Previous - 3.3.1 Bias Adjustments 1850 to 1941 Index Next - 3.3.3 Estimating Uncertainty in Bias Adjustments
In the post-1941 period, Folland and Parker [1995], Smith and Reynolds [2003], Smith and Reynolds [2005] and Rayner et al. [2006] opted not to adjust the data because they found no clear evidence of the need for adjustments. However, Rayner et al. [2006] did identify biases in Japanese and Dutch data after the Second World War that were as large as the biases before the war. Thompson et al. [2008] identified a discontinuity in global-average SST associated with a change in the composition of ICOADS release 2.1 in late 1945. Reynolds et al. [2010] quantified a relative bias between ship and drifting buoy measurements that they thought could lead to an artificial cooling of the global average SST. Kent et al. [1999] applied adjustments to ERI measurements, but removed the adjustment from later versions of their data set.
Kennedy et al. [2011c], Hirahara et al. [2013] and Huang et al. [2015, 2016] developed bias adjustments for the period 1941 onwards. Kennedy et al. [2011c] used metadata from ICOADS, WMO Publication 47, observer instructions, technical reports and scientific papers to estimate biases for individual measurement types and to assign a measurement method to as many observations as possible. Hirahara et al. [2013] used a narrower range of metadata. By comparing subsamples of the data for which the metadata were known, they could estimate appropriate metadata assignments for the remainder. Huang et al. [2015, 2017] extended the method of Smith and Reynolds [2002] to the whole SST record. Uncertainties were estimated in Liu et al. [2015].
To estimate the bias adjustments for long-term analyses, Kennedy et al. [2011] and Hirahara et al. [2013] require an understanding of how biases varied for individual components of the observing system. Several studies have examined ERI and bucket biases in ship data [Brooks, 1926; Brooks, 1928; Lumby, 1927; Collins et al., 1975; Wahl, 1948; Roll, 1951; Kirk and Gordon, 1952; Amot, 1954; Perlroth, 1962; Saur, 1963; Walden, 1966; Knudsen, 1966; Tauber, 1969; James and Fox, 1972; Tabata, 1978a, 1978b; Folland et al., 1993; Kent et al., 1993] but only Kent and Kaplan [2006] provide information that is time-resolved and traceable back to ICOADS. There is a single study of pervasive systematic errors in hull sensor measurements [Kent et al., 1993], which analyzed data from a small number of ships over a two year period and found that hull sensors were relatively unbiased and showed no systematic change of bias with depth.
Few studies have looked at the long-term stability and calibration drifts of drifting buoys. Reverdin et al. [2010] installed 16 drifters with high quality temperature sensors in addition to their usual temperature sensors and found that the temperatures measured by the drifters showed inaccuracies that were larger than the 0.1K target accuracy and that they exhibited significant calibration drifts. This is consistent with the behavior seen by Atkinson et al. [2013].
Previous - 3.3.1 Bias Adjustments 1850 to 1941 Index Next - 3.3.3 Estimating Uncertainty in Bias Adjustments