This paper uses the tail pairwise dependence matrix (TPDM) proposed by Cooley and Thibaud (2019), which we extend to the description of common extremes in two variables. We develop an extreme pattern index (EPI), a pattern-based aggregation to describe spatially extended weather extremes. Our results show that the EPI is suitable for describing heat waves.  We extend the EPI to describe extremes in two variables and obtain an index to describe compound precipitation deficits and heat waves.

In this paper we consider the problem of detecting changes in the distribution of the annual maximum temperature, during the years 1950–2018, across Europe. We find that, on average, the temperature that would be expected to be exceeded approximately once every 100 years in the 1950 climate is expected to be exceeded once every 6 years in the 2018 climate. This is of concern due to the devastating effects on humans and natural systems that are caused by extreme temperatures.

Climate model output has systematic errors which can be reduced with statistical methods. We review existing bias-adjustment methods for climate data and discuss their skills and issues. We define three demands for the method and then evaluate them using real and artificially created daily temperature and precipitation data for Austria to show how biases can also be introduced with bias-adjustment methods themselves.

Most climate time series contain annual and diurnal cycles. However, an objective criterion for deciding whether two time series have statistically equivalent annual and diurnal cycles is lacking, particularly if the residual variability is serially correlated. Such a criterion would be helpful in deciding whether a new version of a climate model better simulates such cycles. This paper derives an objective rule for such decisions based on a rigorous statistical framework.

The objective motivating this study is the assessment of the impacts of winter climate extremes, which requires accurate simulation of snowfall. However, climate simulation models contain physical approximations, which result in biases that must be corrected using past data as a reference. We show how to exploit simulated temperature and precipitation to estimate snowfall from already bias-corrected variables, without requiring the elaboration of complex, multivariate bias adjustment techniques.