From Diversity to Volatility: Probability of Daily Precipitation Extremes

Standard statistical approaches to modeling daily precipitation cannot account for observed volatility and typically underestimate the probability of extreme events. This can result in dangerously incorrect risk assessment for practical applications as well as hinder climatic studies of extreme meteorological events. Stochastic theory is used to quantify volatility of daily precipitation amounts at hundreds of weather stations across North America and to identify a mathematically and physically consistent model for the probability of daily precipitation extremes. Results indicate that, characteristically, precipitation volatility can be explained by heavy-tailed probability models, but not by the customary distributions with exponential tails. Moreover, by examining geographical structures in extreme precipitation behavior, we show that the degree of volatility observed at specific locations is determined by the diversity in precipitation-producing mechanisms. These results represent a fundamental development in stochastic modeling of precipitation extremes in climate. They also have immediate and vital practical implications, especially in view of the fact that climate is changeable and changing.