Faculty Interaction

Indrajeet Chaubey

Department of Agricultural and Biological Engineering
Purdue University

RCAC staff: M. Sayeed

Brief Project Description

Evaluating Best Management Practice Impacts on Water Quality Considering Stochasticity of Weather The goal of this project is to analyze various watershed management practices for phosphorus pollution control and its effects on water quality, considering uncertainties in future weather patterns over a 25-year period. The focus is on the Lincoln Lake Watershed, a 32 km2 (12.4 mi2) watershed located in Washington County of Northwest Arkansas (figure 1) where significant land use changes have occurred in this watershed over the past 10-12 years with the predominant change being the conversion of pastures to urban areas.

The study uses SWAT (Soil and Water Assessment Tool), a physically-based distributed-parameter watershed model to simulate 172 management scenarios incorporating 1000 stochastic weather realizations, thus, 172,000 model runs in total. At the rate of 8-10 minutes/run, 23,000-29,000 CPU hours, or 2.6-3.3 years would be needed to complete all runs on a standard LINUX system. Further, 67.5 Mb/run are needed to store output, translating to 10 Tb of space required. Thus the need to develop suitable and efficient techniques to handle such comprehensive BMP (Best Management Practices) evaluations is very important.


The scale and nature of these computational simulations makes it ideal for a high throughput computing environment such as Condor. The SWAT program was ported from the Windows platform to a Linux system that allows condor runs. Modifications were made to the program to automatically read the appropriate input files. Also, scripts were written to modify the input files and to seemly perform the large number of condor runs. Although the total number of runs was reduced to save on storage, we were able to do a large number of runs within a period of two months.

M. Sayeed from RCAC's Performance Engineering Group enabled the Chaubey group to take advantage of Purdue's Condor pool , and helped them get access to Teragrid resources.

References-
Draft of the paper titled "Computational Approaches to Evaluating BMP Scenarios Considering Stochasticity of Weather; M.W. Gitau, L. Chiang, M. Sayeed, and I. Chaubey" is ready.

 

Last updated 03/09/08