Comparative Time-Series Modeling and Forecasting of Tilapia Broodfish Growth in Pond and Recirculating Aquaculture Systems (RAS) Using ARIMA

This study applied time-series modeling using autoregressive integrated moving average (ARIMA) to compare the growth performance of tilapia broodfish in pond and recirculating aquaculture systems (RAS) from June 2023 to May 2024. Descriptive statistics showed a higher mean percentage weight gain under RAS (26.69%) than pond culture (23.75%), although monthly variability in the RAS dataset was influenced by an outlier, which may be attributed to influential exogenous factors rather than water-quality parameters. Normality, stationarity, and autocorrelation diagnostics confirmed that both datasets were appropriate for ARIMA modeling without differencing. Multiple ARIMA models were evaluated based on RMSE, MAPE, MAE, AIC, BIC, and residual behavior; ARIMA (1,0,1) emerged as the best fit for both systems. Forecasting up to May 2028 revealed stable long-term growth patterns, with RAS consistently showing slightly higher forecasted growth compared to pond culture, although the difference remained small in absolute terms. Predictions remained within model-generated 95% confidence intervals; however, these results indicate internal model consistency rather than independent validation of predictive accuracy. The findings highlight that RAS offers more consistent and slightly superior growth performance, supporting its potential for optimized broodfish production. Recommendations emphasize adopting RAS for enhanced growth predictability and improved management in tilapia aquaculture.