Predicting the habitat suitability of freshwater shrimp in Anzali wetland watershed using support vector machine model

The prediction of the habitat suitability of freshwater shrimp (Macrobrachium nipponense) is an important issue for managing the aquatic ecosystems. Six sampling sites were selected from different parts of the catchment area of ​​Anzali wetland to predict the habitat conditions of freshwater shrimp in which the species was present in three sampling sites (36 instances) and it was absent in the remaining of the sampling sites (36 instances). In total, fifteen environmental factors consisting of water quality, dynamic and structural variables were monthly sampled based on the biological variable (the presence/absence of shrimp) over one-year sampling period (1397-1398). The prediction of SVM was reliable based on the two predictive performances (CCI% and Cohen kappa) since both predictive performances exceeded the threshold values (CCI%> 70% and Cohen kappa>0.70). The results of the support vector machine model showed that the population decline or even the absence of shrimp might have a very close relation with increasing water flow velocity, water depth, salinity, electric conductivity, sodium and chloride concentrations. Based on the model outcomes, increasing the concentration of other water quality variables including potassium, sulfate and total hardness might have an intermediate impact and increasing the concentration of orthophosphate, nitrate and biological oxygen demand might have less impact on the prediction of the shrimp habitat. Contrary to the above-mentioned variables, the results of the applied model showed that increasing the concentration of some variables such as dissolved oxygen, acidity and water turbidity might increase the probability of the shrimp presence in the sampling locations. Based on the factors predicted by the model, the importance of these factors should be considered for freshwater shrimp in future monitoring.