Abstract: How to accurately price the weather derivatives,which as a sort of weather risk-hedging financial derivatives, has always been the focus of academic debate. Based on the O-U model of monthly volatility presented by Alaton^[5], the speed of mean reversion is considered in this paper,which is a constant in the original model as a time-dependent variable. Then the time sequence of the speed of mean reversion is analyied by ARMA(p, q) model and establish the time-varying O-U model. Based on Beijing's daily average temperature data from 1951, three years of daily average temperature from 2010 to 2012 are simulated respectirely, and compared with its true value. After the comparison,it is found that: the residual sum of squares of the improved model are smaller, meanwhile the bias proportion, the variance proportion and the covariance proportion also show the improved model obtains a better temperature prediction. Finally, the CDDs are calculated and HDDs are calculated according to Beijing's daily average temperature data by Monte Carlo simulation, and then price the related futures contract to further validate the adaptability of improved model.

Key words: weather derivatives, O-U model, speed of mean reversion, Monte Carlo simulation