As any good sommelier will tell you, the character of a wine depends on the grapes, which, in turn, depend on the soil.  And what determines the character of the soil? In Oregon’s Willamette Valley, the answer is at least partially tied to a series of prehistoric flood events that occurred between 10 and 40 thousand years ago, carrying tens or hundreds of feet of silt to the valley floor.  HEC-RAS, on the other hand, is a type of modeling software developed by the US Army Corps of Engineers for hydraulic simulations, and Galileo is the easiest way to deploy HEC-RAS runs on remote machines and cloud. The common link between all three is Kleinschmidt Associates’ Chris Goodell, who built an expansive model of the Glacial Lake Missoula floods caused by the failure of a prehistoric ice dam. 

Why Model the Dam Breach and Glacial Lake Missoula Floods in HEC-RAS?

Goodell began work on this model in 2008, around the same time he launched his popular and widely respected blog, The RAS Solution.  His motivation was trifold.  To begin with, he was absolutely fascinated by the geology.   He also wanted to challenge himself. Was he capable of setting up and running such a complex model, and was HEC-RAS capable of handling it?  As a HEC-RAS expert and longtime instructor, Chris had always wondered about the software’s limits, in terms of what it can actually model. To his knowledge at the time of embarking on the study, the Missoula floods were the largest magnitude dam breach event ever to be modeled in HEC-RAS.  

As he started working on this in 2008, before HEC-RAS 2D was available, Goodell was initially working within a 1D framework.  When Hypernet Labs learned that he had also created a model using 2D, we were eager to test Galileo by running this computationally intensive version as our own feat of strength.  If Galileo can handle a remote run of a 2D simulation of the largest dam breach event ever modeled in HEC-RAS, it can run all other HEC-RAS simulations easily!    

Initial inspiration for the model came from Goodell’s discovery of a book by David Alt called Glacial Lake Missoula and its Humongous Floods (2001) at a conference.  He had always liked geology and found it fascinating that there had been a massive flood event right where he lived.  It had shaped many of the spectacular land features of the Pacific Northwest, some of which he drove past on a daily basis.  Willamette Valley agriculture was made possible by the tens or hundreds of feet of nutrient dense silt deposited by the flood events from eastern Washington.   

Setting up the HEC-RAS Model

The book not only sparked his interest, Alt’s work provided much of the information necessary to build the model.  The large amount of available data is what made it possible to simulate the flood event, at all. There’s an advantage to the fact that much of the flooding occurred in eastern Washington because the area is very dry, which means there is not a lot of vegetation to hide the geologic evidence, and there hasn’t been a lot of erosion to cover things up.  As a result, we have excellent aerial imagery. It’s possible to see the scarring resulting from these floods just by looking at Google Earth. In fact, while calibrating the model, Goodell would run it, map the flood extents, and then compare his map to the scarring on Google Earth to see if it matched up.    

To give some idea of the scale of the model, it initially included 592 cross sections with spacings from 1km to 25 km (5 km average).  As Goodell refined the model, there were eventually 2,346 cross sections, 68 reaches, 34 junctions, and 36 external boundaries. The results of the model provided further indication of the outsized proportions of the event, especially with regard to the magnitude of discharge, flow velocities, and inundation extents.

1D model of Missoula floods, Courtesy of Chris Goodell

The 1D and 2D results may look fairly similar to the untrained eye, due to the skillful setup of the 1D version.  However, Chris points out some significant differences between the two, including different areas that are wet or dry.  In addition, 2D allows you to have different water surfaces across any given transect, but 1D allows for only one surface across each cross section.  The 2D model also simply looks more sophisticated because the 2D mesh is much denser than the cross sections that are knitted together in the 1D version. 

1D model maximum inundation, 20-hr breach formation

2D results, images courtesy of Chris Goodell

Comparing HEC-RAS 1D and 2D Modeling–Advantages of 2D

One of the biggest differences between the 1D and 2D models is time, both in terms of model setup and run time.  In general, 1D modeling entails more setup time, while 2D modeling means longer run times and requires more computational power.  This is where Galileo comes in, as it can help minimize the 2D modeling time commitment by quickly and easily connecting modelers to optimized machines and freeing up their local computers.

To illustrate the vast difference in set up time: Goodell spent about a month of evenings and weekends, an hour or two here and there, setting up the 1D model before he could even run it.  Taking troubleshooting into account, he estimates that it was more like a couple of months before he had a working model, actually running and giving him results. Setup for HEC-RAS 2D is much faster because it uses the terrain behind the digital mesh to determine the direction of flow instead of relying on the manual input of the modeler’s assumptions.  With the 2D model, once Chris entered a digital terrain surface, he had it up and running within a couple of hours!

As expected, though, the 2D version took longer to run.  Goodell’s last, fastest run of the 1D model, after years of refining the model and clearing errors, took 44 minutes.  Using Galileo, which connected him to a remote 32 core machine, the 2D runtime was 2 hours, compared to a benchmark on his computer of 3+ hours.

[testimonial_slider autorotate=””]

How to Handle the Computational Intensity of HEC-RAS 2D

According to Goodell, 1D will continue to be useful, especially because 2D is so computationally intensive:  “If you want to model a long reach, it’s just way too prohibitive, and so you can combine 1D elements—for the efficiency—with 2D for model accuracy and make a really powerful model that way.  A lot of times we’ll use 1D to inform the boundary conditions for a more focused 2D area.” 

It’s actually the computational intensity of 2D modeling that makes it such a perfect use case for us at Hypernet Labs.  Galileo helps to remove the barriers to efficiency and make 2D modeling more practical by giving modelers extremely easy access to RAS-optimal machines and allowing them to run their models elsewhere, without tying up their own workstations.