The error residual for TOC analysis is examined with the result that two BOD5/TOC models are suggested, both of which use a multiplicative error structure. A comparison of these models with the historically-assumed linear BOD5/TOC model is created for data collected from the City of Waterloo's Water Pollution Control Plant. The predictive capabilities of the derived BOD5/TOC models are examined for interpolation and extrapolation potentials for augmenting water quality data information bases. Use of the models to reduce the statistical uncertainty associated with BOD measurements is considered.