Department of Atmospheric SciencesDepartment of Atmospheric Sciences

University of WashingtonUniversity of Washington

DDD RRR... GGG AAA BBB RRR III EEE LLL EEE VVV III LLL LLL AAA RRR III NNN III AAA SSS SSS OOO CCC III AAA TTT EEE RRR EEE SSS EEE AAA RRR CCC HHH SSS CCC HHH OOO LLL AAA RRR

DDD EEE PPP AAA RRR TTT MMM EEE NNN TTT OOO FFF CCC III VVV III LLL AAA NNN DDD EEE NNN VVV III RRR OOO NNN MMM EEE NNN TTT AAA LLL EEE NNN GGG III NNN EEE EEE RRR III NNN GGG PPP RRR III NNN CCC EEE TTT OOO NNN UUU NNN III VVV EEE RRR SSS III TTT YYY

“A DATA-DRIVEN PERSPECTIVE

ON FLOODING AND CHANGING CLIMATE”

Date: Tuesday, February 28, 2012 Time: 3:30 p.m. Location: Bagley Hall, Room 154

Abstract: Every year, flooding claims a large toll both in terms of fatalities and economic damage. Under projected human-induced climate warming, modeling results suggest an acceleration of the hydrological cycle, with potentially large effects on the occurrence of extreme events, like flooding. Detailed analyses of the historical records are of paramount importance to improve our understanding of possible future changes in the flood peak distributions. Whether flood magnitude and frequency are bound to change in the future has been the subject of vigorous debate, even though a consensus has yet to be reached. Possible reasons for the ambiguous nature of these changes, and more generally of the validity of the stationarity assumption, are the physical processes at play, sample properties and the methods used for the statistical inference. In this presentation, the validity of the stationarity assumption in flood peak records is examined by analyzing annual peak discharge time series from numerous stations with long records in the continental US and central Europe. Over these areas and during the twentieth century, large human modifications of these catchments (e.g., construction of dams, changes in land use / land cover) represent the norm, and the analyses should reflect and account for these anthropogenic changes. Violation of the stationarity assumption can be caused by abrupt and slowly varying changes. Abrupt changes in the mean and variance of the flood peak distributions are examined through change-point analyses using the non-parametric Pettitt test, while the presence of monotonic patterns is examined through the non-parametric Mann-Kendall and Spearman tests. Abrupt changes, rather than monotonic trends, are typically responsible for non-stationarities in annual flood peak records in the continental US and central Europe. The detected change-points are often linked to regulation of river basins, changes in land use-land cover, and river engineering. These analyses indicate that it is difficult to detect a climate change signal in the flood peak records. Changing flood frequency in urban watersheds is the topic of focused analyses. Urbanization has a large impact on flood hydrology by decreasing infiltration and increasing runoff, resulting in changes in both flood peak magnitude and variability. Nonstationary statistical modeling of these flood peak records is presented, together with a discussion about changes in return period under nonstationary conditions. Given the profound anthropogenic changes that these catchments have undergone (e.g., construction of dams, changes in land use / land cover, river engineering), an alternative way of examining the validity of the stationarity assumption for these time series is by describing the flood peak records in terms of mixtures of different flood generating mechanisms. Results concerning the eastern US are presented, showing the central role played by landfalling tropical cyclones and extratropical systems. Analyses of long-term trends in counts of landfalling tropical cyclones for the eastern US are presented, together with results concerning changes in the frequency of tropical cyclones under different climate change scenarios and for different climate models.