Assessment of biases in merchant ship surface temperatures

Kent, E.C. ORCID: https://orcid.org/0000-0002-6209-4247. 2002 Assessment of biases in merchant ship surface temperatures. Southampton, UK, Southampton Oceanography Centre, 97pp. (Southampton Oceanography Centre Research and Consultancy Report, 58)

Before downloading, please read NORA policies.

Preview

Text
MTD_02_10.pdf
Download (8MB) | Preview

Abstract/Summary

Weather observations from Voluntary Observing Ships (VOS) have been analysed in an attempt to identify biases resulting from different observing practices. The observations were taken from the Comprehensive Ocean Atmosphere Dataset (COADS) for the period 1980 to 1997. A new quality control procedure was developed and implemented based on a comparison of local data rather than on climatological limits and individual ships were tracked to remove data with corrupt positions. Metadata containing information on measurement methods from a World Meteorological Organisation catalogue of VOS was associated with individual ship observations to enhance the information already contained within the COADS. Observational pairs were constructed from the COADS where reports were colocated within 50 km and 1 hour. This procedure reduced uncertainty about the comparisons of different measurement methods due to spatial and temporal variability. Sea surface temperature (SST) reports made using a bucket and thermometer were shown to be biased cool in conditions where surface heat loss was expected to be large. Bucket SST reports, which sample only the near surface, were relatively warm under conditions where a shallow diurnal warm layer in the ocean might be expected to form. Biases in the nighttime bucket SST reports were quantitatively assessed using a statistical technique that accounted for the error structure of the dataset. This suggested that at moderate wind speeds bucket SST data are biased by an approximately constant fraction of the air - sea temperature difference. SST derived from the engine intake thermometer were much noisier than those measured using buckets but, on average, no significant bias could be detected. This contradicts earlier studies which are comprehensively reviewed. Biases in air temperature measurements depend on the exposure of the sensor (which depends on recruiting country), on the incident shortwave radiation, the ventilation of the sensor and the past heating of the sensor. Regional variations in the height of the air temperature measurement are shown to be significant as are changes in observation height with time.