Predictive Analysis: Settings

Predictive Analysis requires completion of the fundamental setup and a successful history matching process (calibrated model). The parameter set found as a result

of the latter must have been applied as initial parameter values to the current FePEST setup.

Optimization Control

The required operation mode in PEST is the Predictive analysis mode, the optimization control section of the Problem Settings dialog contains the respective setting.

Observation and Type

Predictive analysis aims to either maximize or minimize a key prediction while maintaining the objective function below a user-specified value.

The prediction could be any observation (e.g. a specific hydraulic head) that is defined during the fundamental problem setup. Usually, this particular observation assumes zero weight so that it is excluded from the measure­ment objective.

The Predictive Analysis section of the Problem Settings dialog prompts the settings for the choice of the prediction whose value must be maximized/min­imized, and for the maximum value permitted for the objective function.

Objective Function Limits

The maximum/minimum prediction is sought subject to the constraint of main­taining the calibrated state of the model. It is important to define a maximum threshold for the objective function value. The model is considered to be cali­brated if the value of the objective function is less than the maximum thresh­old. This is done by providing objective function limits. The model is considered to be calibrated only if the value of the objective function is less than the defined target objective function.

A common choice is to chose a value which is 5-10% greater than the value of the objective function obtained in the preceding calibration process. How­ever, it should be noted that the choice can significantly impact the result. If the statistics of measurement noise are known, the limiting target objective function could be set in accordance with theory.

 

See the PEST Manual (5th Ed.), Chapter 6.2.2: PEST Variables used for Predic­tive Analysis for a full discussion of these variable.

Other Settings

The default settings are chosen such that they are suitable for doing predic­tive analysis for many cases.