Changes in the characteristics of extreme cold anomaly for the period of 1951-2010 as revealed by a Markov analysis

Abstract

A long-term change in weather extremes seemingly occurs along the change of global/regional mean temperatures. Thus far, this notion is grounded by statistical records of occurrences of weather extremes alone. This study investigates the global changes of the extreme cold anomaly (ECA) with three different aspects, as presented by the so-called Markovian climate descriptors such as recurrence time, persistence, and entropy (or irregularity). Here the ECA is defined as the bottom two percentile of deseasonalized daily surface temperature records for each 2.5°-grid of NCEP/NCAR reanalysis I data for the period of 1951-2010. By the definition, the threshold of temperature for ECA turns out to be 0°C (for the tropical region) to -20°C (for the polar region). As results, the ECA is found to have experienced the increase in recurrence time (25 days per decade), the decrease in persistence (？0.15 days per decade), and the decrease in entropy (？0.01 per decade). These trends indicate that the ECA has occurred less frequently and more regularly, and lasted shorter. These changes have been more rapid over East and Southeast Asia, while the opposite changes are shown over the area around 30°S (i.e., Australia, South Africa, and the center of South America). Time series of the Markovian climate descriptors for each region demonstrate that the regional ECAs do vary on the decadal timescale. To ensure our results, the same analyses are carried out with two additional data: ERA-interim reanalysis data (1979？2010) and Hadley Centre quality-controlled gridded station data (HadGHCND) (1951？2010). The trends of the descriptors from the two additional data are similar to those from the NCEP/NCAR reanalysis data in North America. In the other regions, the time variations in the descriptors during the 2000s (1980s and 1990s) from ERA-interim reanalysis data agree (disagree) with those from the NCEP/NCAR reanalysis data. The present study suggests that the Markovian climate descriptors would serve to systematically cover important characteristics of the changes of weather extremes, which cannot be revealed by a single parameter like the frequency.