A stochastic modification of the basic Vollenweider model was used to represent variability in daily phosphorus concentrations in a small southern hypereutrophic lake. Uncertainty measurements, arising from defining initial conditions such as loading estimates, and a lumped white-noise parameter were used to project probability distributions of mean daily in-lake phosphorus concentrations. Additional consideration of spatial variability permitted comparison of model projections with monitoring data. Model calibration was performed with over 500 temporally and spatially distributed total phosphorus observations, ranging from 0.171 to 1.575 mg/L, taken from a 70-day intensive monitoring program in 1983. Results indicate the model's potential usefulness, particularly in unstable systems, although short-term oxygen depression, stratification, and catastrophic events, such as a 12.7 cm rainstorm, limited its predictive capabilities.