Monday, 11 January 2016: 4:00 PM
Room 242 ( New Orleans Ernest N. Morial Convention Center)
Accurate characterization of extreme rainfall, as extreme hydro-meteorological events, is of utmost importance for a variety of scientific and societal interests. Precipitation Frequency Estimates (PFE) at different return periods are often used by engineers, water managers and decision makers for design of major flood protection projects. Recent interest has been expressed to characterize the PFE's over an areal domain (e.g., a basin or catchment) as opposed to point-based estimates. Such shift in focus requires accurate description of the spatial dependence in extreme rainfall fields. While methods of statistical modeling of univariate extremes are well developed, extending such methods to spatial extreme fields remains as an active area of research. Tools for describing spatial dependence have been extensively developed in the field of geostatistics (e.g. variogram and correlation functions); however, such tools may not be applicable in the context of modeling of spatial extremes. The current study presents an analysis of the spatial dependencies in Annual Maximum precipitation Series (AMS) over the southeast United States using non-parametric measures that were recently developed in the field of extreme value theory (EVT). Examples of these measures are the extremal coefficient and the madogram, which offer a simple and efficient connection between EVT and geostatistics. The current analysis is performed using extreme rainfall data extracted from two sources: (1) spatially-distributed radar-based multi-sensor precipitation estimates (MPE), which provide continuous coverage in space but with fairly short records, and (2) point-based gauge observations, which provide much longer observational records. For comparison purposes, the spatial dependence is also characterized using classical measures such as correlation functions. The study also examines the change in the extremal dependence as a function of several factors besides distance (e.g., rainfall duration, dry-versus-wet seasons, and other relevant regional/climatic characteristics).
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner