Surface Water Utilities

Much can be done in calibration of surface water and land use models that is not being done now. Highly-parameterized, regularized inversion based on a multi-component objective function that blends expert knowledge with historical observations of flow and system state best serve this process. PEST’s surface water utilities perform model post-processing and PEST setup tasks that expedite this process.


History-matching of surface water and land use models is most effectively done if it is based on a strategic, multi-component objective function. The value of this function is reduced as model-to-measurement fit improves. Ideally, each component of this objective function should be comprised of field data on the one hand and its model-generated counterpart on the other hand, processed in a certain way. If done properly, the processing behind each component should distil information of a different type from the measurement dataset. Each of these components should therefore inform different parameters, or groups of parameters. Each objective function component should be weighted for visibility in the total objective function. 

The components of a multi-component objective function may include the following:

  • Flows and/or transformed flows (with weights that are functions of flow);
  • A conveniently-extracted approximation to baseflow (enabling visibility of recessions in the total objective function);
  • Flow duration statistics;
  • Monthly or event volumes;
  • Functions of flow that are similar to those on which decision-critical predictions are based.

Where models are built to assist in the management of water quality, the objective function should include similarly-creative components that support assimilation of information that is resident in historical measurements of water constituents.

Parameterization of surface water and land use models, as currently implemented by the environmental industry, also needs improvement. How else can expert knowledge of parameter value relativity between different land uses, soil types and watersheds be introduced to an inversion process other than by simultaneous calibration of models for multiple watersheds, with Tikhonov regularization enforcing expert-knowledge-based ordering relationships? This is easily done using regularized inversion and uncertainty analysis functionality provided by members of the PEST and PEST++ suites.


TSPROC is a time series processor written specifically to assist calibration of surface water and land use models. This is the flagship of the PEST surface water utility suite.

TSPROC can be run in either of two ways. Firstly, it can act as a model post-processor. In doing so, it undertakes interpolation from model output times to field measurement times. Then it computes paired functions of field measurements, and model-calculated counterparts to field measurements, so that the two can be compared. 

Secondly, TSPROC can be run as a PEST dataset constructor. As such, it writes PEST control and instruction files through which PEST-based, or PEST++-based, history-matching can employ a multi-component objective function constructed in ways described above.

You can obtain TSPROC by downloading the PEST surface water utilities. An enhanced (by the USGS) version of TSPROC can be obtained here.