Predicting Peruvian Anchovy School Density with GAMLSS and BAMLSS Using Morphological Descriptors Extracted from Echograms

Huaman O.M.F.; Zamudio M.B.S.; Nieto G.P.C.; Rojas C.J.P. · Ocean Science Journal (2025) · DOI: 10.1007/s12601-025-00211-3

Resumen

The objective of this research project was to obtain a statistical model to estimate the density of Peruvian anchovy (Engraulis ringens) school based on their morphological descriptors, extracted from SIMRAD EK60 scientific echo sounder data at the frequency of 120 kHz, using the algorithms of the 2D detection module of the Echoview programme, including: height, length, mean depth, volume3D, distance to the coast and geographical position. The statistical tools GAMLSS and BAMLSS were used for the predictive analysis. The Box–Cox power exponential (BCPE) distribution appropriately describes the variation in the density of anchovy schools and predictors. The patterns in the worm and cube diagrams show 95% of the model residuals within the approximate 95% confidence interval bands. In GAMLSS and BAMLSS, the residuals Q-Q plots are distributed on the straight line and the histograms are fitted to a normal density function. The models fitted to the training data allowed estimation of the parameters μ, σ, ν and τ; the average and standard deviation for the prediction of the parameter μ (m2/mn2) with GAMLSS and BAMLSS were 2966.9 ± 4260.2 and 2907.8 ± 4229.4, respectively. The prediction in both models was optimal, with MAPE value of 14.18 and 14.4%, respectively. Kruskal–Wallis and Wilcoxon comparison tests detected no significant differences between the medians of μ prediction between GAMLSS–BAMLSS, GAMLSS–test data and BAMLSS–test data. The application of the model makes it possible to quantify the density of anchovy schools accurately, allowing scientists and fishery managers to take advantage of the information obtained with acoustic systems for commercial fisheries. © The Author(s), under exclusive licence to Korea Institute of Ocean Science & Technology (KIOST) and the Korean Society of Oceanography (KSO) and Springer Nature B.V. 2025.