Development Of Generalized Additive Models (Gams) For Salmo Rizeensis Endemic To North-Eastern Streams Of Turkey

Generalized Additive Models (GAMs) are widely used statistical models for species distribution in habitat and environmental management modeling as they enable incorporation of non-linearity. The objective of this study is to develop relationships between habitat variables, fish absence/presence (logistic GAM) and abundance (Poisson GAM) for the S. rizeensis using on-site observations in Solakli River. Logistic GAMs correctly predicted absence/presence of adult and juvenile S. rizeensis at 79.3% and 74% of sampling areas respectively. Response curves of logistic GAMs results show differences in probability of finding the adult and juvenile fish. Adult S. rizeensis were mostly found in higher velocities (0.5-0.8 m/s) than in juvenile fish (approximate to 0 m/s). Also, the highest number of adult fish were recorded in deeper habitats (approximate to 0.4 m) rather than in juvenile fish (approximate to 0 m). Curves show adult fish are more independent to the presence of cover compared to juvenile fish. Velocity shelter and bedrock formation were the most common cover types. Chi-square test results of predicted values showed that developed Poisson GAMs of adult and juvenile S. rizeensis could not accurately represent fish abundance. The results show that while logistic GAM is applicable, Poisson GAM model is not applicable for the area.WoSScopu