Application of Adaptive Estimation to Temperature Forecasting.
TEXAS UNIV AUSTIN ELECTRONICS RESEARCH CENTER
Pagination or Media Count:
The work is an application of adaptive estimation to temperature forecasting. It is presented as a feasibility study demonstrating the efficacy of the adaptive approach. The local station temperature forecasting problem is chosen to focus the discussion on the efficiency of the filtering algorithm by using only surface level single geographic location data. A diagnostic study is made to ascertain the appropriate statistical properties of the weather data for algorithm selection. A phenomenalistic approach is taken since no differential equation or complete quantitative description exists to describe the temperature process. The Lainiotis Filter is chosen for model identification and classification as well as prediction results. The Lainiotis Filter, given in the Partition Theorem, provides an efficient, powerful tool in the application of adaptive estimation techniques. The feasibility of the adaptive approach is established with comparative results with previous objective forecast methods while greatly reducing the amount and variety of required input data. Author