SWAT-WB

A Water Balance Based Version of the
Soil & Water Assessment Tool

SWAT-WB is a slightly modified version of the USDA's Soil & Water Assessment Tool watershed model.  Instead of using the traditional Curve Number method to model surface runoff, SWAT-WB uses a physically based soil water balance.  This new method results in  a model that predicts runoff that is generated from saturated areas, contrary to the traditional SWAT. 

Introduction

The impetus for creating a version of SWAT free of the Curve Number methodology was born out of the Soil and Water Lab's work in two very different watersheds:  New York City's drinking water supply watersheds in the Catskill Mountains, and the Blue Nile watershed in the Ethiopian Highlands.

Differences between SWAT and SWAT-WB

The main difference between the original version of SWAT and this new version is the fact that SWAT-WB does not rely upon the Curve Number method to determine daily surface runoff volumes.  There are numerous justifications for using a soil water balance, rather than the Curve Number, to determine runoff volumes:

1. The Curve Number method, although widely used, was not designed for use in temporal watershed models.  While it can be successfully used as a calibration paramter for models such as SWAT, the original methodology was developed for use in determining streamflow from single events, not on a day-to-day basis.

2. The Curve Number method can be modified to account for different runoff generating processes, however, as it was designed (and implemented into SWAT) the method is simply a way to match observed rainfall data to predicted streamflows.  The Curve Number method does not indicate how the water in the stream originated within the watershed (see our descripition of runoff generating processes for a full description).  By using a soil water balance, SWAT-WB determines how much rainfall is required to saturate portions of the watershed, at which point surface runoff will be generated.

3. The original version of SWAT does already offer a Curve Number alternative, the Green-Ampt routine.  Green-Ampt is a physically-based method, however it requires hard to come by data (sub-daily precipitation records) and models runoff strictly from infiltration-excess conditions (i.e.  rain is falling faster than it can be absorbed into the soil).  While more physically based than the Curve Number, Green-Ampt is not appropriate for use in watersheds that have significant portions that are saturated.  Furthermore, the required sub-daily precipitation data presents a large obstacle in many minimally instrumented watersheds.  By using SWAT-WB, no extra data is required (whatever is needed for the Curve Number works for SWAT-WB), and runoff is modeled for saturation-excess conditions. 

4.  Although it has been used to model watersheds all over the world, the Curve Number method was developed using only data from a very small portion of the United States.  On the other hand, physically-based water balance models have proven to be simple, yet accurate, tools in predicting runoff and streamflows in virtually every climate around the world.  Using these time-tested water balance models within the SWAT framework results in a version of SWAT that can be successfully applied anywhere in the world, particularly areas in which saturation processes tend to govern surface runoff flows.

As can be seen from the above points, the novelty of SWAT-WB is that it models saturation-excess generated surface runoff with minimal changes to the SWAT data input requirements.  SWAT-WB, being based on a widely applicable soil water balance, is more appropriate for use in regions around the world where the Curve Number might not be the best choice.

With more and more researchers around the world developing SWAT models, we are confident that SWAT-WB provides an easy to use, physically based, accurate alternative to the Curve Number based version of SWAT.

SWAT-WB Results

SWAT-WB would not be a very compelling alternative to SWAT2005 if it did not produce better results.  However, that is not the case; SWAT-WB returned noticeably more accurate streamflow predictions for the two watersheds that we tested it on.  Fsull description of these test watersheds  and the model results are provided in the official SWAT-WB paper (available in the downloads section of this site), but the following tables summarize the improvements in model statistics when SWAT-WB is compared to SWAT2005.

Model Results for the Gumera Watershed in the Ethiopian Highlands

	SWAT-WB		SWAT2005 (CN)
	Calibration	Validation	Calibration	Validation
Nash-Sutcliffe	0.70	0.76	0.64	0.67
R-squared	0.71	0.81	0.65	0.73

Model Results for the Townbrook Watershed in New York State

	SWAT-WB		SWAT2005 (CN)
	Calibration	Validation	Calibration	Validation
Nash-Sutcliffe	0.64	0.52	0.43	0.62
R-squared	0.69	0.65	0.59	0.69

How to Use SWAT-WB

Once a SWAT project has been developed by the user (either using the GIS interface or by creating the text input files by hand), very few changes must be made in order to use SWAT-WB. A new one line text file must be created for each soil type present in the watershed.  These text files will be placed, along with the swatWB.exe executable file, into the project folder which contains all text files developed by the GIS interface.  The program, when run from this folder, will produce the same output files as the original SWAT program.

Resources and Downloads

At the top right of this page you will find files available for download. The first file contains everything required to run SWAT-WB, whereas the other files contain help documents as well as the source code which was used to compile SWAT-WB.