Regression and Mixed-Effects Modeling of Fish Growth, Survival, and Yield in Semi-Intensive Aquaculture Systems in Gusau, Zamfara State, Nigeria

Abstract

This study applied regression-based predictive modeling to estimate fish growth and yield performance in twelve semi-intensive ponds at the Federal College of Education (Technical), Gusau, Nigeria. Water quality indicators (temperature, dissolved oxygen, ammonia, pH), feed quantity, and stocking density were evaluated as key predictors of mean body weight, survival, and total biomass of Oreochromis niloticus and Clarias gariepinus. A repeated-measures design was used, with 20 fish sampled per pond every two weeks throughout the 11-month study period. Linear regression revealed strong positive effects of temperature, dissolved oxygen, feed amount, and pH on growth, while stocking density had a negative influence. Logistic regression showed that dissolved oxygen increased survival likelihood, whereas ammonia and high stocking density significantly reduced survival probabilities. The mixed-effects model demonstrated the highest predictive capability, indicating notable pond-level variation due to micro-environmental differences despite uniform management. Model comparison results confirmed that incorporation of random pond effects improved accuracy (R² = 0.91) over standard linear and logistic models. These findings demonstrate the potential utility of regression-based models for improving decision-making in semi-intensive freshwater aquaculture systems.