In this work dynamic mathematical models for the prediction of biogas production in a UASB reactor were developed. The dynamic modeling technique was applied successfully to a 2-year data record from a potato wastewater treatment plant. The technique used included regression analysis by residuals. Seventeen parameters were examined including the following: wastewater's flow rate, reactor's temperature and pH, total and soluble influent COD, wastewater's temperature and pH, total and soluble effluent COD, volatile fatty acids, alkalinity, biogas production rate and each parameter with a time lag of up to 10 days. Finally, after all parameters and all time lag trials three models were the best fitted models that were developed. The models' adequacy was checked by chi(2) test for a data record of the same UASB reactor but at a different time period and proved to be satisfactory. Additionally, a comparison among the three models was conducted as far as their ability to predict and to control biogas production rate is concerned. Through these models various aspects of the process can be enlighten, such as the fact that the hydrolysis of starch requires a resident time of 9 days. (C) 2004 Elsevier B.V. All rights reserved.