Temperature modelling is a major issue for valuation of weather derivatives. Goodness of fit is usually assessed from historical data. However, estimation errors can result in large price uncertainty that may be problematic for practical applications. In this paper, we consider a temperature ARMA model and quantify the price uncertainties for weather Futures and weather options. Each price is seen as a random variable (which is a function of the parameters estimator), and we assess price uncertainty by giving confidence intervals. In addition, we look for sources of uncertainty, and point out the major defects of the model.