Fish-based Index of Biotic Integrity for wadeable streams from Atlantic Forest of south São Paulo State , Brazil

Understanding the relationship between environmental quality of streams and biological integrity of fish assemblages is critical to successful ecosystem management. Aim: We adapted the Index of Biotic Integrity (IBI) using ecological data of the fish assemblages that occur in headwater streams from the Atlantic Forest in southern São Paulo State. Methods: We sampled the ichthyofauna and collected environmental data in 27 streams stretches during the dry season of 2010. The fish species were categorized into trophic group, position in the water column and preference for rapid meso-habitats. Candidate metrics were screened for range, responsiveness and redundancy. Results: Of the 17 metrics tested, four metrics were included in the IBI. They belonged to attributes species diversity: percentage of individuals as Loricariidae family; habitat use: percentage of individuals as benthic riffles; and trophic function: percentage of individuals as omnivores and percentage of individuals as herbivores/detritivores. Eight streams (30%) were classified as excellent or good and fourteen (50%) as poor or very poor. Conclusions: On a regional scale, many aspects of biological integrity were altered but there are streams that can be used as biological reference.

. Streams at headwater regions are generally less degraded than downstream stretches of the river basins, also presenting smaller natural variations in physical, chemical and biological conditions.Due to these characteristics, these regions can be used as references sites in environmental assessment (Drake, 2004).
Although the natural variability in these systems can be considered low, when the results of anthropogenic modifications are considered, headwaters can be quite heterogeneous spatially, and characterized by diverse interacting patches ranging from human-dominated environments to conservation units.In the present study we developed an IBI based on stream fish communities that can provide potential implications for the management in headwater streams that flow through the Atlantic Forest in the south of the São Paulo state.

Introduction
In urban areas, studies point to dramatic changes in the quality of streams, primarily due to sewage emission (Pinto et al., 2006).In rural areas, these changes are generally related to a decline in riparian vegetation to provide area for cultivation and forestry and to the presence of plumbing and drain-laying work for construction of secondary roads (Ferreira & Casatti, 2006;Roth et al., 1996).
Several studies demonstrated a relationship between environmental quality of streams and biological integrity of fish assemblages.Since the Index of Biotic Integrity -IBI was proposed (Karr, 1981), environmental assessment systems based on biological community attributes have supplemented those based only on water quality.The IBI focus at the biological responses to physical or chemical characteristics of water bodies of the different organisms that make up the assemblages (Karr, 1981), by grouping multiple indicators of abundance, composition and functional organization patterns of these assemblages into a single index (Karr & Dudley, 1981).Thus, it can be used as an indicator for assessing and monitoring the quality of water bodies.
The integrity of the assemblages provides a direct measure of the ecological conditions of water resources (Angermeier & Davideanu, 2004) and, when compared to an intact reference system, these environments can be positioned along a continuum of environmental degradation.In São Paulo state several studies generated streams IBIs based in fish community (Marciano et al., 2004; 140 inhabitants.km - .The Alto Paranapanema river basin presents agricultural characteristics, a population density of about 30 inhabitants.km - with about 15% of native vegetation and headwaters covered by reforestation areas, mainly Eucaliptus and natural forests.In the hydrographic basins of the Ribeira de Iguape river about 60% of its territory has native vegetation and it has a population density of about 15 inhabitants.km - .

Streams classification
During the dry period in 2010 (July -November), we sampled the ichthyofauna and structural environmental data to represent the full range of habitat types and stream quality within each sub-drainage system.Twenty-seven sample sites distributed in the Sorocaba river basin (n=10), Alto Paranapanema river basin (n=8) and in the Ribeira de Iguape river basin (n=9) were sampled (Figure 1).The streams in this study measured 3.9 m ± 2.14 m in width and 28.7 cm ± 11.02 cm in depth (mean +/-sd) (Cetra et al., 2012).To characterize the reaches we used a physical habitat index (PHI) adapted from Barbour et al. (1999).We evaluated the stretches with four habitat parameters: epifaunal substrate/available cover, sediment deposition, frequency of riffles and bank stability (Table 1).The PHI range classification was: 0 to 20 (Poor), 21 to 40 (Marginal), 41 to 60 (Suboptimal) and 61 to 80 (Optimal).

Fish collections
Considering that seasonal fluctuation of the water level is one of the most important factors to influence the structure of fish assemblages (Rodríguez & Lewis, 1997), the sampling period for the ichthyofauna was the dry season.During this period, connections between the structure of the fish assemblage and the habitat structure are more robust, and the effect of temporal variation can be controlled (Willis et al., 2005).Sampling is also more efficient due to the smaller volume of water and consequent increase in the density of fish (Pease et al., 2012).
Sampling was performed with an electrofishing apparatus, between 8:00hs and 17:00hs without contention nets, at the upper and lower limits.The ichthyofauna was collected in 70-meter stretches, a distance that is sufficient for representing the range of available mesohabitats, i.e., a repeating sequence of riffle, pool and run.The apparatus consists of a transformer supplied by a generator (Toyama 2000W) connected to two dip nets, that in the water release a continuous current (2A).Two individuals handled the nets in a single downstream-upstream movement to capture the fish.

