Snapshots of WRF activities in the GFI/Iceland groups

DLR / Bernadett Weinzierl

Layering

Observations from DLR research aircraft near Eyjafjallajökull

Lidar from the DLR flight across the plume on its way from Eyjafjallajökull to Scotland

Strong stratification (layering)

WRF with horizontal resolution of 9 km

FLOW

FLOW

WRF with horizontal resolution of 3 km

Hálfdán Ágústsson and Haraldur Ólafsson, 2010.

Permanent lowering of the flow level

Vertical mixing in a 3000 m deep layer

Vertical mixing at higher levels due to mountain waves

FLOW

None of these important features are visible at dx=9km, or at resolutions of many current forecast models!

WRF with horizontal resolution of 1 km

Hálfdán Ágústsson and Haraldur Ólafsson, 2010.

The model underestimates the sharpness of the inversion, and all smaller details above it

MODELOBSERVATIONS

Keflavík

http://belgingur.is

High concentrations of ash occures several times in Reykjavík, after the end of the eruption

Very clear source

Simulated surface flow

http://belgingur.is

Simulations of winds at the source of the dust (ash)

m/s at the source

Horizontal resolution (km)

The FLOHOF field experiment

1)

2)3)

4)

By far largest errors: Downslope winds

Jonassen et al

The the horizontal extension of the downslope winds is highly non-stationary

The SUMO model aircraft for meteorological observations

Photo: Haraldur Ólafsson

No extra observationsWith in-situ obs. with a model aircraft

A major difference in flow pattern extending far above mountain top level

23 km

Wind speed, ranging from 0 to 12 m/s

Vertical section of simulated flow across mountain

A study of turbulent fluxes inn Spitzbergen(Kilpelainen et al., Tellus – in press)

Dynamics of extreme precipitation in Central Norway (Steenesen et al. subm. to Tellus)

Tveita et al. MAP, in rev. 19

Case studies• Dynamics of mesoscale winds

WSP10m (CTRL-NOGREEN) +24h 5 March 2000 18UTC

– Frontal jet off shore– Cape Tobin jet– Pattern NE of Iceland– Wake

Prominent features

Using the WRF fields to calculate the surface gusts by the Brasseur method

Ágústsson & Ólafsson, MAP, 2009

Predictability studies:

Analysis of the dynamics of forecast errrors

Tveita, Olafsson, Sandvik & Hagen, MAP – in revisionHagen, Olafsson, Sandvik & Tveita, MAP – in revisionSteenesen, Olafsson & Jonassen, Tellus, submitted

RESULTS - FORECASTING

C: LEAD TIME

• increased lead time => decrease in forecast quality

• the decrease is quite regular for all the four cases and for most steps

• however, some steps are larger than the others, e.g. 72 h lead time to 96 h lead time for the 10 November 2006 case.

• ~5 K warmer in GOOD than in BAD

• Temperature difference =>difference in mean sea level pressure

DIFF (GOOD-BAD) MSLP

GOOD: 0h BAD: +24h

Black line: track of slp anomaly originating east of Hudson Bay

Red line: track of another slp anomaly

DIFF (GOOD-BAD) in θ850

RESULTS - FORECASTING