Astroclimatology Forecast Models

Forecasting Local Flows: MM5 mesoscale model

The plots below summarize the results of MM5 test runs performed by Dr. Michel Cure (University of Valparaiso, Chile) in July 2001 when visiting the Institute for Meteorology of Munich Ludwig Maximilian University (MIM), Germany (Prof. Joseph Egger, Leiter Abteilung Theoretische Meteorologie).

These encouraging results triggered a further study by Michael Riemer at MIM with negative conclusions about the use of MM5 for operational predictions of local wind direction at Paranal. The reason is that the global models used for initializing MM5 are not accurate enough in this poorly monitored area of the world. The final report (PDF, 7.6 Mb, 140p) however gives a detailed understanding of the terrain-flow interaction at Paranal and nearby sites.

---

Following the example of Mauna Kea MM5 Operational Forecasts, the terrain model contains 4 domains in a nested grid configuration:
domain 1 is a grid of 99x99 points, grid distance between points 27km
domain 2, 82x82, 9km
domain 3, 76x76, 3km
domain 4, 55x55, 1km

The program TERRAIN from the MM5 suite of programs uses for the domain creation the following global/quasi-global data
domain 1 has 10 arcminutes global terrain and landuse (DEM_10M_GLOBAL)
domain 2, 5 arcminutes, (DEM_05M_GLOBAL)
domain 3, 30 arcseconds, (W100S10.DEM)
domain 4, 30 arcseconds, (W100S10.DEM)

The initialization data are from the NCEP analysis (NCEP_04_2001).

The run covers two nights, starting with a night having average to bad conditions (21-22 April 2001) and ending on a very bad night with high humidity, high wind and very bad seeing (22-23 April 2001). The run took 17 hours on a 1.5GHz PC.
The following plots are extracted using domain 4 terrain: Paranal summit (2634m) is at km10 on both EW and NS cuts, the higher summit at km 28 to the East is Cerro Armazones (3064m), the lower apparent elevations are due to spatial averaging by the 1km mesh.

Check local situation on the same night (21-22 April 2001).

East-West Vertical Cut MM5 Output

night 21 to 22 April 2001

North-South Vertical Cut MM5 Output

night 21 to 22 April 2001

Check local situation on the same night (22-23 April 2001).

East-West Vertical Cut MM5 Output

night 22 to 23 April 2001

North-South Vertical Cut MM5 Output

night 22 to 23 April 2001

Full data set (12MB) is available.

MM5, a public domain model also gave promising results (credit: http://hokukea.soest.hawaii.edu/forecast/mko/ ) when compared to SCIDAR profiles recorded at Mauna Kea Observatory. See also similar studies using the french developed Meso-Nh

Send comments to <Marc.Sarazin@eso.org> Last update: May 23, 2003 Back to Astroclimatology homepage.