Yong Wang (ZAMG) and Eric Bazile (Météo-France)
Because of strong influence of mountain on weather processes, ALADIN have some specific problems over complex mountainous areas, especially the precipitation forecast. The aim of the work is to improve the ALADIN precipitation forecast. Special attentions are paid to investigating the impact of the model topography and LBC's on the ALADIN precipitation forecast.
The domains used for the numerical simulations are shown in Fig. 1 (Ahrens et al., 2001). The coarse-grid domain LACE (12.2 km horizontal resolution, 37 level in vertical, hydrostatic) is centred over Central Europe and coupled to ARPEGE. The intermediate domain VIENNA (9.6 km horizontal resolution, hydrostatic) is for the studies over the Alps and coupled to ALADIN/LACE; ALADIN/VIENNA is taken as the control run. The domain LMTA is positioned so as to include the Lago Maggiore Target Area, our most interest target area for investigation.
Figure 1. ALADIN domains for the numerical simulations. LACE : over central Europe, 12.2 km horizontal resolution; VIENNA : over the Alps, 9.6 km horizontal resolution; LMTA : Lago Maggiore Target Area; the sub-domain in red is used for the detailed investigation.
Heavy rainfall over the Alps occurs very often in autumn on the southern slopes of the Alps (Frei and Schär, 1998). The LMTA domain belongs to these maximums and was chosen for studying the mechanisms of persistent rain. This area is well investigated and observed within the Mesoscale Alpine Programme, MAP (Binder et al., 1996, Bougeault et al., 2001) and the EU Project RAPHAEL (Bacchi & Ranzi 2000). The dense operational observation network in the Alpine region was complemented during the special observation period (autumn 1999) by research instruments: ground-based and airborne radars, supplementary radio-soundings,... The dataset collected during the Intensive Observation Periods (IOPs) allows the direct study of heavy rain and the validation of fine-scale research and operational models over a mountainous area. One of the most intense rainfall episodes, IOP2b (19 and 20 September 1999), is used here for the study.
The synoptic situation of IOP2b is characterized by (details in Asencio et al., 2002) : a deep cyclone located to the West of Ireland at 00 UTC on 19th Sept., and moving to France at 12 UTC on 20th Sept.; a strong stationary anticyclone over Russia, which extends by a ridge even to South of the Mediterranean Sea East of 20°E. In North Africa, a second cyclone moves from Morocco to Tunisia during the IOP2b.
An active cold front is associated with the Atlantic cyclone. At 00 UTC, it extends along an approximately North-South axis from the British Islands to Portugal. It crosses France on 19th Sept. and reaches the Alpine orography during the night. The eastwards evolution across the Po Valley during the 20th Sept. is slowed down by the stationary high pressure over Eastern Europe.
This is a typical synoptic situation for heavy rainfall over the Alps because the cold front movement will intensify the West-East pressure gradient and therefore the low-level jet ahead of it. Moreover, warm air originating from Africa crosses the Mediterranean Sea and this leads to the formation of a conditionally unstable air mass.
For investigation of the ALADIN precipitation forecast over the complex mountainous area, the precipitation analysis (version 2.0, 25 km resolution) from ETH, Zürich, (Frei and Häller, 2001); and radar precipitation analysis provided by the MAP Data Center, Zürich, and M. Hagen, DLR, are used, which are shown in Fig. 2.
Figure 2. 24 h mean precipitation rate, in mm/h, from 06 UTC 20.09.1999 to 06 UTC 21.09.1999.
The representation of the orography in the model plays a crucial role in the precipitation forecast. For having more knowledge about the influence of the model orography on the precipitation forecast we simulate the IOP2b case by using different model orographies in the ALADIN/VIENNA, i.e. 9.6 km horizontal resolution, coupled to ALADIN/LACE. The model orographies we have chosen are:
The height-latitude cross-sections of the different model orographies and the original input data at longitudes 7.1°E, 8.5°E and 9.0°E are shown in Fig. 3. The mean orography is closer to the reality than the envelope and the semi-envelope ones, which increase the mass of the mountain. The comparison of the precipitation forecasts using envelope and mean orography is presented in Fig. 4. The impact of using the mean orography is positive, it reduces the precipitation maximum on the top of the mountain, and the forecast is closer to the analyses, but the other problem is still there, e.g. too dry in the lee side of the mountain. We have also compared the forecasts using full envelope and semi-envelope orographies in the model, there is indeed an improvement on the precipitation maximum, but not so evident.
One may be afraid that by introducing no envelope orography the wind forecast will be deteriorated. For this a comparison between the wind forecast with and without the envelope orography is presented in Fig. 5. This is the time-height cross-section of the averaged wind intensity. The result is very similar, and we can't conclude to any deterioration in the wind forecast by using no envelope orography. The validation of the wind intensity with the radar observations (see Asencio et al., 2002) shows that the forecasted upper-level jet is weaker in intensity and few hours earlier than observed, and in the low levels the wind intensity is overestimated a bit in the period 06 UTC to 12 UTC 20.09.1999. In this period, the cold front is over the area.
An experiment was done for investigating the impact of the coupling LBCs on the model precipitation forecast. In the simulation, the initial conditions and lateral boundary conditions, linearly interpolated in time between 6 hourly analyses (not as in the control run with a forecast), are given by the ARPEGE operational analyses. To analyse the impact, we focus on the western side of the Po Valley, and divide it into 3 areas. For the definition of those three areas, please see Asencio et al. (2002). The temporal evolution of the rain averaged over each area (Fig. 6) shows that the simulation with forecast as coupling overestimates the rain over Alpine slopes and Piedmont area, and is quite reasonable over central Po Valley. All of them exhibit a temporal shift. The experimental simulation is close to the radar observation in intensity over the Alpine slopes and Piedmont area, but underestimates the rain over central Po Valley. As in the control one, the temporal shift is still there.
