MeteoSwiss - COST 728_scientific

Figure 1a: Comparison of the COSMO simulation

Figure 1: Comparison of the COSMO simulation with LES data for an ideal convective case. (a) Scaled TKE profile;

Figure 1b: Comparison of the COSMO simulation

Figure 1b: TKE budget terms from the LES;

Figure 1c: Comparison of the COSMO simulation

Figure 1c: TKE budget terms of COSMO.

The performance of the COSMO models turbulence scheme was tested in several case studies. The investigation of an ideal convective case showed several important drawbacks of the scheme [3]. As our goal was to understand the behaviour of the turbulence scheme in details, it was not enough to verify the predicted Turbulent Kinetic Energy (TKE), which is the main turbulence variable of COSMO, but the sources and sinks of the TKE budget had to be analyzed separately. The results show that the turbulent transport of TKE is too weak in the COSMO model, compared to the LES results (Figure 1). Consequently, TKE values at the top of the planetary boundary layer (PBL) are low and the negative buoyancy flux in the entrainment zone is nearly completely missing. The improvement of the turbulence scheme on the basis of this analysis is subject to ongoing work at MeteoSwiss.

Turbulence characteristics diagnosed by the coupling interface were also verified using measurement data from the CN-Met Project. The verification results show an overall good performance of the modelling system (Figure 2), with all the selected turbulence parameters being in an acceptable range (20-30% relative bias).

Figure_2.pdf, 103 KB

Turbulent kinetic energy, which is the only turbulence related model variable in COSMO, is generally underestimated by the model. Two different approaches for the post-diagnosis of turbulence variables, as required by the LPDM dispersion model, were compared. Very good performance was observed in the case of vertical turbulence, which is the most important turbulence variable with respect to mesoscale dispersion modelling.

The standard deviations of horizontal wind speed are not as well predicted as that for the vertical component. The method based on similarity theory shows slightly better performance than that based on the direct use of TKE from the COSMO model. The turbulence variables used by ENSI (Swiss Federal Nuclear Safety Inspectorate) are also predicted well by the COSMO model, with a moderate underestimation of the standard deviation of wind direction.

First, the impact of diagnosed turbulence characteristics on the dispersion process was evaluated on hypothetical pollutant releases (Figure 3). Results show that the different interfacing approaches can lead to substantial changes in the forecasted concentrations, with the method based on the direct usage of TKE resulting in lower concentrations [4].

Figure 3. Forecasted near-surface (below 500 m AGL) mean concentration fields by the COSMO-7 - LPDM system for 8 September 2008 18 UTC (18 hour forecasts, hypothetical case). Simulations differ from each other in the interfacing approach applied: (a) Similarity approach with PBL heights determined with the bulk Richardson number method; (b) Similarity approach with PBL heights determined from TKE profiles; (c) "Direct" approach.

The scaling relationships used in dispersion modelling systems are usually based on turbulence datasets over flat and homogeneous surface. However, turbulence structure in complex terrain, such as in steep and narrow Alpine valleys, can be substantially different from flat conditions. In this work, a first attempt was made to account for the modified turbulence characteristics in complex terrain, and a new scaling approach suited for steep and narrow Alpine valleys [5] was investigated with respect to pollutant dispersion [6]. The new approach was tested on the TRANSALP-89 tracer experiment which was conducted in highly complex terrain in southern Switzerland. The sensitivity of the modelling system towards the soil moisture, horizontal grid resolution, and boundary layer height determination was investigated. It was shown that if the flow field is correctly reproduced the new scaling approach improves the tracer concentration simulation, when compared to the classical coupling methods (Figure 4).

Figure 4. Simulated tracer concentrations on 19 October 1989 for a selected measurement station with different interfacing approaches: "Direct" (dash-dotted line), similarity approach (dotted line) and new scaling approach (dashed line), compared to tracer measurements (solid line).

For further information, please contact: Pirmin Kaufmann

COST 728: Scientific Results