US Army Corps of Engineers
Northwestern Division Website

Corps partners with Boise flood control district to create predictive model

Walla Walla District Corps of Engineers
Published Jan. 15, 2021
Caldwell Idaho 2017 flood model with aerial image background

Caldwell Idaho 2017 flood model as seen in 2D modelling software with aerial image background. Corps engineers used their 2D model to replicate the 2017 flooding event. Recreating a past event is a good way to test if a model is accurate and helps determine whether it can be trusted to predict future events.

Caldwell Idaho 2017 flood model. Terrain Background

Caldwell Idaho 2017 flood model as seen in 2D modelling software with terrain background.

The Corps is using the latest technology to develop tools that communities in Idaho can use to predict flooding. The technology: Two-dimensional modelling.

In an effort to help anticipate and manage future flood events on the Boise River, the Corps of Engineers, Walla Walla District partnered with Flood Control District 10 to develop a two-dimensional (2D) model of a 75-mile stretch of the Boise River.

During the summer of 2019, Flood Control District 10 reached out to the Corps of Engineers, requesting a river model that could be used to predict flooding events. The model needed to extend from the Bureau of Reclamation Diversion Dam, just downriver from Lucky Peak Dam, all the way to the mouth of the Boise River, a total of about 75 miles of river and shoreline.

A team of Corps engineers was assembled to tackle the project, including hydraulic engineers Tracy Krause, Russel Lodge and Tracy Schwarz, eGIS Manager Sean Redar and Hydraulic Technician Gary Slack. Hydraulic Engineer Brandon Hobbs served as project manager.

The Corps partnered with Flood Control District 10 to split the costs of the project. The flood control district put together a coalition of stakeholders to contribute both funds and essential lidar data to help create a product to aid in the management of the Boise River. The list of stakeholders includes: the Idaho Water Resource Board; the cities of Boise, Caldwell, Eagle, Garden City, and Middleton; the Eagle Sewer District; the Ada County Highway District; the Treasure Valley Water Users Association; and the Pioneer Irrigation District.

The team of Corps engineers took topography data, mostly in the form of lidar data and geographic information system (GIS) geospatial data, and began assembling a 2D model. Lidar data is collected from an aircraft flying over the floodplain.  The aircraft repeatedly shines a laser at the ground and collects the reflections to determine the distance from the aircraft to the ground. When all the individual measurements are collected and put together, it forms a detailed representation of the shape of the ground. 

One-dimensional (1D) modelling, normally the standard for these sorts of projects, calculates water flow based on a series of cross sections. Cross sections are essentially slices across the flood plain. From the sky, each slice would look like a straight line. However, pulling the slice out and laying it on its side, as if it were a slice of bread, reveals the shape of the land, how it rises and falls. In a floodplain, the slice would show how the land dips down into the river channel then curves up again to the opposite bank. Extending beyond that, it might show the slope of a road or the curve of a hill, features that would affect how and where water would flow in the event of a flood.

Unlike slices of bread, slices in a 1D model have no thickness. The model computes what water does at each of those lines. Based on that information, assumptions can then be made for what water would do between each of the lines.

2D modelling, instead of taking slices, creates a grid of cells of multiple sizes. The model then runs a series of 1D computations for each side of each cell. This means it calculates the shape of each side of the cell and can then determine the overall shape of the cell, how the ground curves and slopes and how water would flow within that cell.

When all the cells are placed together the model can determine which way water would flow from any cell based on the relative elevations and slopes of nearby cells. This sort of modelling removes a lot of the guesswork present in 1D modelling and provides more accurate results.

“The point of the 2D modelling is instead of working on slices, we take the world, put a grid on it, and we let the water move where it will. Our software allows it to spread out, to move just like you would think it would, through the low spots first and then filling up to the higher spots if there’s enough water,” Brandon Hobbs, project manager for the Walla Walla District, said.

The downside to 2D modelling is that it requires a lot more computing power. Running the model for the entire 75-mile stretch of river takes approximately 5 days. However, there is also the option to run small sections of the model, which would take a fraction of that time. This is useful for groups or individuals wanting to predict flooding behavior on a specific stretch of the river, say near a specific community or city.

“The Corps is really working to support the flood control district, giving them the latest technology at an affordable price that’s usable for the end user. We’re trying to provide them with a tool that is reasonable for people who don’t necessarily have the fastest computers,” Tracy Schwarz, hydraulic engineer for the Walla Walla District, said.

Corps engineers spent months fine-tuning the model. They had to make adjustments to account for structures like bridges, irrigation canals and gravel quarry pits. Bridges were a problem because they obstructed the lidar data (gathered overhead by an aircraft) that mapped the bottom of the river. To ensure a more accurate model, engineers had to insert information on the shape of the river channel under each bridge where lidar data could not see. Irrigation canals and gravel quarry pits were also challenging to account for because they are not always present or accurately depicted in geospatial data.

“It took a lot of time. It took months to work the issues out,” Schwarz said.

Finally, Corps engineers were able to use their 2D model to replicate the 2017 flooding event. Recreating a past event is a good way to test if a model is accurate and helps determine whether it can be trusted to predict future events. In September, the draft model was sent out and reviewed by the flood control district’s technical review team who sent back technical comments and recommendations.

The completed model is expected to be sent to Flood Control District 10 sometime in January or February, just in time for the 2021 flood season.