Advancing Fisheries and Aquaculture Management with Machine Learning: Bycatch Risk Prediction and Autonomous Mortality Modeling

Machine learning is proving to play an increasingly important role in many fields, including ecology and fisheries sciences. Machine learning models offer many advantages over traditional statistical analysis methods, such as being well suited for analyzing large data sets, capable of incorporating multiple data types in a single model, and able to handle cross-correlated variables without severely affecting model performance. They can also potentially detect complex patterns among variables that other modeling methods cannot. As seafood demand continues to rise, it becomes crucial for managers of both wild capture fisheries and aquaculture to maximize their production while minimizing potential negative environmental and ecological impacts. This study aimed to demonstrate the utility of applying machine learning methods to fisheries and aquaculture management, specifically for bycatch analysis and for predicting caged-fish mortality and productivity in offshore aquaculture. Our study found that, machine learning methods effectively assess how fishing tactics, spatial, temporal and environmental variables impact bycatch risk of vulnerable organisms such as seabirds, helping refine fishing regulations to reduce future bycatch events. Hook density and float density had a very strong influence on bycatch risk and had been previously unexamined in bycatch models. This study also identified two machine learning model frameworks for effective bycatch prediction. Machine learning models were developed to predict fish growth and mortality rates, and facilitate determining the optimal harvest schedule for productivity and profit. The linear model was most effective for estimating daily growth of the cobia. ARIMA-ANN hybrid model framework was the best the five model frameworks for assessing mortality patterns in Cobia and in predicting future mortality rates. This automation is particularly important in offshore fish farms, which are in remote and hazardous environments. In conclusion, machine learning models can be extremely beneficial to advancing fisheries and aquaculture management.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

doctoral

Discipline thesis:degree_discipline

Fisheries and Wildlife Science

Department dc:contributor.department

Fish and Wildlife Conservation

Grantor dc:publisher

Virginia Tech

Year dc:date.issued

2026

Author dc:creator

• Pakzad, Iman Yeganeh

Chair dc:contributor.committeechair

• Jiao, Yan

Committee members dc:contributor.committeemember

• Galappaththi, Eranga
• Schwarz, Michael H.
• Zuo, Lei

• Seabird bycatch
• Pelagic longline fishery
• offshore aquaculture
• mortality
• U.S. Atlantic coast
• Random Forest
• XGBoost
• ARIMA
• Neural Network

Statement dc:rights

• In Copyright

Licence dc:rights.uri

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language.iso

en

Dc Identifier Other

vt_gsexam:45597

OAI identifier oai:identifier

oai:vtechworks.lib.vt.edu:10919/140951