Due to dependency of energy demand on temperature, weather derivatives enable the effective hedging of temperature related fluctuations. However, temperature varies in space and time and therefore the contingent weather derivatives also vary. The spatial derivative price distribution involves a risk premium. We examine functional principal components of temperature variation for this spatial risk premium. We employ a pricing model for temperature derivatives based on dynamics modelled via a vectorial Ornstein-Uhlenbeck process with seasonal variation. We use an analytical expression for the risk premia depending on variation curves of temperature in the measurement period. The dependence is exploited by a functional principal component analysis of the curves. We compute risk premia on cumulative average temperature futures for locations traded on CME and fit to it a geographically weighted regression on functional principal component scores. It allows us to predict risk premia for nontraded locations and to adopt, on this basis, a hedging strategy, which we illustrate in the example of Leipzig.
risk premium weather derivatives Ornstein-Uhlenbeck process functional principal components geographically weighted regression