Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: Parallelization of the SWAT Watershed Model for Application on the Supercomputers at the Hawaii Supercomputer Center
2012 Annual Report


1a.Objectives (from AD-416):
To efficiently run the SWAT watershed model on a supercomputer for application to water management and sustainability on Maui.


1b.Approach (from AD-416):
To develop spatially detailed surface and groundwater models for the island of Maui and to take advantage of the power of supercomputers, the SWAT model needs to be parallelized to run on multiple processors. Two approaches will be analyzed and developed to parallelize the SWAT model simulations. The first approach is to disaggregate SWAT input files to simulate each individual subwatershed within a basin as a separate model run. This allows each subwatershed to be run simultaneously on different processes. It requires the subwatersheds to be run in sequence defined by stream order and passing large output files between simulations. The second approach requires SWAT to be recoded to run the entire basin in parallel. Scientists at the University of Hawaii have considerable expertise in developing parallel code and have programming techniques to run subwatersheds on different processors within the same simulation. A SWAT simulation for the entire island of Maui will be developed and both parallelization methods will be run on at the Hawaii Open Supercomputer Center. Advantages and disadvantages of each approach will be analyzed. Also, we will examine the potential link of SWAT to the groundwater model, MODFLOW, and define issues related to linking the models and running the combined model in parallel on a supercomputer.


3.Progress Report:

The goal of this project is to efficiently run the SWAT watershed model on a supercomputer for application to water management and sustainability on Maui; this understanding directly contributes to objective 1 of the in-house project 5320-21610-001-00D which this project was originally established under (expired in FY2011).

A collaboration with the UH International Pacific Research Center (IPRC) has been established to complete a micro climate model of Maui which will be used as inputs to the SWAT model. The climate simulation will be completed at a minimum grid size of 1 km and forward for over twenty years.

A meeting at University of Hawaii Manoa was conducted on the week of June 25, 2012 to review progress on the overall Navy/USDA effort to determine the feasibility of converting the HC&S commercial sugar operation on 40,000 acres to biofuel production. On June 26 HOSC briefed USDA/ARS, Temple Texas, the developer of SWAT on the early results of the micro climate modeling for Maui. The micro climate is an innovation in climate modeling which typically involves a much larger geographical region. The micro climate model required a series of nesting computations which grew increasing more time consuming. At the current time the climate model for Maui has been completed for a few years but the objective will be to reach a prediction period of fifty years.

The UH HOSC team is working with USDA/ARS, Temple, Texas on the parallelization of the SWAT code (pSWAT). We received a copy of the SWAT code which will be paralyzed at UH. The code has been installed on the HOSC supercomputer and is currently being tested on a single core. Initial changes in the code are being implemented and tested to insure that no spurious effects have been introduced. We will compare results from a single core computation with the pSWAT results and validate the accuracy of the parallelized code. We have received two large datasets for two previous studies which we will compare with the results from pSWAT. We are also preparing a Microsoft operating system version of pSWAT to operate on smaller multiple processors.


Last Modified: 11/27/2014
Footer Content Back to Top of Page