Seasonal Space Time Models for Climate Systems

A class of seasonal space-time models for general lattice systems is proposed. Co-variance properties of spatial rst-order models are studied. Estimation approaches in time series analysis are adopted and forecasting techniques using the seasonal space-time models are discussed. The models are applied to 516 consecutive elds of monthly-averaged 500 mb geopotential heights over a 1010 lattice in the extra-tropical Northern Hemisphere for the purpose of understanding the underlying statistical structure. It is found that space-time models with instantaneous spatial component give the best t compared to other models in terms of maximizing the conditional likelihood function. The models are potentially useful for assessing the consistency of outputs from laboratory-based numerical models with eld observations. Forecasting ability of the seasonal space-time models is also investigated.