Estimating Monotonic and Cyclic Trends in Extreme Rainfall over the Northeast United States Using Hierarchical Bayesian Regression

Several observational-based studies indicate that the Northeast United States has experienced noticeable changes in intense precipitation events over the last several decades. Projecting whether these trends may continue, reverse, or weaken into the future first requires building an understanding of the mechanisms underlying these trends (i.e. whether these trends are primarily manifestations of anthropogenic warming or multi-decadal natural variability). This paper investigates how extreme rainfall in the Northeast United States has varied spatially and temporally using a model that estimates monotonic and cyclic trends of 7-day extreme rainfall as a function of climate predictors. We present a hierarchical Bayesian regression model to estimate the secular trend and the effects of climate predictors on the annual 7-day extreme rainfall at 52 rain gauge stations over the period of 1948-2013. In this multilevel structure model, the sensitivity of the extreme rainfall to the climate indices is informed with locational characteristics of each station. Results indicate that over 34% of the stations, the trend can be explained by El Niño–Southern Oscillation, Summer North Atlantic Oscillation, Atlantic Multidecadal Oscillation and Pacific North American teleconnection. 42% of the stations still exhibit monotonic trend even after including the climate effects. This modeling approach is introduced as an alternative to the regional climate trend analysis that involves clustering sufficiently homogenous sub-regions.