A comparison of surface and free-air temperature variability and trends at radiosonde sites and nearby high elevation surface stations



Previous research has illustrated differences in temperature trends as measured by high elevation surface stations versus free-air temperatures interpolated to the same locations from the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research's (NCAR) Re-analysis R1. This paper examines the extent to which the decision to use R1 rather than radiosonde data influenced these results. Temperatures at selected high elevation surface sites (1948–998) are paired with nearby homogeneity adjusted radiosonde data from Lanzante, Klein and Seidel (LKS). For each station pair, four mean monthly temperature anomaly time series are examined, consisting of (1) surface Global Historical Climate Network (GHCNv2)/Climate Research Unit (CRUv2) station anomalies (SF), (2) LKS radiosonde anomalies, and R1 anomalies interpolated to (3) the surface (RNSF) and (4) radiosonde (RNLKS) locations respectively. Analyses demonstrate the extent of common variance, the mean climatology of each of the four series, and differences or similarities in trends. The surface record is decoupled from the other three series especially in locations of incised topography. In 15 out of 18 pairs RNSF shows greater affinity with LKS than with SF and there is a high degree of common variance between LKS and RNSF. There is a high degree of correlation between secular trends in the two R1 series, both of which are much more similar to radiosonde than to surface trends. Trends in raw temperature and in ΔT (the surface/free-air temperature difference) both therefore show limited sensitivity to the choice of LKS radiosonde versus R1 (RNSF), apart from in a few locations in the Eurasian continent. Copyright © 2007 Royal Meteorological Society