Circularity in fisheries data weakens real world prediction

Area of Science:

• Fisheries science
• Ecological modeling
• Stock assessment

Background:

• Synthesized data from models is a cost-effective alternative to direct data collection in fisheries stock assessments.
• The impact of using synthesized data on predictive skill, particularly for recruitment forecasting, is not fully understood.

Purpose of the Study:

• To investigate how using synthesized data in stock assessments affects the accuracy of recruitment predictions.
• To compare the predictive performance of Standard Fisheries Models (SFMs) and an equation-free approach using both synthesized and observational data.

Main Methods:

• Utilized a global database of stock assessments.
• Compared the ability of SFMs to predict synthesized and observational recruitment data.
• Evaluated an equation-free approach against SFMs for predicting both data types.

Main Results:

• SFMs accurately predict synthesized data but show lower skill in predicting raw or minimally filtered observational data.
• The equation-free approach outperforms SFMs in predicting synthesized data.
• The equation-free approach demonstrates strong predictive capability for observational recruitment data.

Conclusions:

• While synthesized datasets offer short-term cost savings, they can limit the utility of stock assessments for real-world recruitment prediction.
• An equation-free modeling approach shows promise for more reliable recruitment forecasting in fisheries management.