Nonlinear characteristics of daily temperature data and nonlinear modeling

We present a multifaceted analysis of west Tennessee temperature data using nonlinear dynamical tools and to investigate the extent to which the temperatures are predictable. The data set that will be used in this analysis consists of daily temperatures recorded at several stations in west Tennessee. Such nonlinear characteristics as the test of nonlinear dependence (due to Savit and Green), BDS test (Brock, Dechert, and Scheinkman), Lyapunov exponents, and correlation dimension are applied to daily temperatures with an eye toward improving the forecastibility. Our preliminary results represent a comprehensive approach to exploiting forecastability and include an example of forecasting daily temperatures using a state-space prediction method. As a byproduct of this research, we propose to apply nonlinear dynamical tools to other climatology data to contrast and compare regional weather characteristics.