A comparison of regional trends in 1979-1997 depth-averaged tropospheric temperatures

Abstract

This study examines regional temperature trends during the period 1979–1997 from the Microwave Sounding Unit (MSU) 2r satellite measurements and compares them with the same trends in depth-averaged tropospheric temperatures derived from the National Center for Environmental Prediction (NCEP) reanalysis, in an attempt to determine whether regional trends exist which are larger than known inhomogeneities in the data. Large, statistically significant regional trends were found in both the NCEP and the MSU data that are of both signs and have larger magnitude than documented biases in the data. The datasets have overall agreement on the location and strength of these significant regional trends at mid and high latitudes but agreement decreases in the tropics. A global annual average of the significant regional trends with larger amplitudes than reported data biases and areally weighted over the globe yields −0.02°C over the 19-year period of the record in the MSU 2r Version C dataset, and −0.05°C/19 years in the NCEP data in the 1000–500 mb layer. Increasing the bias threshold by as much as five times still results in an average cooling in both datasets. Subjecting the surface temperature record to the same regional analysis yields a regionally significant trend of 0.17°C/19 years, approximately halving the trend obtained when all regions, regardless of significance, are considered. In addition, many regions with significant warming trends in the surface network occur in areas with limited observations over oceans and are not confirmed by the other datasets. Discrepancies between significant regional trends in the surface record and the upper-air observations are not systematic. In no case are regionally significant, tropical, warming trends at the surface magnified at higher levels in the MSU and NCEP tropospheric data. In the case of the NCEP reanalysis, both warming and cooling trends on average become larger, more significant, and cover larger areas in shallower tropospheric layers. These results suggest that the disparity between global trends in satellite/rawinsonde/reanalysis datasets and those of the surface record are not simply the result of large-scale changes in the vertical structure of the atmosphere or to large-scale biases in the satellite observations, but instead are linked to processes which are regional in nature. Copyright © 2000 Royal Meteorological Society