Analysis of the Merging Procedure for the MSU Daily Temperature Time Series

Abstract

Abstract The merging procedure utilized to generate homogeneous time series of three deep-layer atmospheric temperature products from the nine microwave sounding units (MSUs) is described. A critically important aspect in the process is determining and removing the bias each instrument possesses relative to a common base (here being NOAA-6). Special attention is given to the lower-tropospheric layer and the calculation of the bias of the NOAA-9 MSU and its rather considerable impact on the trend of the overall time series. We show that the bias is best calculated by a direct comparison between NOAA-6 and NOAA-9, though there other possible methods available, and is determined to be +0.50°C. Spurious variations of individual MSUs due to orbital drift and/or cyclic variations tied to the annual cycle are also identified and eliminated. In general, intersatellite biases for the three instruments that form the backbone of the time series (MSUs on NOAA-6, -10 and -12) are known to within 0.01°C. After slight modifications in the treatment of the bias, drift-error, and cyclic fluctuations, the authors produced a time series in which the decadal trend is +0.03°C warmer than previously reported for the lower troposphere. Because they are of much higher precision, the midtropospheric and lower-stratospheric products are only slightly affected by alterations to procedures applied in this study. Recent suggestions that spurious jumps were present in the lower-tropospheric time series of earlier versions of the MSU data based on SST comparisons are addressed. Using independent comparisons of different satellites, radiosondes, and night marine air temperatures, no indication is found of the presence of these “spurious” jumps.