Discussion
A conspicuous absence of stationarity between fish species richness and the tested hypotheses was determined, but climate was observed to contribute the most to the richness distribution of stream fishes. The water-energy dynamics were the most probable metabolic restriction mechanism acting on the community structuring of stream fishes. Regarding species richness distribution at a macro scale, two characteristics must be considered: i) spatial data autocorrelation and ii) stationarity in fish richness-macroecological variables relationship. Autocorrelation of data modifies both the relationship and the significance of the relationship between the variable of interest and the predictor(s)(BINI et al. 2009). In the present study, the spatial autocorrelation was controlled when the GWR radius that had a Moran’s I index close to zero was chosen, isolating the second characteristic (stationarity) and facilitating its analysis.
The absence of stationarity relationship found can be derived from environmental heterogeneity, usually associated to altitude variation that causes climatic anomalies and modifications of local conditions(Kerr and Packer 1997; O’Brien et al. 2000; Rahbek and Graves 2001). Depending on the location and altitude variation, this landscape heterogeneity can increase or decrease diversity. The presence of mountain ranges, like Serra do Mar (southwest) in this study, results in an increased humidity on its windward side and the formation of drier and warmer areas on its leeward side, since the wind and humidity are blocked by the windward side. Local variation of temperature, precipitation and wind regimen influence on microclimate, which consequently change habitat availability and quality. As a result, sites favored by the windward effect can display greater species richness whereas those under the leeward influence could show less species richness, as predicted for all models. The topography gradient was observed as a functional factor structuring fish assemblages in streams of the Tocantins-Araguaia basin, (lower altitudes in the Araguaia (lower) and elevated ones in the Tocantins basin(Carvalho and Tejerina-Garro 2015)). Additionally, geographic heterogeneity increases geographic area(O’Brien, Field, and Whittaker 2000) and allows events of allopatric speciation to occur by interrupting geneflow between populations due to physical discontinuities in the riverbed (waterfalls and dams) or physiochemical changes (pH, temperature(Rahbek and Graves 2001). An increase of available area favors more individuals and species that can occupy a region. In regions with high elevation, such as Serra do Mar (2,366 m) and Espinhaço (2,072 m), it is common to observe low temperatures and, in lower elevations (near the ocean), higher temperatures. This thermal difference can make species diversity smaller than what was predicted for the region, due to local extinction of species less tolerant to cold weather(Girard et al. 2015; Mas-Martí et al. 2014). This mechanism could occur in regions that have an elevated altitude, such as observed in the Brazilian central region in this study. On the other hand, geographic heterogeneity (quantified by the topography) can create more complex habitats, and allow the coexistence of more species, than regions with not so conspicuous elevations(Bickford and Laffan 2006).
Regarding the macroecological variables, climate was observed to be what most influences richness distribution of stream fishes. Sixty percent of the diversity gradients had their observed patterns explained by climatic factors, some of them with R-squared close to 90%(Hawkins et al. 2003). In this case, the most important factors for determining species richness are water availability and energy input (Hawkins et al. 2003). The non-stationarity in the relationship between richness and climate was also studied by Hawkins et al.(Hawkins et al. 2003), who observed temperature as more important in high latitudes (colder places) than in low ones (tropical regions). In this study, the variation of temperature was observed as the factor of greatest influence on fish richness, presenting positive relationships in the Brazilian central region and negative in the Amazonian one. The non-stationarity of the relationship between stream fish richness and temperature oscillation can be explained by the climatic heterogeneity of the study area and the climate influence on the taxonomic diversification of the fish. Fish populations found in the Brazilian central region are inserted in a savannah landscape characterized by a tropical climate, with a well-defined dry season and rainfall concentrated in only one period of the year(Marengo and Valverde 2007). This climate type is characterized by seasons with 250 mm or <10 mm of precipitation per month and soil temperature varying between 20 and 40°C(N. B. F. dos Santos, Júnior, and Ferreira 2011). Fishes from the Amazonian region are located in areas with equatorial climate, where annual precipitation is 2,000 mm distributed equally throughout the months of the year, presenting an average soil temperature of 27°C varying less than 3°C(S. R. Q. dos Santos et al. 2011). Therefore, fish populations present in savannah areas predominant in the Brazilian central region sampling sites are exposed to a greater range of temperature variation, thus eliminating the occurrence of species with a small thermal range. On the other hand, in Amazonian areas, where the thermal variation is lower, tolerance to changes in temperature should not be a key factor in species selection. This could explain why we observed both tolerant and intolerant species to temperature variation in this region. Consequently, a negative correlation pattern is observed between temperature variation and stream fish richness.
