ADM1 was simulated using the Rosén and Jeppsson (2006)
implementation. Reaction rates are taken according to Table 1. Moreover, feeding with impulses generates transients in the transfer from the gas to the liquid (i.e. a negative specific mass transfer rate of CO2which can become smaller than with pressure in the headspace. A modification of the physicochemical reaction is therefore necessary when modelling the liquid–gas transfer. Computing the gas flow rate as the sum of the liquid-gas transfer rates (as proposed in Batstone et al. (2002))is therefore incorrect. The alternative solution consists in computing the gas flow rate from an overpressure in the headspace: atm gas