MetCast: a new model for cloud forecasting in the very short range (0 - 12 h)

S. H. van der Veen
<

Royal Netherlands Meteorological Institute (KNMI)

PO Box 201

3730 AE De Bilt

The Netherlands

The acronym MetCast stands for Meteosat cloud assimilation and advection. This model, which has recently been developed at the KNMI, is 3-dimensional and time dependent, and it can compute liquid water contents in clouds explicitly. In particular cloud cover, cloud base heights and cloud top heights can be predicted. Cloud initialisation is performed by the MetClock (ref) scheme, which computes degree of cloud cover and cloud top temperatures from meteosat ir and vis observations. Also synoptic observations of cloud base heights are used.

Subsequently, the 3-D clouds are advected (allowing for evaporation and/or condensation in vertical movement) with wind fields that are forecast by the HIRLAM model.

We have computed the skill of both MetCast and HIRLAM in the period January 19 through 31, 1998, by comparing predicted total cloud cover to synoptic observations over the Netherlands and its close surroundings (appr. 80 stations). MetCast started its cloud analysis at 10 UTC, and it computed 12 hours ahead, until 22 UTC. As MetCast computations are available at 11 UTC, we compared its results to the HIRLAM forecasts that start at 6 UTC.

In the two figures below it is seen that during the first hours of the MetCast run the rms error in total cloud cover is nearly a factor 3 smaller than that of HIRLAM. The accuracy of the MetCast forecast gradually decreases, and it finally reaches the HIRLAM value after some 12 hours. The bias in the computed cloud cover is nearly zero for MetCast, whereas HIRLAM clearly predicts too few clouds (at least during these two weeks).

As the (preliminary) results of verification of the total cloud cover of MetCast show considerable skill over predictions made by HIRLAM, we can draw two conclusions:

1. It seems worthwhile for the HIRLAM model to make more use of synoptic observations (cloud base height) and especially Meteosat observations (via MetClock) in the analysis, thus improving the initial humidity fields.

2. It takes on the average at least half a day until dynamic forcing of cloud formation/evaporation dominates over the influence of initial cloudiness.

*) Feijt, A. J. and S. H. van der Veen: Cloud detection in a short range cloud prediction model using Meteosat and NWP model data. To be submitted to: Journal of Applied Meteorology.