Figure 3. South-North cross-sections of the orography used in the simulations, at 7.1°E, 8.5°E and 9°E. Grille_8proche : average of the 8 surrounding gridpoints. Grille_proche : average of the 4 surrounding gridpoints. NEW_ORO : without envelope. OLD_ORO : envelope. Relief_input923 : original input data. Semi2Oro : semi-envelope.
Figure 4. Comparison between the 24 h accumulated precipitation forecasts with (upper) and without (lower) envelope orography, valid for 06 UTC 20.09.1999 - 06 UTC 21.09.1999.
Figure 5. Time-height cross-sections of the wind intensity, in m/s. Left : forecast without envelope orography. Right : forecast with envelope. Valid from 00 UTC 19.09.1999 to 00 UTC 21.09.1999.
Figure 6. Time-height cross-sections of the hourly precipitation over the three areas. Black : radar analysis; red : research model Meso-NH; blue : coupled with forecast; green : coupled with analyses.
A comparison with Meso-NH, the red curves in Fig. 6, shows that ALADIN is better in intensity than Meso-NH, but the forecast timing is rather poor.
Based on the meteorological situation we consider the IOP2b case as 3 periods :
a) pre-frontal (00 UTC 19.09.1999 to 06 UTC 20.09.1999) : During this period, the cold front approaches the Po valley, the integration with analyses as coupling quite well simulates the quasi-stationary rainfall over LMTA, whereas this is not the case over the two other areas, where the duration of the rainfall episodes is more variable. In the last 6 hours of the period, the integration with forecast as coupling intensifies the rainfall too much, except over the central Po Valley.
b) frontal (06 UTC to 18 UTC 20.09.1999), it is associated with the end of rain over Piedmont, the intensification over LMTA and variable over central Po Valley. Both simulations don't recognize the timing of the rainfall and overestimate much, although the one with analyses as coupling is better.
c) the post-frontal period, the last part of 20. 09.1999, which is characterized by residual orographic precipitation and ended the whole episode. The good estimation of the rainfall over every area in both simulations is confirmed, by intensity and timing.
Again, as in Part 3.1, we have studied the wind forecast of both simulations, which is shown in Fig. 7. Both wind forecasts are alike, and in good agreement with the observations. In this aspect the model with the analyses as coupling improves the wind forecasts. However, the model overestimates the wind during the 20.09.1999 morning. In particular the simulated upper-level jet is in excess of 8 m/s compared with the observations. In the low levels, the simulation with no envelope and analyses as coupling improves the wind forecast; there is no overestimation of the wind intensity any more.
Figure 7. As in Fig. 5, but with ARPEGE analyses as coupling. Left : wind forecast with envelope orography. Right : forecast without envelope.
We have studied the MAP IOP2b case for investigating the possibilities of improvement on the ALADIN precipitation forecast over complex mountainous areas. This case is typical for heavy precipitation in the Alpine region. The emphasis has been put on the Lago Maggiore area in the Alps because of the dense observations collected by research and operational measurements during the MAP IOPs. ALADIN simulations with different orographies and coupling strategies have been carried on. The main conclusions of this work are summarized in the following :
Mean orography in the model does improve the precipitation forecast, and doesn't deteriorate the wind prognoses.
The best result for rainfall and wind forecast we get is by using the ARPEGE analyses as coupling LBCs, which indicates that obtaining a good ALADIN forecast largely depends on the quality of the coupling model.
Further works on this subject are still to be done, especially on the influence of the microphysics and the data assimilation for the initial state of the model. Mesoscale predictability of the quantitative precipitation forecast over complex mountainous area is another aspect of the future investigation.
We acknowledge D. Giard and F. Bouyssel for advice and technical help. This project has received funding from the Österreicher Austauschdienst (ÖAD), Republic of Austria. Cooperations with the Institut für Meteorologie unf Geophysik, Universität Wien and the MAP archive are kindly acknowledged.
P. Binder and C. Schär (eds) 1996, MAP Design Proposal,
P.Bougeault, P. Binder, A. Buzzi, R. Dirks, R. Houze, J. Kuettner, R. B. Smith, R. Steinacker, and H. Volkert 2001, The MAP Special Observing Period, Bull. Am. Meteor. Soc. 82, 433-462.
C. Frei and C. Schär 1998, A precipitation climatology of the Alps from high-resolution rain-gauge observations, Int. J. Climatol. 18, 873-900.
C. Frei and E. Häller 2001, Mesoscale precipitation analysis from MAP SOP rain-gauge data, MAP newsletter 15, 257-260.
B. Ahrens, Y. Wang and K. Jasper 2001, On ALADIN quantitative precipitation forecasts in Vienna. ALADIN Newsletter, 19, 64-68.
Y. Bouteloup 1994, Improvement of the spectral representation of the earth topography with a variational method, Mon. Wea. Rev., 123, 1560-1573
N. Asencio, J. Stein, M. Chong and F. Gheusi 2002. Analysis and simulation of the local and regional conditions for the rainfall over Lago Maggiore Target Area during MAP IOP2B. Submitted to Q. J. R. Meteor. Soc.