Stochastic modeling of hot weather spells and their characteristics

Climate Research

L. Sushama¹  T. B. M. J. Ouarda¹  P. Gachon¹  M. N. Khaliq¹ 

关键词: Climate change;    Compound Poisson process;    Hot weather spells;    Logarithmic ­distribution;    Negative binomial distribution;    Poisson-gamma model;   

DOI  :  10.3354/cr01003

来源: Inter-Research Science Publishing

【 摘 要 】

ABSTRACT: Stochastic characterization of hot weather events (HWEs) is useful for developing probabilistic climate change information. We assumed that simultaneous exceedances of daily minimum and maximum temperatures (i.e. T_min and T_max, respectively), above the selected thresholds for these temperatures, form alternating sequences of `hot weather´ and `non-hot weather´ spells. This behavior is stochastically characterized by a generalization of the compound Poisson process. The generalization is achieved by randomizing the mean rate of occurrence of the original Poisson process with the gamma distribution. The resulting compound Poisson process gamma (CPPG) model consists of 2 basic components: one relates to the occurrence of HWEs and the other to their durations. Detailed validation of the CPPG model is presented for HWEs derived from homogenized T_min and T_max observations from McTavish station (located in the centre of Montreal, southern Quebec, Canada) for the June-August hot summer season over the period 1941-2000. Results of the study suggest that the CPPG model can adequately describe occurrences of HWEs and the distributions of their durations (including distributions of extreme durations), as well as the distributions of the number of hot days, when assessed on the basis of the Kolmogorov-Smirnov and chi-squared goodness-of-fit tests. A non-stationary framework is introduced and applied to 2 contrasting sets of non-stationary observations of HWEs from Les Cedres and La Tuque (southern Quebec) for the 1941-2000 period. The results of the non-stationary modeling approach demonstrate that the developed methodology is generally applicable and could be useful in developing probabilistic information about various ­characteristics of hot weather climate for a particular location or a larger spatial domain.

【 授权许可】

Unknown