Observations of marine air temperature made by Voluntary Observing Ships (VOS) are known to contain significant biases which depend both on environmental factors, such as solar radiation, and on the exposure of the instruments. These errors can be greater than 2°C when the solar radiation is strong and the instruments poorly sited. The potential existence of such large errors has prompted the development of air temperature climatologies based only on night-time data. Previous attempts to correct for this heating error have assumed a steady state: relating the heating errors to the incident short wave radiation and the relative wind speed existing at the time of measurement. However this type of correction scheme is shown to over correct the air temperature in the morning, under correct in the afternoon, and to fail to remove biases which persist after sunset.

An analytical model of the heat budget of the ship and sensor environment has therefore been developed. With certain assumptions the resulting equations can be solved for time-varying incoming solar radiation to give estimates of the difference between the ship environmental temperature and the ambient air temperature, and hence the error in ship measured air temperature. The analytical model contains constants relating to the thermal and physical properties of the ship and also the relative wind speed. These constants need to be empirically estimated using the ship air temperature data. Preliminary investigations used the VOS Special Observing Project for the North Atlantic (VSOP-NA) dataset of paired ship and Numerical Weather prediction (NWP) model air temperatures and showed that differences between ship and NWP model air temperature were well approximated by the heating model. The analysis shows that ships recruited by different countries can have very different heating characteristics, largely due to different exposure of the air temperature sensors. This implies that the empirical constants need to be derived separately for different countries where possible; this significantly reduces the residual heating error after correction.

The analytical model is then fitted using data from the Comprehensive Ocean Atmosphere Data Set (COADS) to extend the correction technique to the global ocean. The resulting corrections are verified by examining the correlation of air temperature and sea temperature anomalies and evaluating the decrease in the air temperature random errors.