Mathematical modelling of greenhouse gas emissions from membrane bioreactors: A comprehensive comparison of two mathematical models.

This paper compares two mathematical models (Model I and Model II) to predict greenhouse gases emission from a University Cape Town (UCT) - membrane bioreactor (MBR) plant. Model I considers NO production only during denitrification. Model II takes into account the ammonia-oxidizing bacteria (AOB) formation pathways for NO. Both models were calibrated adopting real data. Model comparison was performed in terms of (i) sensitivity analysis (ii) best fit and (iii) model prediction uncertainty. On average 6% of factors of Model I and 9% of Model II resulted to be important. In terms of best fit, Model II had a better capability of reproducing the measured data. The average efficiency related to the NO model outputs was equal to 0.33 and 0.38 for Model I and Model II, respectively. On average, 73% (Model I) and 86% (Model II) of measured data lay inside the uncertainty bands.