When observed data are used to infer on parameters that are not directly observable, usually an inverse problem needs to be solved. Characteristic for inverse problems is their ill-posedness, which in practice means that small errors in data may propagate to huge inconsistencies in the solution if the problem is not properly regularized or augmented with prior information. An important part of solving inverse problems is to carefully model the noise that in addition to the exogenous noise contains possible modeling errors. One potentially significant source of modeling error is inadequate modeling of the sensor. Poorly modeled sensor cause a discrepancy between the model prediction and observation that often has a non-vanishing mean and high correlations between channels, thus producing artifacts in the solution of the inverse problem. In this talk, we briefly revisit the electrode modeling in electrical impedance tomography, and discuss in more detail the modeling challenges of surface pH electrodes used in the study of cell membrane permeability.