How spatially coherent and statistically robust are temporal changes in extreme precipitation in the contiguous USA?

Century-long precipitation records from stations in the contiguous USA indicate an increased frequency of rainy days over the past century and some evolution in the probability distributions of precipitation amount. Temporal trends in eight metrics of the precipitation climate are of similar magnitude and sign regardless of whether they are derived from bootstrapping of regression residuals or using the Kendall's tau statistic, though the bootstrap technique generally resolved a larger number of ‘significant’ trends. There is substantial variability in terms of the magnitude, significance and sign of the linear trends with specific metrics, and they are sensitive to recording bias of light precipitation events in the early part of the 20th century. The majority of stations that exhibit significant linear trends show evidence of increases in the intensity of events above the 95th percentile. The resolved trends tend to have a larger magnitude at the end of the century. Spatial variability as manifest in the spatial autocorrelation in the interannual variability and trends in extreme metrics is manifest at a range of scales, but in general the correlation between stations is significant only for separation of distances of a few tens of kilometres. The largest trends towards increased annual total precipitation, number of rainy days and intense precipitation (e.g. fraction of precipitation derived from events in excess of the 90th percentile value) are focussed on the Central Plains/northwestern Midwest. Copyright © 2008 Royal Meteorological Society