Inferring Trends and Change Attribution Using Regression-Based Unimpaired Runoff to San Francisco Estuary, 1872-2022

Abstract This work utilized recently extended time series of “unimpaired” or “virgin” runoff from ten Sierra Nevada watersheds upstream of California’s San Francisco Estuary to study long-term hydrologic changes on annual and seasonal time scales. The runoff time series, spanning 1872–2022, were reconstructed using a multivariate regression approach that was informed with monthly temperature and precipitation data. Although the annual runoff series were found to be highly variable and were characterized by multi-year wet and dry periods, they were nevertheless found to be stationary over the full study period. This work also showed that April to July runoff, where snowmelt is a large component, exhibited a statistically significant decrease in most watersheds when reported as a fraction of total annual runoff. A linear trend method revealed statistically significant decreases in this metric in nine watersheds over the period of record; in six watersheds, a piecewise linear trend approach identified significant mid-20th century breakpoints signifying stationary conditions followed by substantial declining trends. The last decade of record showed the highest temperatures and greatest temperature-related declines in this metric. A change attribution analysis suggests that not only are warming temperatures driving runoff timing changes, but they are also impacting annual runoff volumes. This work confirms previous research indicating earlier runoff from snow-fed watersheds in western North America (including California’s Sierra Nevada Mountain range) in recent decades; it also highlights the stationarity in this quantity from the late 19th century to the middle of the 20th century before significant warming trends were apparent.