Distribuição espacial de cianobactérias em reservatórios tropicais: Identificando as variáveis determinantes e o papel da autocorrelação espacial

Abstract

Little is known about the spatial autocorrelation (ACE) and pattern of distribution of cyanobacteria in tropical reservoirs when considering the effect of space based on environmental variables. Spatial autocorrelation (ACE) measures the relationship between biotic and abiotic variables considering the spatial proximity in which spatially close observations have more similar values than distant ones. The present study sought to evaluate the spatial distribution of cyanobacteria and identify predictive variables in tropical reservoirs, using spatial regression models to find out whether the incorporation of spatial autocorrelation (ACE) affects the estimates of model coefficients and the inference of statistical models. We sampled the phytoplankton community in thirty-seven tropical reservoirs distributed in different river basins in the state of Paraíba/Northeast, Brazil, during February 2020 to December 2021, with a quarterly frequency in the subsurface of the water column. We evaluated the ACE through the Global and Local Moran Index and then performed spatial and non-spatial regression models to determine the explanatory variables of the distribution of cyanobacterial biomass. Our results showed that out of the 37 reservoirs we evaluated, nineteen (19) showed dominance of cyanobacteria, that is, they showed cyanobacteria biomass greater than 50% of the total biomass of the phytoplankton community. The cyanobacterial biomass showed ACE, especially when considering the watershed level; although this effect varies according to the hydrographic basin studied. Cyanobacteria biomass within the Paraíba river basin showed significant ACE (I = 0.404; p = 0.008) with a high-high and low-low spatial distribution pattern, while the spatial effects within the state of Paraíba where we have the clusters from all basins were not significant. The results of the regressions showed that the determining variables for the cyanobacterial biomass were precipitation, water volume, temperature, dissolved oxygen, total phosphorus, ammoniacal nitrogen and total alkalinity. We observed that the values were higher in the spatial lag and error regressions and lower AIC values, in addition to showing a greater number of predictive variables when compared to the classical regression models. Therefore, the incorporation of spatial autocorrelation analyzes can help define which locations can be prioritized for conservation and management strategies.