The non-stationary relationship between temperature oscillation and fish richness found in this paper was also observed in snakes (Elapidae) and attributed to historical factors of the group’s recent diversification(Braga et al. 2014). The influence of temperature (as well as precipitation) driven diversification in recent taxonomic groups and favoring diversity gradients has ample acceptance in recent literature (Hawkins and Porter 2003b; Rodríguez et al. 2005; e.g.: Hawkins et al. 2003) Two mechanisms, the trophic cascade (greater amount of energy available in the system results in an increase of primary productivity) and the metabolic requirements (different species with different temperature tolerances) are proposed to explain the influence of temperature over the richness gradient(Hawkins et al. 2003).
The results presented here, suggest the relation between the stream fish sampled, the trophic cascade and the metabolic requirements mechanisms. The annual estimate evapotranspiration (AET) in June, which represents the measurement of energy input to the system, is the variable with the second greatest magnitude in determining the observed richness pattern. This variable had a negative relationship with stream fish in the Amazonian region and a positive one in the Brazilian Central region, therefore supporting the idea of physiological restriction. This result strengthens the hypothesis that Amazonian fish have low tolerance to thermal variation, the inverse occurring in fish from the central regions. Additionally, terrestrial primary production predicted fish richness, suggesting the influence of the trophic cascade mechanism. High terrestrial primary productivity is associated with areas that have dense vegetation coverage (England and Rosemond 2004). Forested riparian zones make available large inputs of leaves and terrestrial insects to the instream environment(Meyer et al. 2007), as is the case with the sampled streams (1st and 3rd order). The input of resources from terrestrial vegetation occurs in two ways; i) vertically – leaves, fruits, seeds and plant parts directly falling into the streams; and ii) horizontally – lixiviation of these resources from adjacent areas into the waterbody during the rainy season and/or pulses of inundation(Junk W., Bayley E.P. 1989; Junk and Wantzen 2004). With the entry of allochthonous resources, there is an increased resource availability for primary consumers, thus supporting a richer and more abundant food web.
A particularity of terrestrial primary production observed in this study is its negative effect on fish richness. It suggests that the metabolic restriction mechanism is more important than the trophic cascade mechanism. Organisms in this region, including aquatic ones, are exposed to a greater thermal amplitude(Marengo and Valverde 2007), which, together with increased terrestrial primary productivity, limits species richness. This effect possibly occurs due to increased surface shading of the streams’ main channel caused by dense riparian vegetation, since a greater primary productivity is related to areas with denser vegetation(England and Rosemond 2004). The dense vegetation stabilizes local microclimate(Monadjem and Reside 2008; Vieira, Dias-Silva, and Pacífico 2013) reducing climatic heterogeneity (cold water) and consequently species richness, possibly due to local extinction of fishes that had a higher optimum temperature.
The Water-Energy hypothesis is the main predictor of species richness considering the physiological mechanism². This hypothesis predicts a positive relationship between species richness and water quantity in lower latitudes and energy in places of higher latitudes². This relationship was observed in the present study for fish richness in Brazilian streams, where portions located close to the equator (Amazonian region) had a positive relationship to water quantity (average annual precipitation) and negative to energy input (AET in June). Portions with higher latitudes (Brazilian Central region) had a positive relationship with fish richness, while water had no relationship. The tradeoff between water and energy and diversity seems to be more dependent on water scarcity than energy restriction. Hawkins et al.² found that annual precipitation is the variable that determines diversity patterns in birds from the Australian continent, challenging what was expected by the literature, since the region is in an area of high latitude. In another study, Kessler(Kessler 2001) found that pteridophyte richness was a function of precipitation. This relationship was observed in Andean regions, where the expected would be energy (temperature, AET) acting as the limiting factor². These two relationships demonstrate that geographic and climatic heterogeneity generate non-stationary relationships, supporting the hypothesis stated in this paper, that is, the metabolic mechanism acts in a more deterministic way than the food web mechanism, although both are not mutually exclusive².
In conclusion, the diversity pattern of fishes in streams is a function of climatic variables and terrestrial primary production, where both the Water-Energy dynamic and metabolic restriction mechanism are more evident. The metabolic restriction mechanism divides Brazil in two regions: i) Amazonian, with a more stable climate and populations with low tolerance to thermal variation; and ii) Central, with greater temperature amplitude and populations more resistant to thermal variation.