The Leading Patterns of Climate Variability and Winter Weather Extremes in the United States Part I: Temperature

An emerging area of research is to understand how the natural modes of climate variability might account for changes in extreme weather events and to develop new methodologies to forecast these events based on short-term climate prediction. This approach offers an opportunity to anticipate such events with significantly more advanced warning than has been possible in the past.

Understanding changes in climate variability and extreme weather events is made difficult by interactions between these two characteristics of climate. In particular, climate variations cause changes in the atmospheric circulation patterns. These changes, in turn, cause shifts in the probability distributions of weather events, e.g. shifts towards warmer/colder/wetter/drier conditions, by changing the mean of the distribution and / or its shape (i.e. variance). Both effects magnify the changes in the probability of extreme events on the tails of the distributions.

It is difficult to assess the characteristics of extreme weather events for different climate states because the mean conditions and extreme events during a season are the result of contributions from a number of factors, which can act independently, and which may not be predictable. Possible contributors include El Niño-Southern Oscillation (ENSO), the Arctic Oscillation (AO), the Pacific Decadal Oscillation (PDO), random weather events, and the impacts of long term trends related to global climate change. One way to approach this problem is to isolate the contributions of the natural modes, which are linked to changes in the large-scale circulation of the atmosphere.

In this study this approach is used to examine relationships between the natural modes and changes in the distributions of surface air temperature and precipitation over the U.S.. The study is based on 50-years (JFM 1950-1999) of daily-mean surface air temperature and daily accumulations of precipitation over the conterminous United States. Extreme events are defined as those ranked in the top (bottom) 10% of the appropriate distribution. Changes in the number of extreme events over the last 50 years are examined locally, regionally and for the U.S. as a whole. Relationships between numbers of extreme events and the natural modes are established using composite analysis together with standardized ENSO, PDO and AO indices. The analysis is confined to boreal winter since this is the time of year when the natural modes exert the strongest influence upon temperature and precipitation patterns. The temperature and precipitation time series are also detrended, and class limits are fixed, in order to clarify relationships between the patterns of extreme weather events and anomaly patterns associated with the natural modes. The results are used to highlight some of the challenges that arise in probabilistic forecasting of extreme events.