The use of in-stream structures for habitat and stream restoration dates back to the early 1900’s (Thompson 2005); however, the design, effectiveness, and performance of these types of structures have not been well documented (Nagato 1998). River-spanning rock structures (rock weirs) are being constructed for water delivery, to enable fish passage at barriers, and to provide or improve the aquatic habitat for endangered fish species. Current design methods are based upon anecdotal information applicable to a narrow range of channel conditions with mixed success (Holmquist-Johnson 2011). Without accurate understanding of the hydraulics associated with rock weirs, designers cannot address the failure mechanisms of these structures (i.e., scour undermining the structure, rock movement, bank erosion).
In 2008, USGS Fort Collins Science Center (FORT) scientists began collaborating with Colorado State University (CSU) and the Bureau of Reclamation (Reclamation) on research that incorporates a multifaceted approach. This approach employs (1) field data to identify prototype conditions that provide practical and applied information, (2) laboratory physical modeling to isolate key variables and study individual processes in detail, and (3) computer modeling to develop design tools and predictive capability for analysis and design. Collecting enough detailed field and laboratory data to include a wide range of design parameters (structure geometry, sediment grain sizes, channel characteristics, etc.) and conducting an analysis structure performance for each design would be costly and take years to accomplish. Numerical modeling provides an opportunity to test design parameters over large ranges for lower cost and in less time than field measurements or physical models, for all possibilities.
While numerical modeling provides a method for testing design parameters over a large range of values, the computation time required for a single simulation can vary from 4 hours to multiple days, depending on the complexity of the model. Assuming that multiple simulations need to be performed, this can expend a considerable amount of time for just running the model before any analysis can be conducted. For the case where a Monte Carlo-type simulation or large parametric-type study is used, as with the rock weirs study, the total amount of computation time can easily turn into months.
To accommodate a sufficient number of parameter sweeps (enumerating sets of parameters to generate a model replication) for these numerical simulations, FORT is using High Throughput Computing (HTC). HTC provides a more efficient and much easier mechanism for distributing the simulations among a pool of computers by facilitating job management, matchmaking jobs with available machines, handling failed jobs, and utilizing all available computer resources belonging to the HTC system). At FORT, the HTCondor1 HTC system was used to execute both 2D and 3D simulations.
Using HTC for 2D/3D hydrodynamic numerical modeling is advantageous because an HTC system reduces the amount of time a scientist would require to manually run all simulations on several machines, and the results of the simulations can be achieved in a matter of hours or days rather than months. For example, if a simulation requires 4 hours of processing on a single machine, generally it will take the same amount of processing time per simulation when using HTC. Therefore, if we had to run 200 simulations, where each simulation runs for 1 hour, the total amount of time required to run all simulations on a single machine is approximately 200 hours (about 8 days if there is no down time after starting jobs). However, with HTC and a network of 100 machines available to run the simulations, these jobs could be completed in approximately 2 hours (see HTC Computing Times). Running these simulations first requires an individual to set up parameter files, preprocessing, and processing between 2D and 3D applications. Once that’s done, running the simulations using HTC, as FORT has demonstrated, turns the job around in a very short time.
The analysis and results of the numerical modeling along with field and laboratory data provide a process-based method for understanding how structure geometry affects flow characteristics, scour development, fish passage, water delivery, and overall structure stability. Work in progress is using the output from the numerical model simulations to develop quantitative design and analysis techniques. These developments will be integrated into a final report on the hydraulics and scour of river-spanning rock structures. Accordingly, the final report will provide tools and design considerations for engineers and managers to employ in designing more robust structures or retrofits based upon predictable engineering and hydraulic performance criteria.
1The use of any trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.