It’s Normal: The Probability Distribution of Temperature Extremes

The probability of heat extremes is often estimated using the nonstationary generalized extreme value GEV distribution (GEVD) applied to time series of annual maximum temperature. Here, this practice is assessed using a global sample of temperature time series, from reanalysis (both at the grid point and at the region scale) as well as station observations. It is found that the computationally simpler normal distribution outperforms the GEVD for providing probabilistic year-ahead forecasts of temperature extremes using as a measure forecast negative log likelihood, which is particularly sensitive to the most extreme heatwaves. It is therefore suggested to consider alternatives to the GEVD for assessing the risk of extreme heat.