| This dissertation is concerned with spatio-temporal processes in the atmospheric sciences.; In the first chapter, a comprehensive overview of spatio-temporal methods from the atmospheric science literature is presented. Focus is on Empirical Orthogonal Function (EOF), Principal Interaction Pattern (PIP), Principal Oscillation Pattern (POP), and spatio-temporal Canonical Correlation Analysis (CCA) methods. Previously unexamined issues related to measurement error, continuous space, and Bayesian ideas are considered.; In the second chapter, harmonic analysis is used to make diagnostic inference about the spatial variation of the semiannual oscillation (SAO) in the Northern Hemisphere (NH) 500-hPa height field. The SAO is explained by the spatial and temporal asymmetries in the annual variation of stationary eddies. The SAO in the NH extratropics is a result of east-west land-sea contrasts, analogous to the well-known Southern Hemisphere (SH) SAO, which is explained by north-south land-sea contrasts.; The third chapter examines the seasonal variability of mixed Rossby-gravity waves (MRGWs) in the lower stratosphere over the tropical western Pacific. Thirty-one years of lower stratospheric wind observations from four tropical Pacific stations are examined with seasonally varying cross-spectral analysis, which suggests significant twice-yearly peaks in the v-wind power and the mean squared coherence between the u- and v-winds, with peaks occurring in the winter-early spring and in summer-early fall. Horizontal momentum flux convergence is found with these waves, with the sign of the convergence opposite during the two seasonal maxima. Cyclic spectral analyses show that the frequency of the maximum v-wind power in the MRGW frequency band shifts seasonally.; In the fourth chapter, a spatio-temporal statistical model is proposed that assumes a first-order Markov dynamic process combined with a spatially descriptive colored noise process. With a measurement error equation, a spatio-temporal Kalman filter gives predictions in time and at any spatial location. The model prediction equation includes a simple kriging analog as a special case. The model predicts well with simulated spatio-temporal data, and is superior to simple kriging applied independently at each time. Predictions of precipitation over the data-sparse South China Sea captures the dynamic variation of the spatial precipitation. |
