Innovative Satellite Image Data Analysis System to Improve Snowmelt Runoff Modeling Across Idaho
September 12, 2019
U of I Department of Biological Engineering doctoral student Craig Woodruff and College of Engineering associate professor and Idaho State Climatologist Russell Qualls have developed a mathematical model that processes thousands of remotely sensed images to synthesize the annually recurrent pattern of snowmelt from mountainous watersheds and have applied this process to images of the Upper Snake River basin. Their research, "Recurrent Snowmelt Pattern Synthesis Using Principal Component Analysis of Multiyear Remotely Sensed Snow Cover," published in the journal Water Resources Research, was recently selected for recognition in a research spotlight article in the American Geophysical Union's (AGU) online magazine, Eos.
In the Western United States, snowmelt makes up between 50% to 80% of the annual water flow used for irrigation, drinking water and other uses. Tracking snowmelt runoff and streamflow helps water management and agricultural industries plan better business operations around water availability.
AGU is a nonprofit organization regarded as the premier institute for geophysicists and hydrologists, with more than 62,000 members from 144 countries. The editors of AGU's 22 journals select a few articles per month from among those published for recognition in Eos.
Remote sensing involves using satellites to detect the physical characteristics of an area, in this case the spatial distribution of snow, by measuring reflected radiation from a grid of points or pixels to produce an image. These images would be useful to model runoff, but cloud cover often renders days to weeks worth of remotely sensed images unusable, disrupting the modeling process.
"A cloud-covered image is like an incomplete jigsaw puzzle," Qualls said. "Current methods to complete the 'puzzle' primarily wait until subsequent images become available that reveal the condition, or presence or absence of snow, of the cloud-covered pixels. This delays image-based runoff modeling until after runoff has already occurred."
The recurrent snowmelt pattern that their model generates, "is like the picture on the puzzle box; it allows one to fill in the gaps of daily cloud-covered images in near-real time, based solely on what is visible in a small portion of the image," Qualls said.
The pattern followed by snowmelt on a watershed is relatively static even though the timing and duration of melt changes from year to year, which is what allows a single pattern to be extracted and produced from many years of remote sensing images. The snowmelt pattern looks much like the contour lines on a map, except the contours represent the location of the snowline as it moves progressively up through the watershed during the melt season.
Woodruff and Qualls applied their model to the headwaters of the Snake River, a 3,500-square-mile watershed which supplies about one-third of the flow of the Snake River across southern Idaho, to synthesize the recurrent snowmelt pattern of the watershed from 17 years of daily remotely sensed satellite imagery. Using the synthesized snowmelt pattern, they achieved 85% to 97% accuracy in cloud-covered image recovery.
"This is a pretty big advancement for our understanding of snow," Woodruff said. "A lot of places are reliant on stream flow for drinking water, and we're starting to understand that process a little more."
Woodruff said the U.S. Army Corps of Engineers conducts an annual helicopter flight over the Bitterroot Mountains to obtain a one-day snapshot of snow cover conditions of the watershed supplying Dworshak Reservoir in northern Idaho. They use this information for runoff modeling to make reservoir water-release decisions. He said this research could provide snow cover data on a daily basis for input to their runoff model, serving as a valuable decision support tool.
Qualls said he anticipates widespread interest in this research. Agencies that track snow and model runoff including the Idaho Department of Water Resources, National Oceanic and Atmospheric Administration, National Weather Service, the Natural Resource Conservation Service, the U.S. Bureau of Reclamation, and Idaho Power.
The research has been funded through the Idaho Water Resources Research Institute, the United States Geological Survey, as well as by the USDA National Institute of Food and Agriculture through Agricultural Experiment Station research in the U of I College of Agricultural and Life Sciences. Qualls said additional funding is being sought through NASA.
"The research has applications beyond this," said Woodruff. "This can be used for all types of modeling and climate change research."
Qualls said the work has potential to shift the paradigm of how remotely sensed snow images are processed and how snowmelt runoff modeling is performed, benefiting water users in Idaho and across the globe.
About the University of Idaho
The University of Idaho, home of the Vandals, is Idaho’s land-grant, national research university. From its residential campus in Moscow, U of I serves the state of Idaho through educational centers in Boise, Coeur d’Alene and Idaho Falls, nine research and Extension centers, plus Extension offices in 42 counties. Home to nearly 12,000 students statewide, U of I is a leader in student-centered learning and excels at interdisciplinary research, service to businesses and communities, and in advancing diversity, citizenship and global outreach. U of I competes in the Big Sky and Western Athletic conferences. Learn more at uidaho.edu