Session 2.2 Assessment of numerical weather forecasts against near-surface observations over an annual cycle

Monday, 9 June 2008: 10:45 AM
Aula Magna (Aula Magna)
John M. Edwards, Met Office, Exeter, United Kingdom; and J. McGregor, M. R. Bush, and F. J. Bornemann

Presentation PDF (263.1 kB)

Boundary layer schemes in numerical weather forecasting models are typically assessed against observational data gathered from field experiments operating over fairly short periods of time. A complementary perspective is provided by examining the statistical behaviour of forecasts over a longer period of time.

The Met Office runs three forecasting models of different resolution over the United Kingdom: a global forecasting model with a resolution of 40 km, a North Atlantic and European model with a resolution of 12 km and a high-resolution model covering the UK, with a resolution of 4 km. Within the domains of these models, surface observations are routinely made at a field site at Cardington, Bedfordshire.

For the period from October 2006 to September 2007, comprising a complete annual cycle, comparisons have been made between the forecasts and the observations at Cardington, focusing on the surface flux budget and the near-surface temperature profile.

The closure of the annual mean surface flux budget in the observations provides a check on their consistency and closure is achieved to an accuracy of around 5 Wm-2. The boundary layer is dominantly stable from October to February, matching the solar forcing and emphasising the importance of the stable boundary layer in the winter months. In the summer months, the forecasting models all show high net shortwave fluxes at the surface, consistent with a deficit in cloud amounts, though the size of the error varies between the different models.

Forecast surface and screen temperatures show a cold bias during the winter months (December to February), whilst a warm bias is seen at night during the spring and summer. In very stable conditions, particularly on calm evenings in spring, the forecast surface temperature is generally too high while the difference between the surface and screen temperatures is underestimated by the models, so that the error in the screen temperature is smaller than that in the surface temperature. Generally, wind speeds in the model are too high in these cases, but if the observed and forecast winds are very weak screen temperatures may show a pronounced cold error around the evening transition. The link between these errors and the surface flux budget is being analysed.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner