Modelling the presence/absence of Samii's riffle minnow (Alburnoides samiii) in river

Abstract

Determining the most important environmental variables influencing the habitat of fish species is a very important issue to achieve an effective management of river ecosystem. The present research aimed to implement a data-driven model (classification tree; CT) to predict the presence and absence of Samii's spirlin (Alburnoides samiii) in a river basin system. 96 real observations together with various water quality and physical-structural variables were monitored monthly for one-year study period (2017–2018) in the Sefidroud River, northern Iran. The fish's presence/absence was successfully predicted by the CT model leading to a high model reliability (CCI> 75 % and Cohen Kappa> 0.60). The outcomes of the CT model at high levels of tree pruning (PCFs = 0.01 and 0.10, and with five-time randomization effort) confirmed that an increase in the water turbidity, biological oxygen demand, electric conductivity, river depth and orthophosphate may restrict the fish's presence, while the probability of fish's presence in the river may show an increase at higher slope and flow velocity. A significant difference (Kruskal-Wallis's test) was found between all predictors decided by the CT model and the samples with fish and without fish (p < 0.05 for the selected variables). The fish's abundance significantly differed with the sampling seasons/sites (Pearson Chi-Square= 40.6; p < 0.01). The generalized linear model (GLM) also indicated that the probability of the fish's occurrence may be diminished with increasing water depth and nutrient pollution (e.g., orthophosphate) in the river. The outcomes of the CT model, thus, can be applied for predicting the presence/absence of A. samiii and similar fish species in other ecoregions.