21.06.2015 Views

Proceedings - C-SRNWP Project

Proceedings - C-SRNWP Project

Proceedings - C-SRNWP Project

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

The setup is based on a new numerical kernel, 2-timelevel 3rd order Runge-Kutta, with a<br />

main time step of about 15 seconds. It uses improved physics with e.g. new schemes for<br />

graupel, improved turbulence and shallow convection (the deep convection being resolved),<br />

as well as with implementation of the topographic effects on radiation. A new multi-layer soil<br />

model with 8 layers for energy and 6 for moisture is used; moisture is updated every 24h form<br />

IFS in the lowest levels. A measurement driven soil moisture analysis is being implemented,<br />

as well as a mosaic approach will enable to use the same soil for aLMo and aLMo2. A snow<br />

analysis at 2.2 km refined with the HRV channel of MSG will be used.<br />

It will have its own assimilation cycle with a rapid update cycle having a short cut-off of 60<br />

min. Additional measurements will be assimilated like radar data using the 2-dimension latent<br />

heat nudging scheme and the analysis of the boundary layer will be improved through a better<br />

usage of surface measurements and of the lower part of wind profiles.<br />

Validation of aLMo2 winds (P. Kaufmann and O. Marchand)<br />

The local dispersion modeling for the Swiss emergency response system for nuclear hazards<br />

is currently based on a set of pre-determined wind fields (WINDBANK). The project<br />

“Centrale Nucléaire et Météorologie” (CN-MET) has the purpose to replace the current<br />

system with a system based directly on aLMo2. The comparison between aLMo2 and a<br />

temporal, dense wind measurement network during the WINDBANK campaigns conducted<br />

by the Paul Scherrer Institute is a first step to establish the validity of this concept. The main<br />

advantage of aLMo2 over the current system, to deliver fully three-dimensional wind<br />

information over the whole volume of interest, does however not come into play in this<br />

comparison.<br />

The mean error and standard deviation of the error were calculated for hourly values of 72<br />

days. In addition, histograms and scatter-plots for direction and speed serve to compare model<br />

and measurements. Directions at speeds smaller than 3 ms -1 were filtered out. When looking<br />

at the results, one should keep in mind that the WINDBANK observations were pointobservations,<br />

whereas the aLMo2 forecasts represent a spatial mean over a whole grid cell,<br />

and thus a perfect match cannot be expected.<br />

The wind direction bias (ME) is well below 10° and the wind speed bias far below 1 ms -1 for<br />

all model versions. The direction bias is significantly larger for the coarser-resolution aLMo<br />

analysis than for the others (below 5°, see Table 1). In regards of standard deviation of the<br />

error (STDE), the wind direction of the high-resolution aLMo2 analysis is clearly better. The<br />

wind speed standard deviation however does not show a significant difference (Table 1).<br />

wind direction (°) wind speed (ms -1 )<br />

ME STDE N ME STDE N<br />

aLMo analysis 8.26 53. 15326 0.36 1.743 44155<br />

aLMo2 analysis 3.23 47. 23032 0.17 1.749 64280<br />

aLMo2 19-24h 3.53 57. 5424 0.09 1.789 16347<br />

Table 1: Bias and standard deviation of the difference model minus measurement, for wind direction and wind<br />

speed.<br />

The wind direction histograms for the WINDBANK measurement show a channeling of the<br />

flow over the Swiss plateau (e.g. Fig. 2a). This channeling is underestimated by aLMo2 at<br />

157

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!