The Automatic Parameter Optimization Program (OPT3) was designed to enable users of NWSRFS models to further refine parameter estimates previously developed through manual calibration procedures.
However, these automatic optimization procedures are not a substitute for the manual calibration process.
The following optimization strategies are currently available:
- Pattern Search (PATSERCH) algorithm
- Adaptive Random Search (ARS) algorithm
- Shuffled Complex Evolution (SCE) algorithm
An automatic optimization technique consists of two main components:
- a search algorithm
- objective function or optimization criterion
The objective function is a statistical measure of the difference between the observed and simulated hydrographs.
If N parameters are being optimized, the optimization criterion is a function of N-dimensional space.
The surface created by this criterion is called the response surface.
The automatic optimization procedure searches this surface to locate a minimum point, corresponding to an optimal set of parameter values.
The response surface is usually quite irregular with many bumps and dips, thus, it is difficult to know if the point at which the search algorithm stops is the global optimum or a local optimum corresponding to an inferior parameter set.