Activity Number:
|
619
- Spatial and Spatial-Temporal Statistics
|
Type:
|
Topic Contributed
|
Date/Time:
|
Thursday, August 2, 2018 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Statistics and the Environment
|
Abstract #327225
|
|
Title:
|
Spatial Statistics for Improving Collective Estimates of Extreme Precipitation at Weather Stations
|
Author(s):
|
Mark Risser* and Christopher Paciorek and Michael F Wehner
|
Companies:
|
Lawrence Berkeley National Laboratory and University of California, Berkeley and Lawrence Berkeley National Laboratory
|
Keywords:
|
extreme value analysis;
spatial extremes;
nonstationary;
bootstrap;
precipitation
|
Abstract:
|
While there are well-established univariate statistical methods for modeling the climatological properties of extreme precipitation, characterizing the heterogeneous properties of multivariate extremes over space for a large geographic area is a much more difficult task. This presentation describes the application of spatial statistical analysis tools that incorporate the heterogeneous spatial dependence patterns present in the climatological features of extreme precipitation. These tools scale to accommodate large spatial data sets, and we show that the borrowing of strength over space can improve the signal-to-noise ratio in estimates of the climatological features of extremes (e.g., r-year return values). Furthermore, our approach accounts for spatial dependence in daily precipitation due to individual storm systems and characterizes changes over time in extreme statistics. Uncertainty estimates are obtained via a block bootstrap technique that requires no assumption of temporal independence within each year of data. We illustrate our approach via seasonal analyses of daily measurements of precipitation from the GHCN stations over the continental United States.
|
Authors who are presenting talks have a * after their name.