| || |
Submitted by pscully on Fri, 01/21/2011 - 09:51.
01/27/2011 - 16:10
01/27/2011 - 17:30
STA/BST 290: Richard Grotjahn (UC Davis)
Some extreme statistics applied to a scheme that downscales upper air data to identify California Central Valley hottest days
Thursday, January 27th, 2011 at 4.10pm, MSB 1147 (Colloquium room)
Refreshments: 3.30pm, MSB 4110 (Statistics Lounge)
Speaker: Richard Grotjahn
(Dept Land, Air and Water Resources Atmospheric Science Program, UC Davis)
Title: Some extreme statistics applied to a scheme that downscales upper air data to identify California Central Valley hottest days
Abstract: Extreme hot days as defined for the California central valley (CV) have a characteristic large scale structure that affects much of the west coast of North America as well. The patterns for this type of extreme event (EE) will be briefly summarized, including statistically highly significant regions far from California. Simple statistics (bootstrap resampling and simple tail tests) for assigning significance to parts of the EE patterns will be summarized.
The EE patterns for the hottest 1% of summer days have been compared to daily weather maps in a simple calculation of a daily anomaly 'circulation index' to hindcast (and forecast) hottest days in daily anomaly values. The 'circulation index' captures ~1/2 of the hottest 1% days from a 25 year period using a peaks over a threshold test to find the highest 1% of circulation index and CV observed temperature. The circulation index has high correlation with maximum temperatures on near-normal and below-normal dates as well (overall correlation between daily maximum temperature at central valley stations and HDA index: 0.84) but higher correlation does not guarantee better skill in capturing the rare hottest events. Some simple statistical quantities for rare events will be shown including POD, FAR, CSI, and EDS scores as well as parameters for a GPD fit.
Remaining questions include how the large scale patterns develop dynamically and improving extreme statistics for an imperfect predictor and imperfect scheme.