Using GEBCO digital bathymetry to infer depth biases in the XBT data

A global ocean 30 arc-second gridded GEBCO bathymetry is used as a reference to infer depth biases in expendable bathythermograph (XBT) data. In agreement with earlier studies we diagnose a depth varying XBT depth bias, with XBT sample depth being overestimated within the upper 50–200 m and underestimated below 200 m. The depth bias also exhibits certain time variability, with the fall rate for the majority of the XBT types being slower in the 1970s, with a minimum around 1975–77, again in agreement with earlier studies. However, when applied to the global XBT dataset, the new depth- and time varying depth corrections do not eliminate the total XBT temperature bias relative to the quasi-collocated Conductivity-Temperature-Depth (CTD) and bottle data thus implying the existence of a fall-rate-independent XBT thermal bias. In agreement with earlier works, the largest positive thermal bias is observed before 1980–85. The analysis also suggests different fall-rates for the probes produced by the two main XBT manufacturers.