Several observational studies suggest that there are changes in the precipitation characteristics for different regions of the globe. These results are consistent with modeling studies that suggest a likely manifestation of a rise in global temperatures would be an enhancement of the hydrologic cycle. This enhancement of the global water cycle is expected to result in an increase in the rates of evaporation and precipitation as well as an increase in the frequency of extreme precipitation events.

Precipitation is produced by a complex combination of interacting processes operating over a wide range of space and time scales. It has significant diurnal, synoptic, seasonal, inter-annual and decadal scale variability that makes it difficult to identify and isolate trends in space and time. Another complication is introduced by changes in natural modes of atmospheric circulation associated with teleconnections such as ENSO and the resulting changes in precipitation patterns. Because an anthropogenic signal is expected to project onto natural modes of climate variability, it is necessary to quantify the fluctuations associated with natural modes of variability. This analysis needs to be performed on timescales that explain most of the space-time variance of precipitation, but it is not clear from previous studies which scale to use. We use gridded hourly and daily precipitation data to identify dominant scale of variability of precipitation and determine how it changes from season to season, year to year and location to location.

We use Multi Taper Method – Singular Value Decomposition on gridded hourly and daily precipitation records from the United States to identify coherent patterns of variability as a function of frequency. This technique combines a multi-taper spectrum estimate with a complex singular value decomposition to identify patterns of coherent, narrowband oscillations in a multivariate dataset. The analysis is performed on the entire 32 year period covered by the dataset and then for each season (year) to identify and quantify the dominant patterns of spatiotemporal variability over the United States and how they differ during El Nino, La Nina and neutral years.

The orthogonal discrete wavelet transform is then used to decompose the variance of each time series into components representing time scales ranging from hours to several days. A “scale series” representing the total energy from each scale is then constructed for each season and a rotated principal component analysis performed to identify coherent modes of scale-dependent variability across the United States and how they differ in El Nino, La Nina and neutral years. Trends in the scale-dependent variability are quantified and discussed.