|Document Type:||Journal Article|
|Title:||Assessing the quality of life history information in publicly available databases|
|Author:||James T. Thorson, Jason Cope, W. S. Patrick|
|Keywords:||life history parameter,Bayesian,error-in-variables,global database|
Single-species life history parameters are central to the research and management of environmental systems, and are increasingly stored in centralized and global databases. However, there has been little independent evaluation of the precision and accuracy of the values in these databases. We therefore develop a Bayesian error-in-variables model that compares database entries with estimates from local experts, and illustrate this process by comparing entries in FishBase, a global fish database, with life history estimates compiled by local experts through the U.S. We do not assume that local experts have perfect information, but instead that they have obtained unbiased parameter estimates on average. This model distinguishes biases among seven life history parameters, two types of information available in FishBase (i.e. published values, or those estimated from other parameters) and two taxa (i.e. bony vs. cartilaginous fishes) while accounting for additional variance caused by sex- and region-specific life history traits. For published values in FishBase, the model identifies a small positive bias in natural mortality and negative bias in life span, perhaps caused by unacknowledged fishing mortality. For values calculating from other parameters, the model identifies large and divergent biases. It also demonstrates greatest precision for body size parameters, decreased precision for values derived from geographically distant populations, and greatest between-sex differences in age at maturity. We recommend that parameter estimates be used in future error-in-variables models as a prior on measurement errors. We hope that this independent evaluation will be replicated for other global databases, and that it will encourage further development and improvements in FishBase, which remains an indispensible tool for marine research.
We compared stock assessment life history data previously compiled for NMFS Productity/Susceptibility Analyses with data in FishBase, a publically accessible life history database. We used a Bayesian error-in-variables model to estimate accuracy and precision of FishBase data, and FishBase estimates of missing data using their life history tool. We show that FishBase data are generally unbiased, although they have small biases attributable to unacknowledged fishing mortality. We also show that FishBase life history estimates are often unbiased. We hope that these results will lead to improvements in FishBase, and increased usage in stock assessment.
|Full Text URL:||http://www.esajournals.org/doi/abs/10.1890/12-1855.1|
|Theme:||Recovery, Rebuilding and Sustainability of Marine and Anadromous Species|
Characterize vital rates and other demographic parameters for key species, and develop and improve methods for predicting risk and viability/sustainability from population dynamics and demographic information.