Local summer temperature dynamics: Bayesian Markov-switching to forecast annual frequency and duration of heat waves
提出一种动态隐马尔可夫模型，结合自然季节性、大尺度气候指数和尾部建模，利用当地最高温度数据预测热浪的年发生频率和持续时间，并以意大利两个地理位置接近但气候异质的地区为例进行说明。
Summer temperatures may exhibit prolonged periods of extremely high temperatures, informally referred to as heat waves. Operative definitions entail temperature abnormality, persistence over time, and, possibly, impact on ecosystems and human life. Historical episodes and stochastic patterns of heat waves can be described by adopting a specific definition using observed temperature data. Atmospheric and local conditions may affect the phenomenon’s variability, complicating the study and prediction of temperature dynamics and heat waves. Hence, the proposal of a dynamic hidden Markov model that incorporates natural seasonality, a large-scale climate index, and tail modelling to explore the annual distribution of heat wave metrics. Various classification methods for describing and forecasting heat wave periods are considered. The proposal is illustrated using the local maximum daily temperatures of two geographically close and heterogeneous locations of the Italian region Friuli Venezia Giulia .