With one exception, the measures of simulation error or objective functions (OFs) (later transformed to likelihoods) were formulated in terms of the absolute magnitude of the relative deviation from the observed value, i.e., Fs = ∣SS − SO∣/SN, where SS and SO denote values for the simulated and observed shape descriptors, respectively. The normalizer (SN) is either the observed value or a user-selected constant value (e.g., a representative radar time step of 275 s for timing-based OFs like FTP in Table 3). This formulation makes the OFs comparable across different peaks, and for magnitude-based OFs like FQP (Table 3), accounts for heteroscedasticity (i.e., nonconstant variance) in the residuals [Sorooshian and Dracup, 1980]. Apart from the five basic shape descriptors, other OFs (Table 3) considered flow magnitude, volume, timing, slope, skew and net time series errors. Skew was defined as the hydrograph risetime divided by the total hydrograph time. For the slope-based OFs over different portions of the hydrograph, the magnitude and time terms in the slope equation were normalized using respective observed values to impart a 45° angle to the observed slope. The OFs apply to single peaks, except the last two OFs in Table 3 that apply to the entire event.