Growth parameters with traditional and artificial neural networks methods of big-scale sand smelt (Atherina boyeri Risso, 1810)

Abstract

In this study, the growth parameters of big-scale sand smelt (Atherina boyeri Risso, 1810) in İznik Lake has been determined with traditional (length weight relationships (LWRs), von Bertalanffy (VB), condition factor (CF)) and modern approaches (Artificial Neural Networks - ANNs). A total of 635 specimens (44.84% female and 55.16% male) were collected from the local fisherman during the hunting season between April 2018 to April 2019. Mean fork length (FL) (mm, min-max), mean W (g, min-max) and mean CF (value, min-max) were estimated as 67.31 mm (40.10 - 97.77 mm), 2.57g (0.53 - 7.50 g), and 0.790 (0.170-1.520) for all individuals. The length-weight relationships were determined W=0.00001437L2.8602 for female, W=0.00001570L2.8266 for male and W=0.00001328L2.8717 for all individuals. The von Bertalanffy equations were determined Lt=136.218 [1-e(-0.240(t+0.51))] for female, Lt=155.042 [1-e(-0.185(t+0.73))] for male, and Lt=146.916 [1-e(-0.205(t+0.64))] for all individuals. The values in training (MSE (Mean Squared Error) 4.52559e-5, R (correlation coefficients) 9.09347e-1), verification (MSE 4.86111e-5, R 9.00931e-1) and test data (MSE 3.391999e-5, R 9.43465e-1) were found in calculations made with ANNs. It was determined that ANNs could be an alternative for evaluating growth estimation.