How reliable are the Powell-Wetherall plot method and the maximum-length approach? Implications for length-based studies of growth and mortality

TL;DR

This study critically evaluates two common length-based methods in population studies, revealing significant biases and limitations, and emphasizes the need for new approaches to improve reliability in growth and mortality estimations.

Contribution

It provides a comprehensive assessment of the biases in Powell-Wetherall and Lmax methods, highlighting their limitations and proposing the necessity for developing new, more reliable techniques.

Findings

Powell-Wetherall estimates of Z/K are highly biased and unreliable.

Linf estimates are sensitive to intra-cohort variability and sampling issues.

Lmax cannot be reliably used to estimate Linf due to lack of a consistent relationship.

Abstract

Length-based methods are the cornerstone of many population studies and stock assessments. This study tested two widely used methods: the Powell-Wetherall (P-W) plot and the Lmax approach, i.e., estimating Linf directly from Lmax. In most simulations, P-W estimates of the ratio total mortality / growth (Z/K ratio) were biased beyond acceptable limits (bias > 30%). Bias in Z/K showed a complex behavior, without possible corrections. Estimates of asymptotic length (Linf) were less biased than Z/K, but were very sensible to intra-cohort variability in growth and to changes in the occurrence of large individuals in the sample. Exclusion of the largest size classes during the regression procedure or weighing by abundance does not solve these issues. Perfect linearization of the data and absurdly narrow confidence intervals for Z/K will lead users to erroneous overconfidence in outputs. Clearly, the P-W method is not suitable for the assessment of Z/K ratios of natural populations. Estimation of Linf may be tentatively possible under very specific conditions, with necessary external verifications. Also, this study demonstrates that there is no way to estimate Linf directly from Lmax, since there is no particular relationship to expect a priori between Linf and Lmax. Errors in estimating Linf directly affect the estimate of the growth constant K and all other subsequent calculations in population studies, stock assessments and ecosystem models. New approaches are urgently needed for length-based studies of body growth (e.g., unconstrained curve fit with subsequent bootstrapping), that consider the inherent uncertainty regarding the underlying data and processes.