HSPF is a comprehensive, conceptual, continuous watershed simulation model designed to simulate all the water quantity and water quality processes that occur in a watershed, including sediment transport and movement of contaminants. Although it is usually classified as a lumped model, it can reproduce spatial variability by dividing the basin into hydrologically homogeneous land segments, simulating runoff for each land segment independently, using different meteorological input data and watershed parameters. The model includes fitted parameters as well as parameters that can be measured in the watershed.

HSPF is a public domain software program distributed by the U.S. EPA's Center for Exposure Assessment Modeling (CEAM). It is available for DOS systems from:

        http://water.usgs.gov/software/surface_water.html

However when using PEST with HSPF it is best to use a slightly modified version whose data
input protocol has been slightly altered for easier data transfer between PEST and HSPF. You
can download such a version (slightly modified from version 12 of HSPF - the latest version)
by clicking here.

Using PEST with HSPF

The PEST Surface Water Utilities make it very easy to use PEST with HSPF. Using these utilities you can do the following:

• automate the construction of PEST input files for calibration of HSPF; whether you are calibrating against a handful of observations or thousands of observations, setup time can be reduced to minutes.

• allow PEST to run a HSPF post-processing program as part of a "composite model" used during the calibration process. This post-processor time-interpolates HSPF outputs to the times and dates of corresponding flow and/or constituent measurements so that the two can be readily compared by PEST as part of its parameter adjustment process.

• incorporate "volumetric measurements", exceedence times and other statistics based on observations and model outputs into the calibration process. A comparison between "observed" flow volumes (calculated over the whole simulation time or various parts of it) against volumes calculated by the model over similar time intervals, reduces parameter nonuniqueness, brings stability to the parameter estimation process, and heightens the ability of the calibrated model to predict future flow volumes accurately.

• adjust the observation weights on a PEST input file with ease, no matter how many observations are included in the calibration dataset. A proper weights assignment strategy can assist the model calibration process enormously.

• calibrate multiple watershed models simultaneously. PEST can be asked to assign the same parameter values to all watersheds, or (using its amazing regularization functionality), assign parameters that differ between watersheds only to the extent needed to ensure calibration of all watershed models.

• explore the uncertainty of key model predictions; this can be done in a variety of ways including the use of PEST's predictive analyzer, or through "calibration-constrained monte-carlo analysis" (the latter being made possible through the use of PEST's advanced regularization capabilities).

• convert HSPF outputs to formats where they can be easily plotted by commercial plotting and spreadsheet software for comparison with field data.

• undertake baseflow separation, based on digital filtering. It is now possible to include quickflow and baseflow as separate components of an overall calibration objective function, tailored to best suit the purposes for which the model is constructed.
   

A comparison between modeled and observed flows over part of a 16 year calibration period.
 

 A comparison between modeled and observed monthly volumes over the entire 16 year
calibration period.
 

 A comparison between modeled and observed exceedence probabilities.
 

 "Calibration-controlled Monte-Carlo analysis" allows the range of predictive uncertainty
to be explored.

Learn How to Use PEST with HSPF

Download a training document providing a detailed practical exercise in calibrating a HSPF model using PEST. Download the files you need to run the calibration exercise. You will also need the PEST Surface Water Utilities, as well as a special version of HSPF (see above) whose data input protocol has been altered slightly for optimal use with PEST.

Further Information

While the use of PEST to calibrate a model can often lead to a much better fit between model outputs and corresponding field measurements in a much shorter time than if calibration was carried out done by hand, the best thing about using PEST is that you will understand far more about the information content of your data, the uncertainty surrounding parameters estimated on the basis of that data, and about the reliability of predictions made using the calibrated model. Using PEST for model calibration is a whole new way of doing things, engendering a much better understanding of the model calibration and deployment process, and a far more informed use of the model in the making of management decisions than has hitherto been possible.

A recent publication documenting the use of PEST with HSPF is: 

Doherty, John, and John M. Johnston, 2003.   Methodologies for calibration and predictive analysis of a watershed model, J. American Water Resources Association, 39(2):251-265. 

Another paper describing the use of PEST in TSS model calibration has also been submitted; contact us if you would like a copy.

The use of PEST, and the role of calibration in the overall model construction and deployment process, is taught in PEST courses which are presented from time to time in different parts of the world.