21 – Spatial Modeling to Identify Relationships and Trends
Comparison of Time and Spatial Scales in Global Temperature Data
Ming Luo
SUNY at Albany
Igor Zurbenko
SUNY at Albany
To capture the important spatio-temporal patterns in a global temperature dataset, we applied correlation analysis on recovered signals based on a nonparametric smoothing algorithm, the Kolmogorov-Zurbenko Spline (KZS). In our study, KZS was applied on monthly average temperature observations from thousand of weather stations over the global to generate smoothed monthly temperature deviations from normal levels over the time period of 1893 to 2008. These signals, which were separated out on different time and spatial scales, can be viewed as global long term climate trend and El Nino-like movement. Then the correlation relationships of the reconstructed signals were studied with bootstrapping. We found this method was decisive for understanding the key features of the data, and it revealed an interesting corresponding relation between time and geographical distance. A movie for global maps of long term temperature trend has been developed as well as El Nino scales over the global, and both provide striking revelations in global temperature anomalies for specific regions.