Practical Glass-Box Machine Learning for Seasonal Water Supply Forecasting, with Applications to the Owyhee and Yellowstone Rivers: Regression Using Climate Indices Derived from SNOTEL Data Using Nonnegative Matrix Factorization with k-Means Clustering
W. Fleming, V.V. Vesselinov, A. Goodbody
AGU Fall Meeting2022
Summary
Hydroclimatic explainability – a straightforward, concise, intuitive, and defensible ‘storyline’ framed around commonly accepted atmospheric and terrestrial hydrologic processes and conditions – is a key requirement for many forms of hydrologic modeling. This includes seasonal water supply forecasts (WSFs), which are the information backbone of the massive water management infrastructure in the mostly dry and increasingly thirsty American West.