Predicting fish distribution in a mesotrophic lake by hydroacoustic survey and artificial neural networks. Limnology and Oceanography 44:1293–1303

The present work describes the development and validation of Artificial Neural Networks (ANN) by comparison with classical and more advanced parametric and nonparametric statistical modeling methods such as Multiple Regression (MR), Generalized Additive Models (GAM), and Alternating Conditional Expectations (ACE) to esti-mate spatial distribution of fish in a mesotrophic reservoir. The modeling approaches were developed and tested using 60 hydroacoustic transects covering the whole lake. Each transect was divided into 100-m-long sections, constituting a total of 732 sampling units. For each of them, the relationships between topology, chemical charac-teristics, and fish abundance were studied. The models had six independent topological (i.e., depth, distance from the bank, slope, and stratum) and chemical (i.e., temperature and dissolved oxygen) variables and one dependent output variable (fish density, FD). The data matrix was divided into two parts. The first contained units where FD was nonnil (i.e., 399 sampling units), and the second contained only cases without fish (i.e., 333 sampling units). Model training and testing procedures were run on the first submatrix after log(FD 1 1) transformation. As linear MR results were not satisfactory (r2 5 0.42 in the training set, and r2 5 0.51 in the testing set) compared with ANN (r2 5 0.81 in the training set, and r2 5 0.77 in the testing set), we tried nonlinear transformations of the variables such as logarithmic, lowess (for the GAM), and an optimal nonlinear transformation using the SA