Systematic Errors in Length Frequency Data and their Effect on Age‐Structured Stock Assessment Models and Management

Abstract

AbstractLength frequency data are a central component of many statistical age‐structured models, which are used widely to conduct fisheries stock assessments and produce estimates of stock status for management. However, these data can be biased due to systematic errors during the sampling process. This study examined how bias in length frequency data affected stock status estimates from an age‐structured assessment model and subsequent prescribed management actions. A simulation approach was used to test the effect of two different scenarios of bias on assessment results. Populations were simulated based on life history parameters for three different serranid species in the southeastern United States and followed a trajectory that increased and then decreased over time. From these simulated populations, fishery‐dependent length frequency data were generated and then altered to represent different types of systematic sampling error. Data that were biased towards larger fish had a considerable effect on the age‐structured assessment model, producing more‐optimistic estimates of stock status, which tended to support management actions that would not adequately protect stocks from overfishing. This may be particularly problematic for longer‐lived species and for stocks that have been depleted well below their original population size. Stock status estimates were much less affected by data that contained a higher frequency of small fish, although this type of biased length frequency information could support unnecessarily stringent management actions in the most extreme cases. The results of this study suggest that error and possible bias in length frequency data should be thoroughly considered when selecting an appropriate modeling approach for fisheries stock assessment.