Adding Snow to Estimates of Spring Flooding

By Corydon Ireland

The risk of flooding in the continental United States peaks every spring. Experts point to deep snowpack, late winter storms, and rapid melting.

In 2023, springtime floods in California got a lot of media attention. But such risk was universal in most of the forty-eight states south of Canada and north of Mexico.

National Weather Service spring flooding estimates in 2023 show that 146 million Americans were at risk, excluding Hawaii and Alaska.

The agency’s risk map predicted flooding across a wide swath of twenty-eight states in mid-America, plus patches in eight mountain and far-western states.

Even small-scale flooding presents risk of injury or death from drowning, electrocution, respiratory illness, and the harms that result from power failures.

A Pacific Northwest National Laboratory (PNNL) team of researchers has developed a new infrastructure design and flood-modeling technique to help anticipate flooding on a local scale.

Adding snow to flood calculations

To date, most small-scale engineering solutions for directing the sudden flow of water―catch basins, swales, detention ponds, and the like―are designed based on rain-only events. This decades-old predictive design standard is called―warning: engineering jargon ahead―the precipitation Intensity-Duration-Frequency curves technique.

The new-generation PNNL approach is based on a dataset that takes into consideration not just rain, but also regional snowpack and rain-on-snow events.

To create their dataset, PNNL researchers used more than 200,000 data points derived from rain and snow predictions nationwide. The data are based on simulation grids a little less than 4 square miles each.

“What makes our approach unique,” said Ning Sun, a PNNL hydrologist, “is that it can provide the exact time and magnitude of snowmelt at different locations.”

‘Simple, elegant, and useful’

Mark Wigmosta, a PNNL chief scientist, conceived of the project in 2017. Since then, he has joined Sun and PNNL surface water hydrologist Hongxiang Yan in writing a nine-paper series of studies that tell the tale of the approach’s evolution.

These include a proof-of-concept study (2018), an introduction of the idea to engineering communities (2019), and the first paper to validate the new method (2019).

Interest was immediate. A 2018 commentary in Water Resources Research called the PNNL approach “simple, elegant, and useful.”

The team’s most recent paper appeared in April 2022. It demonstrates that PNNL datasets show how much water reaches the land surface because of distinct contributions from rain, snowmelt, and rain-on-snow events.

Next: land-cover data

The team is working on the next evolution of the approach, which will incorporate eight new categories of land use and land cover. Original datasets in their new-generation approach are based on “open-area” land cover―that is, an imagined surface with minimal vegetation.

“Vegetation canopy intercepts some snowfall, which impacts the timing and final amount of water that reaches the land surface,” said Wigmosta. “It’s important to pre-calculate curves for all these variables.”

Even more dense and more comprehensive datasets are part of future versions of the approach, said Yan. “We have more datasets, more mature datasets, and we have done more theoretical work.”

The PNNL framework has matured beyond its original 2018 data, which represented only 376 western U.S. observation sites. The current dataset represents more than 200,000 sites. Soon there will be millions, thanks to simulations based on added vegetation cover.