On long period trend estimates of upper-air extreme and sub-extreme temperatures by use of quantile regression

Abstract

Abstract The traditional studies of climate change are based on estimation of trends in average values of climate variables (such as monthly, seasonal, or annual average values of temperature, air pressure, precipitation, wind speed, etc.). However, these estimates do not provide detailed information on changes in the distributions and are not sufficient to answer questions about changes in extreme and sub-extreme values, as well as questions on left and right “tails” of the distributions and changes in the measures of variability. A mechanism called quantile regression (QR) is an instrument for performing such a study. Unlike the traditional regression methods that are based on the Ordinary Least Squares (OLS) methodology, QR can provide detailed information about the structure of climate trends for a whole range of values, i.e. including extremal and sub-extremal values. While the known climatological studies of QR trends in extremes and sub-extremes are related to surface meteorological variables, such as surface temperature, it is essential to make similar QR-based studies of trends for the upper-air (UA) temperature. This paper contains results of such an analysis for the UA temperature based on a collection of radiosonde data for more than 30 years of observation. We discuss typical patterns of a detailed structure of climate trends for the upper-air temperature in the troposphere and in the lower stratosphere for certain geographic sectors of the Northern Hemisphere. The existing difficulties to obtain realistic results in such a study are shown to be related to various kinds of gaps in the radiosonde data. Therefore, some data quality requirements are vital for such upper-air studies.