There is no doubt that a reasonable set of parameters can be obtained using only manual methods.
However, the procedure is time-consuming in terms of person-hours and requires a degree of interplay with the computer, often not available from larger systems.
In addition, the hydrologist performing the optimization must possess a considerable degree of skill acquired through experience with the model.
Automatic methods, on the other hand, are fast and simple to use but have some inherent disadvantages.
Some of these are complete dependency on one error function, failure to attain an optimal solution due to non-convexity of the response surface in the vicinity of the starting point, and failure to recognize the effect of perturbing a group of parameters simultaneously.
Experience in fitting the model to a large number of catchments under operational conditions indicates that the procedure should be one involving both manual and automatic fitting where the strong points of each compensate the weak points of the other.
Generally, much more is achieved by fitting manually first, then using the automatic optimizer after a reasonable fit has been obtained.
Data requirements for the model are somewhat greater than for simpler 'event' type models, since the model utilizes a continuous record rather than a fragmentary one covering selected periods.
The length of the database required for adequate calibration depends on a number of factors including the hydroclimatic characteristics of the catchment and the amount of hydrologic activity during the period in question.
Typically, however, a continuous record of 8 to 10 years is required for reasonable calibrations.
