Adding smoothing splines to the SAM model improves stock assessment

TL;DR

This paper introduces smoothing spline functions into the SAM stock assessment model to produce smoother, more automatic, and less subjective age-dependent parameter estimates, improving model performance and diagnostics.

Contribution

It develops and compares spline-based SAM models with existing models, demonstrating improved accuracy and usability across multiple fish stocks.

Findings

Spline models outperform existing SAM models in validation tests

Spline models produce smoother, more realistic parameter estimates

Spline models serve as effective diagnostic tools for model improvement

Abstract

The stock assessment model SAM contains a large number of age-dependent parameters that must be manually grouped together to obtain robust inference. This can make the model selection process slow, non-extensive and highly subjective, while producing unrealistic looking parameter estimates with discrete jumps. We propose to model age-dependent SAM parameters using smoothing spline functions. This can lead to more smooth parameter estimates, while speeding up and making the model selection process more automatic and less subjective. We develop different spline models and compare them with already existing SAM models for a selection of 17 different fish stocks, using cross- and forward-validation methods. The results show that our automated spline models overall outcompete the officially developed SAM models. We also demonstrate how the developed spline models can be employed as a diagnostics tool for improving and better understanding properties of the officially developed SAM models.