Intercomparison of single-column numerical models for the prediction of radiation fog

1. Intercomparison of single-column numerical models for the prediction of radiation fog Bergot, T., E. Terradellas, J. Cuxart, A. Mira, O. Liechti, M. Mueller and N. W. Nielsen 6th. Annual Meeting of the E.M.S., Ljubljana 4-8 Sep. 2006

2. Introduction The presence of fog has a critical impact on ground, maritime and, especially, airborne transportation. Availability of precise forecasts may improve both the safety and the efficiency of the air traffic management.

3. Shortcomings of operational NWP models in predicting fog • Horizontal and vertical resolutions are too coarse • Surface and boundary layer processes are not accurately enough parameterised, especially under stable conditions • Initialisation of surface and boundary layer is not good enough. Photo: Ted Eckmann, UCSB Geography Department

4. 1D (single column) models: an alternative • May improve vertical resolution • May use more expensive parameterisations • May assess new schemes of physical processes • May modify the initialisation, using or discarding specific data or introducing data from dedicated observational systems. • May introduce climatological knowledge

5. Main goals of the experiment • Identify capabilities and limitations of SCM in fog forecast • Find out reasons behind different evolutions • Assess the importance of vertical resolution

6. Strategy • SCM are run and compared for well documented cases from the Paris-CdG field experiment • Models are run with the same initial conditions (IC): observations + assimilation scheme • 2 events: 1 fog, 1 near-fog • 4 sets of IC for every event

7. The models

8. Case 1: fog1-2 Oct 2003 Classical radiation fog between 2030 and 0600. Its depth progressively grows Visibility Wind speed 10 m  T 

9. Case 1: fog. Initialisation: 18 UTC All models predict fog, but at different times and with very different depths and liquid water contents LWC

10. Case 1: fog. Initialisation: 18 UTC T 45 m q 45 m Although average evolutions of temperature and specific humidity are quite correct, individual low-level evolutions considerably diverge, partly because of the data assimilation. T 2 m q 2 m

11. Case 1: fog. Initialisation: 21 UTC All models predict a late dissipation Different fog layers  T and q at 00 UTC (HH+03)

12. Case 1: fog. Initialisation: 00 UTC With a thick fog layer, the evolution is not so fast and the simulations tend to converge. The resolution of HIRLAM/INM is too coarse. MESO-NH has been run without gravitational settling. Profile of liquid water at HH+03

13. Case 1: fog. Initialisation: 03 UTC The dispersion in the burn-off time forecast is similar to that in the onset time.

14. Case 2: near-fog 11-12 Oct 2003 RH 2 m T wind speed 10 m Weak stability (moderate wind speed and weak inversion) Strong cooling High dew deposition (decreasing mixing ratio) q

15. Case 2: near-fog All models, except HIRLAM/INM predict fog at least in one simulation. HIRLAM/INM does not predict fog because of its underestimation of the cooling rate.

16. Case 2: near-fog Init.: 21 UTC 00 UTC The evolution of the screen temperature and humidity is correctly simulated by all models. Init.: 03 UTC 06 UTC

17. Conclusions • Under conditions of strong stability, the models present very different behaviour. • The simulation of fog needs models with a high vertical resolution. • Hi-res. does not release the models from the need of accurate parameterisations. • The adaptation of parameterisations to the resolution is crucial • The role of the gravitational settling and the dew deposition rate has to be highlighted