Low frequency variability in fall precipitation across the United States

Precipitation and streamflow have increased across the United States during the 20th century, with the largest increases observed in the fall. In this study, we apply the multitaper method with singular value decomposition (MTM‐SVD) multivariate spectral analysis to characterize the space‐time variability of fall precipitation over the United States from 1895 to 2004. MTM‐SVD analysis of precipitation demonstrates the presence of statistically significant secular and decadal signals. The decadal signal in fall precipitation, with a period of approximately 12 years, is large and spatially correlated across most of the central United States. The secular signal, with a period greater than 55 years, is weaker, much less spatially correlated and is strongest in the eastern United States. The decadal and secular variations increased in magnitude after 1950, the period when many previous studies have identified large linear trends. Linear trends in fall precipitation are found to be more spatially consistent with the pattern of decadal variations than the secular variations, suggesting that increasing decadal variations in precipitation after 1950 have contributed to large linear trends reported by several previous studies. Our results suggest that the strong spatial correlation in decadal precipitation variations across the central United States are associated with a large SLP gradient across North America that is conducive for southerly flow from the Gulf of Mexico. Secular variations in SLP associated with precipitation over the United States are located primarily in the high latitudes, and are not consistent with correlated variations in precipitation across the central United States.