version.

Table S2 .
. High numbers indicate good quality of the habitat.Number of individuals of each fish species sampled in three lakes of different categories according to the Amazon Lakes Management System (ALMS) on Paciência Island, lower Solimões River (Brazil).Data from five sampling events between April 2017 and April 2018.ALMS category: PR = preservation lakes; MN = management lakes; CM = commercial use lakes.