In this chapter and the next we will consider how weather forecasting and seasonal forecasting can potentially improve our valuation of weather derivative contracts. This chapter starts with a discussion of which weather forecasts are relevant to weather derivatives, and how these forecasts are made. We discuss forecasts of the expected temperature first, and forecasts of the whole distribution of future temperatures (probabilistic forecasts) after that. In each case we describe how to compare two forecasts to find out which is better, and briefly mention statistical methods with which one can try to improve forecasts. In the second part of this chapter we give a brief description of seasonal forecasts.

It is interesting to note that weather is somewhat different from financial market prices in the extent to which it is predictable. Forecasts of changes in financial market prices are possible, and may be successful, but there is a feedback between the forecast and the price, which means that, over time, all forecasting systems are likely to fail. Weather, on the other hand, is not affected by weather forecasts, and the dynamics of weather is constant. This allows forecast systems to be constructed that will continue to make useful predictions now and into the future. There is, however, a parallel between the unpredictability of changes in market prices and weather in that changes in expected weather are unpredictable.