Modelling is a process by which mathematical expression is used to describe a real quantitative situation in a system. The development of scientific understanding through quantitative expression of current knowledge of a system helps managers and planners to make tactical and strategic decisions. The study was designed to model catches for important cyprinid species (E. sardella) from Lake Malawi. The Box-Jenkins approach to modelling ARIMA (p, d, q) processes was adopted in this study. The Box-Jenkins methodology involved an iterative three -stage process of cautious model identification from ARIMA class, parameter estimation and diagnostic test. The results from the study showed that ARIMA (0,1,1) model had lowest Normalized Bayesian Information Criterion (NBIC), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) making it a suitable model for the study. The ARIMA (0,1,1) model showed that Lake Malawi E. sardella annual catches are fluctuating with a positive trend. The model further predicted that Lake Malawi E. sardella annual catches will increase from the annual average level of 68,742.29 metric tons to an average of 142,006.83 metric tons in the next 18 years. The positive annual catch trend could be due to harvesting pressure from the Lake. Therefore, the study provides critical information for future policy making and formulation of strategies to sustain fisheries resources in the lake.

Keywords: ARIMA, annual catch, Box-Jenkins models, forecasting, E. sardella, Lake Malawi, Modelling, Usipa