The present study deals with the respiratory activity of the microbial consortia established on surface-mould cheese. A software sensor was developed, from a model linking the operating conditions established in the ripening room (temperature, relative humidity, carbon dioxide and oxygen concentration) to the cheese mass dynamic. It estimated the respiratory activity of the microbial consortia based on mass measurement only, without atmospheric gas composition sensors or off-line measurements. After some simplifications of the model, this observer can be seen as an integral Luenberger observer, with analytically defined gain values. This approach was validated on two ripening trials. Relative errors between estimated and calculated (from carbon dioxide and oxygen concentration measurements) respiratory activity were equal to 3.02% and 8.83% for Run 1 and Run 2, respectively. (c) 2007 Elsevier Ltd. All rights reserved.