A model-combination approach to pricing for weather derivatives

Abstract

The main object of this study is to find a model that can explain and predict the behavior of weather time series. Different models of GARCH-type are fitted on the series of the average temperature. Autoregressive, seasonal and trend components are included in mean and variance equations. The seasonal components are constructed as a combination of simple harmonic functions, the frequencies of which are found with Fast Fourier Transform. Forecast combination approach is applied to define the best model and thus this model is used as an underlying to price weather derivatives. Numerical examples of pricing some exotic options are given, found with Monte Carlo simulations.