Data analysis
Based on the literature (Oyakawa et al., 2006;Casatti et al., 2012) species were categorized into trophic group, position in the water column and preference for rapid meso-habitats (Table 2).Seventeen metrics were considered to define the list of candidate metrics of the IBI (Table 3).Candidate metrics were screened for range, responsiveness, and redundancy.First, a principal component analysis (PCA) was used to detect the metrics with a low variance.We used a broken-stick model to decided which axes are important and representative.Metrics with a factor loading > 0.7 were rejected.Second, a Spearman's rank correlation coefficient (rs) significance (α < 0.05) was used to examine the responsiveness of the remaining candidate metrics discriminating the minimally and the most disturbed sites based in the PHI index.Third, Pearson's correlation coefficient was used to test redundancy.Pairs of the metrics with strong positive correlations (r > 0.75) were considered redundant.The redundant metrics were then selected basing on their responsiveness and the applicability to the study area (Casatti et al. 2009;Jia et al., 2013).
The metrics were trissected in the 75th and 25th percentiles.Values above the 75th were given score 5 and those below the 25th were given score 1; and intermediate values were given score 3.For the metrics negatively related to the PHI 5 and 1 were reversed.The IBI was composed from the sum of scores for each metric by stretch stream.

Results
Frequency of riffles and epifaunal substrate parameters varied from poor to optimal.About 50% of the streams has a sediment deposition classified as optimal.The streams bank stability was classified as poor, marginal and suboptimal.The PHI streams ranged from 12 (poor) to 66 (optimal).About 67% of the streams was classified as suboptimal and marginal (Table 4).
A total of 2,892 fish was caught.Of the 17 metrics tested, 7 metrics were first removed because of the low factor loading on PC1, PC2 and PC3.These axes were selected to analysis the range of metrics, because they Acta Limnologica Brasiliensia, 2016, vol.28, e-22 accounted for 70.16% of the total variation (35.18, 22.85, 12.14%, respectively), were higher than broken-stick model and most metrics had the highest values on them.From the remaining 10 metrics, 4 failed the responsiveness test due to P value > 0.05 (Table 5).We exclude two redundant metrics: (i) number of herbivores/detritivores species (Sherb) was considered redundant with proportion of herbivores/detritivores individuals (%Nherb) and; (ii) proportion of benthic riffles individuals (%Nrif ) was considered redundant with number of benthic riffles species (Srif ).Four metrics were finally included in the IBI.They belonged to attributes species diversity (%Nlor), habitat use (%Nrif ) and trophic function (%Noniv and %Nherb) (Table 6).Eight streams (30%) was classified as excellent or good and fourteen (50%) as poor or very poor (Table 7), indicating that, on a regional scale, many aspects of biological integrity are altered but there are streams that can be used as biological reference.Barbour et al., 1999).nec -nektonic; bent -benthonic, trophic group (oni -omnivorous; inv -invertivorous; car -carnivorous; herb-det -herbivorous/detritivorous); riffles (R) and river basin (P -Paranapanema; S -Sorocaba; RI -Ribeira de Iguape.
Table 3. Candidate metrics for the Index of Biotic Integrity and the expected response to an increase in environmental degradation.

Discussion
The current Index of Biotic Integrity (IBI) used the stream fish community and consisted of four metrics that reflected, in a satisfactory manner, the environmental gradient found in wadeable headwater streams, which generally show little environmental change.
The four selected metrics for the IBI reflected the structure of the assemblages mainly in terms of proportional species abundance of the attributes species diversity, habitat use and trophic function.
In the present study, surprisingly, we did not have the expected response of species richness that has been extensively used to infer the quality of ecological systems (Roth et al., 2000).Perhaps the species richness variation is best explained by stream position in the longitudinal gradient than in the present disturbance gradient.
The predominance of individuals as Loricariidae family and herbivores/detritivores in preserved streams is expected in streams with shading because the presence of riparian vegetation increasing primary productivity, especially Periphyton that can be used as a food resource by these organisms.
In degraded conditions, the intense sediment supply increase the unstable substrate, preventing the accumulation or attachment of periphyton and thus the individuals with such food habits become rare in the impacted environment (Ferreira & Casatti, 2006).Our explanation would be linked to the bank stability rather than shading, because the streams with poor classification have low scores on the parameter bank stability.
The relative contribution of individuals as benthic riffles is high in streams with low sediment deposition with little or no enlargement of islands or point bars and less than 5% of the bottom affected by sediment deposition.With this metric we were able to discriminate excellent from poor or very poor biotic integrity in the streams.Benthic rifles species lived on or near the bottom and usually did not feed on the surface (Noble et al., 2007), and were sensitive to the stream siltation (Oberdoff & Hughes, 1992) Alterations in habitat quality, including bank stability, commonly result in changing availabilities of food resources (Karr, 1981).The omnivores dominated streams with biotic integrity very poor.In tropical and sub-tropical rivers and streams, the omnivore food category is predominant (Esteves & Aranha, 1999;Castro et al., 2003;Pinto et al., 2006), although the same behavior is expected where the increase in the degree of omnivory reflects the destructuring of aquatic habitats.

Conclusion
The current adaptation of the IBI is a first step in the establishment of an assessment protocol and the monitoring of Atlantic Forest streams in the headstreams of the Alto Paranapanema, Ribeira de Iguape and Sorocaba river basins.The survey of the four metrics that integrate the IBI and their scores should be sufficient to accompany the environmental changes in these systems.Thus, given the great utility of these results in the biomonitoring of headwater streams in the region, we have confidence in transmitting this information to the decision makers so that sophisticated statistical techniques will no longer be necessary for the adequate accompaniment of anthropogenic changes to these systems.

Figure 1 .
Figure 1.Limits of the Sorocaba, Paranapanema and Ribeira de Iguape river basins and streams sampled.

Table 1 .
Habitat parameters, condition categories and scores of the sampled streams from Atlantic Forest of south São Paulo State (adapted from

Table 2 .
Classification of sampled species according to attributes related to position in the water column.

Table 4 .
Number of streams by PHI parameter classification.

Table 6 .
IBI metrics and scoring criteria for Atlantic Forest of south São Paulo State streams.

Table 7 .
Detailed descriptions of stream biotic integrity, IBI values and number of streams by category (n).