COS 149-5
Detection of climate extremes with functional data analytic approach

Friday, August 14, 2015: 9:20 AM
343, Baltimore Convention Center
Jien Zhang, Department of Earth & Environmental Sciences, Lehigh University, Bethlehem, PA
Ping-shi Wu, Department of Mathematics, Lehigh University, Bethlehem, PA
Background/Question/Methods

Climate extremes have important implications for ecosystem productivity, stability and service in a changing climate. An accurate detection of extreme event periods and locations is an important step of understanding the shifts of climate extremes. Current studies focusing on the trend of historical and future climate extremes are based on traditionally statistical index, which usually fail to detect the spatiotemporal patterns of extreme events. In this study, we applied a set of functional data analytic approaches to assess the spatiotemporal patterns both analytically and graphically in the extreme high and low temperatures as well as extreme precipitations across the state of California (CA). The functional boxplot is based on the center outward ordering that is induced by band depth for functional data. It provides an innovative way to visually evaluate the distributional status of climate profiles and to detect the climate extremes. One 31-year blocks (1948-1978) of daily maximum and minimum temperatures were first used for the annual analysis.

Results/Conclusions

Preliminary results show that, among 38 weather stations, two stations locating in southern CA and two stations locating in northern CA are detected as “outliers” in both maximum and minimum temperatures, which indicates that southern CA likely had hot extremes, while northern CA experienced cold extremes over those 31 years. Gridded daily temperature and precipitation datasets is used to better visualize extreme temperatures and precipitations over CA. Furthermore, effects of the detected climate extremes on modeled soil moisture is discussed to indicate potential changes to ecosystems. Despite the limitation of being sensitive to the annual variability, the functional data analytic approach is superior and robust when dealing with diverse temporal pattern of climate extremes in comparison with traditional methods.