Nonstationarity of summer temperature extremes in Texas
Nonstationarity of summer temperature extremes in Texas
复制标题
DOI：
10.1002/joc.6212
复制
发表时间：
2019-07
期刊：
影响因子：
--
通讯作者：
Meagan Carney;R. Azencott;M. Nicol
中科院分区：
文献类型：
--
作者：
Meagan Carney;R. Azencott;M. Nicol
在天气模式中模拟季节性极端温度可以更好地预测和预报。给定时间段内的极值分析通常通过将广义极值（GEV）分布拟合到数据中的最大值来完成;然而，由于缺少数据或违反独立性假设（块最大值应该在很长的时间间隔内）而缺乏足够的天气记录可能导致GEV模型拟合不佳。在更大的集群区域进行建模可以克服其中一些问题，并可以深入了解宏观天气和气候变化。在这篇文章中，我们分析了7月和8月从德克萨斯州和路易斯安那州新奥尔良的气象站获得的温度测量结果。我们引入聚类技术，分组站的温度趋势和互信息进行极值分析的集群的时间序列。这样就避免了分析单站天气数据时通常遇到的一些问题。极端分析所产生的集群提供了令人信服的证据的非平稳性的分布参数的GEV模型，并指出，从1980年左右的时间到目前为止，观察到更高的极端温度的7月和8月的可能性增加。我们根据一个非平稳的模型，在集群中的极端温度的概率列表。我们的技术可以很容易地适应各种气候问题。
Modelling seasonal temperature extremes in weather patterns allows for better forecasting and prediction. Analysis of extreme values over a given time period is usually done by fitting a generalized extreme value (GEV) distribution to the maximum values in the data; however, lack of sufficient weather recordings due to missing data or violation of independence assumptions (block maxima should be over a large time interval) may result in a poor fit for the GEV model. Modelling over larger, clustered regions may overcome some of these problems and can provide insight into macroscopic weather and climate changes. In this article, we analyse temperature measurements in July and August taken from stations across Texas and New Orleans, Louisiana. We introduce clustering techniques which group stations by temperature trends and mutual information before performing extreme value analysis on the clusters of time series. This obviates some of the problems commonly encountered in analysing single station weather data. Extreme analysis of the resulting clusters provides compelling evidence of nonstationarity of the distributional parameters in the GEV model and points to an increased likelihood from the period roughly 1980 to present of observing higher extreme temperatures for the months of July and August. We tabulate the probabilities of extreme temperatures in the clusters according to a nonstationary model. Our techniques can be easily adapted to a wide range of climatological problems.