H53A-1646
Non-stationary and Trend Assessment of Flash Flood in the Carolinas

Friday, 18 December 2015
Poster Hall (Moscone South)
S Samadi1, Thomas Martin2 and Michael Meadows1, (1)University of South Carolina, Columbia, SC, United States, (2)2. élève-ingénieur, École des Ponts ParisTech, Champs sur Marne, 77420, Île-de-France, France
Abstract:
Evidence that frequency of flash flood is increasing at the global scale has strengthened considerably in recent years. Our study provides insight into how changes in maximum precipitation and large-scale atmospheric circulations modulated the modes of regional scale short- and long-term flood magnitudes during last several decades (i.e.1920-2015) in North and South Carolina. The non-stationary univariate/bivariate extreme value analysis (EVA) models were fit to the short-term (5-day and 11-day) and long-term (monthly, seasonal and annual) maximum flows by incorporating linear/non-linear combination of covariates like maximum precipitation and large scale atmospheric circulations (the El Niño-Southern Oscillation (ENSO) teleconnection and Pacific Decadal Oscillation (PDO)). Eleven of 13 piedmont and the coastal plain gauges show upward trends in short-term (5-day and 11-day) flood magnitudes with changes in the location (mean), scale (standard deviation) and shape (skewness) parameters of the probability distribution. In addition, the trend and nonstationarity in annual flood magnitudes of mountain gauges are monotonically positive potentially leading to increased rainfall induced annual floods. Incorporating large scale covariates in the functionality of the EVA procedure indicated a lagged positive relationship between fall to early winter ENSO and PDO with winter (Jan- Mar) flood magnitudes and frequency. Many study gauges show evidence for step increases in flood frequency around 1940 and 1980 for long- and short-term periods of major catchments in the Carolinas. Our study further used copulas to model the bivariate dependence between ENSO and flood anomaly and the trivariate dependence between ENSO, PDO, and flood anomaly. Both the bivariate and trivariate copulas resulted with almost the same interquartile ranges and simultaneously predicted flood magnitude and frequency. Our results indicated that the greatest monotonic trends occurred in both short-and long-duration floods through a strong nonstationary relationship with maximum precipitation and large-scale circulation modes, potentially leading to an increase in the magnitudes and frequency of the diurnal to annual patterns of flash-floods in the Carolinas.