Assessment of numerical weather forecasts using satellite land surface temperatures

The forecasting of near surface air temperatures (at heights of about 2 m) is an important goal of numerical weather prediction, but the assessment of forecasting models using synoptic observations of near-surface air temperatures is complicated by questions such as the representativity of the data.

Over past decade considerable progress has been made in the retrieval of land surface temperatures (LST) using the infra-red channels of satellite instruments and an accuracies of about 1-2 K are claimed for current schemes. Satellite data have the advantages of broad and homogeneous spatial coverage and high temporal resolution, although retrievals are restricted to cloud-free conditions.

Here, land surface temperatures from the SEVIRI instrument are used in conjunction with synoptic observations of near-surface air temperature to assess forecasts made with the UK Met Office's operational forecasting system, concentrating on cases without clouds. Examples will be shown from contrasting seasons. In very settled conditions in the winter months, the combination of synoptic and surface measurements points to decoupling of the surface in the hours following the evening transition. In summer and autumn, the diurnal range of LST is underestimated over more arid regions, with the model showing apparent cold biases around the middle of the day, while errors in near-surface air temperatures are of a smaller magnitude.