A Comparative Study of Photosynthetic Unit (PSU) Models for Algal Growth Rate and Fluorescence Prediction Under Light/Dark Cycles
AIChE Annual Meeting
Monday, November 14, 2016 - 10:00am to 12:30pm
Accurate description of light-limited algal growth, especially under short light/dark (L/D) cycles, is crucial for prediction of photobioreactor performance and for optimization of operating conditions. Short light/dark cycles have been shown to increase algal light usage efficiency and productivity. A variety of widely used photosynthetic unit (PSU)-based models were evaluated to determine their ability to predict algal mean specific growth rate and mean photochemical efficiency under a range of light/dark cycle conditions. Analytical or numerical solutions were derived, where necessary, for six models for quasi-steady-state light-limited growth under L/D cycles. These models were fit to previously published experimental data for algal mean specific growth rate and mean photochemical efficiency. Subsequently the weighted sum of squared error (SSE) values, normalized sensitivities to parameter change, and corrected Akaike Information Criterion (AICc) scores were compared. The quality of the fits and the model sensitivities were used to evaluate the assumptions and relative merits of the models considered. Sensitivity analysis revealed multiple adjustable parameters which had minimal effect on model predictions or showed strong correlations with other parameters, indicating that they could potentially be fixed to improve model quality. Each model generated qualitatively similar predictions of mean specific growth rate but their relative qualities could be distinguished through differences in mean photochemical efficiency predictions. An important model assumption was the use of adjustable rather than fixed exponents for the power-law relations governing PSU inhibition and total PSU concentration with respect to light intensity. Additionally, the assumption of PSU repair to the resting state results in better fit to experimental data than does the assumption of PSU repair to the active state. The Nikolaou and Papadakis models, which had more adjustable parameters, were shown to be unnecessarily complex for accurate prediction under L/D cycle conditions. The simpler Bernardi model scored significantly better on the AICc measure and shows good potential for future use in predicting algal behavior under L/D cycles.