Improved AI process could higher predict water supplies

Publishing within the Proceedings of the AAAI Conference on Artificial Intelligence, the interdisciplinary group of Washington State University researchers predict water availability from areas within the West where snow amounts aren’t being physically measured.

Comparing their results to measurements from greater than 300 snow measuring stations within the Western U.S., they showed that their model outperformed other models that use the AI process often called machine learning. Previous models focused on time-related measures, taking data at different time points from only a couple of locations. The improved model uses each time and space into consideration, leading to more accurate predictions.

The knowledge is critically vital for water planners throughout the West because “every drop of water” is appropriated for irrigation, hydropower, drinking water, and environmental needs, said Krishu Thapa, a Washington State University computer science graduate student who led the study.

Water management agencies throughout the West every spring make decisions on the way to use water based on how much snow is within the mountains.

“It is a problem that is deeply related to our own lifestyle continuing on this region within the Western U.S.,” said co-author Kirti Rajagopalan, professor in WSU’s Department of Biological Systems Engineering. “Snow is unquestionably key in an area where greater than half of the streamflow comes from snow melt. Understanding the dynamics of how that is formed and the way that changes, and the way it varies spatially is de facto vital for all decisions.”

There are 822 snow measurement stations throughout the Western U.S. that provide each day information on the potential water availability at each site, a measurement called the snow-water equivalent (SWE). The stations also provide information on snow depth, temperature, precipitation and relative humidity.

Nevertheless, the stations are sparsely distributed with roughly one every 1,500 square miles. Even a brief distance away from a station, the SWE can change dramatically depending on aspects like the world’s topography.

“Decision makers take a look at a couple of stations which might be nearby and make a call based on that, but how the snow melts and the way the several topography or the opposite features are playing a job in between, that is not accounted for, and that may result in over predicting or under predicting water supplies,” said co-author Bhupinderjeet Singh, a WSU graduate student in biological systems engineering. “Using these machine learning models, we are attempting to predict it in a greater way.”

The researchers developed a modeling framework for SWE prediction and adapted it to capture information in space and time, aiming to predict the each day SWE for any location, whether or not there may be a station there. Earlier machine learning models could only deal with the one temporal variable, taking data for one location for multiple days and using that data, making predictions for the opposite days.

“Using our latest technique, we’re using each and spatial and temporal models to make decisions, and we’re using the extra information to make the actual prediction for the SWE value,” said Thapa. “With our work, we’re trying to remodel that physically sparse network of stations to dense points from which we will predict the worth of SWE from those points that have no stations.”

While this work won’t be used for directly informing decisions yet, it’s a step in helping with future forecasting and improving the inputs for models for predicting stream flows, said Rajagopalan. The researchers can be working to increase the model to make it spatially complete and eventually make it right into a real-world forecasting model.

The work was conducted through the AI Institute for Transforming Workforce and Decision Support (AgAID Institute) and supported by the USDA’s National Institute of Food and Agriculture.