Modeling and Forecasting Primary Production Rates Using Box–Jenkins Transfer Function Models

Abstract

Box–Jenkins transfer function models were developed for time series of integrated hourly primary production rates. A 28-mo record of 56 biweekly measurements collected from seven mesocosms during a nutrient addition experiment was analyzed. Incorporation of two input variables (phytoplankton biomass and hourly light) significantly improved the fit of the models. When compared with standard regression models, the time series models all had reduced residual variance. The forecasting ability of the final fitted model for a control system was demonstrated with independent data from the two replicate control mesocosms.