Observation data

The "observation data" section of a PEST control must contain as many lines as there are observations. Each line must contain four entries. Variables that appear on each line are shown below.

The first character on each line is the observation name. This can be up to 20 characters in length. (It can be up to 200 characters for programs of the PEST++ suite). The observed value of the observation is next. This is followed by the observation weight, and then by the name of the observation group to which the observation belongs (12 characters or less for PEST-suite programs and 200 characters or less for programs of the PEST++ suite).

The first part of the "observation data" section of a PEST control file follows.

Members of the PEST++ suite can read observation data from one or more external files. This protocol is supported by only one PEST-suite program, namely RSI_HP. RSI_HP is able to read observation data from a single external file. At the same time, it is able to read the value of one extra variable, namely STANDARD_DEVIATION. The "observation data" section of a PEST control file that supports this protocol is shown below. In this case, observation data is read from a file named observ.dat.

The first part of file observ.dat is shown below.

Note that the header to this file is not commented out. It is part of its protocol. Note also the header for the final column, namely "STANDARD_DEVIATION".

Now, you may be asking yourself why observation standard deviations are supplied if each observation weight is supposed to be the inverse of the standard deviation of measurement noise. The many uses of weights are discussed elsewhere. For the moment it is important to point out that weights are used in calculation of the objective function. Standard deviations are used in adding realisations of random noise to observations during ensemble-based history-matching.

Observation groups

Part of the "observation groups" section of a PEST control file is shown below.

The "observation groups" section of a PEST control file should provide a list of observation group names, one to a line. If a covariance matrix is assigned to an observation group, then the name of the file that holds this matrix should follow the name of the observation group. Where a covariance matrix is assigned to an observation group, weights assigned to observations belonging to that group in the "observation data" section of the PEST control file are ignored.

A covariance matrix must be positive definite. This means that you cannot build it yourself; you need software help. The PEST groundwater utilities provide this help for covariance matrices which accompany prior information equations that are used for regularisation and uncertainty analysis of pilot point parameters. These matrices specify prior parameter spatial covariances. See PPCOV, PPCOV3D, PPCOV_SVA and PPCOV3D_SVA; these utilities are provided with the groundwater utility suite.

A covariance matrix is assigned to an observation group. This matrix must be square, with N rows and N columns where N is the number of observations in the observation group to which the covariance matrix is assigned. The matrix must be supplied in a text file. Optionally, the matrix may be preceded by a three-number header which specifies the dimensions of the following matrix. Optionally, the matrix may be followed by row and column names. See Part 1 of the PEST manual for details. See also documentation of PEST matrix file format in Part 2 of the PEST manual.