Structure and Detectability of Trends in Hydrological Measures over the western United States

This study examines, at 1/8 degree spatial resolution, the geographic structure of observed trends in key hydrologically relevant variables across the western United States (U.S.) during the period 1950-1999, and investigates whether these trends are examples of non-stationarities significantly different from naturally occurring climate variations. Variables analyzed included late winter and spring temperature, winter-total snowy days as a fraction of winter-total wet days, 1st April Snow Water Equivalent (SWE) as a fraction of October through March precipitation total (PONDJFM), and seasonal (January-February-March; JFM) accumulated runoff as a fraction of water year accumulated runoff. To distinguish variability within the bounds of naturally occurring stationary distributions of climate, observed trends were compared to the low-frequency natural internal climate variations simulated by an 850-year control run of the CCSM3-FV climate model linked to the Variable Infiltration Capacity (VIC) hydrological model. Large trends (with spatial coherence and magnitudes found less than 5% of the time in the long control run) are common in the observations and occupy substantial parts (37 – 42%) of the mountainous western U.S. These trends are strongly related to recent warming observed over 89% of the domain. The strongest changes in the hydrologic variables, unlikely to be associated with natural variability alone, have occurred at middle elevations (750 m to 2500 m for JFM runoff fractions and 500 m -- 3000 m for SWE/PONDJFM) where warming has pushed temperatures from slightly below to slightly above freezing. Further analyses focused on 66 river basins indicated that hydroclimatic variables must have changed significantly over at least 45% of total catchment areas to yield detectable trends in measures accumulated to the basin scale. The present study, with its fine scale grid, yields positive detections of change in hydrologic variables that could not be expected from natural variability in many sub-areas within the western U.S., and helps bring the results of previous regional-scale detection-and-attribution studies down to scales needed for water management, studies of ecosystem diversity, and anticipation of wildfires. If recent warming continues into future decades as projected by climate models, there will be serious implications for the hydrological cycle and water supplies of the western U.S.