This paper deals with the design of a software sensor based on a high-order sliding mode observer (HOSMO) for online estimation of the concentration of volatile fatty acids (VFA's) in a class of continuous anaerobic digestion processes (AD). Taking into account the limited availability for online monitoring of AD processes, in this contribution it is assumed that only the methane outflow rate is available for online measurement. Considering a simple two-dimensional AD model, it is shown that the VFA concentration can be detectable from the online measurement of the methane outflow rate. The estimation method is based on a local coordinate transformation that converts the original model into an observable system. In the observable form, a supertwisting algorithm is proposed in order to provide robustness under perturbation and uncertainty conditions. The observer convergence is analyzed by using Lyapunov stability techniques. Numerical simulations illustrate the effectiveness of the proposed estimation scheme for a four-dimensional AD model with uncertainties associated with unmodeled dynamics and disturbances in the inflow composition.