Speaker: Dr. Elizabeth Kent from National Oceanography Centre

Abstract: Results from the "Historical Ocean Surface Temperature: Accuracy, Characterisation and Evaluation (HOSTACE)" project will be presented. HOSTACE set out to better understand uncertainty in sea surface temperature (SST) observations and to produce a new gridded analysis of monthly SST from 1850 to present. There are many sources of bias in observations of SST that can depend on the method of observation, protocols used, ambient conditions and ship-specific issues such as a poorly-calibrated thermometer. Adding to the uncertainty are more generic problems such as miscoding, transcription or transmission errors, which can affect the observations and the associated location, date and time.

The steps taken toward quantifying biases in the historical record will be described. Observations thought to be made on the same ship were clustered based on either their identifiers (names, callsigns, etc) or their trajectories. Laboratory measurements of water samples in buckets were used to confirm and simplify a model of temperature change in samples in different types of bucket as used by the Met Office. Unknown measurement methods were estimated using characteristic differences in the diurnal cycle. The expected variability in SST was quantified using high-quality satellite and drifting buoy data from the past 3 decades. The estimation of bias then used a Bayesian approach that allowed for uncertainty in the reference field and the observations to show previously-unidentified regional biases in SST observations from ships. [Paper]