Yong Wang (wang++at++zamg.ac.at)
To investigate the problems of NWP precipitation forecast over mountainous areas, two ALADIN model simulations with different model resolution and dynamics were carried out, a) at 9.6 km horizontal resolution, hydrostatic; b) at 4 km horizontal resolution, non-hydrostatic. The area of investigation is the "Ticino-Verzasca-Maggie", which has been well investigated and observed within MAP (Mesoscale Alpine Programme) and the EU-Project RAPHAEL. This area is a complex mountainous area with elevation differences up to 3000 m and typical slopes of about 30 degrees. It is characterized as a region with exceptional rainfall amounts and intensities. The cases we studied are MAP IOP2 (Sep. 19, 00 UTC - Sep. 22, 00 UTC, 1999) and IOP3 (Sep. 25, 00 UTC - Sep. 28, 00 UTC 1999). The methods we used are prognostic (the forecast used is a long model integration, e.g. 24 hours) and semi-prognostic (only one time-step forecast). For validation of the forecasts the precipitation analysis from ETH, Zürich, and radar data are used.
Weather situation for MAP case IOP2 :
In the pre-frontal air over the western Alps ahead of a low pressure system there was a strong convergence zone in the western Po Valley. With warm and moist air advected towards the Alps from the Mediterranean and an easterly low level jet in the lower layers of the troposphere this produced heavy precipitation in the Lago Maggiore area. The event can be described as orographic precipitation, caused by large-scale ascent of moist flow over a high mountain barrier with embedded convection.
Summary of results :
Both ALADIN forecasts (9.6 km hydrostatic and 4 km non-hydrostatic) overestimate the precipitation, the 4 km simulation even more. This is especially true for the south (upslope) side of the mountain. On the north (downslope) side and in the central areas the forecasts look too dry. The precipitation patterns of the 4 km have more spatial variability than the observations. The large-scale part of the total precipitation is larger in the 4 km than in the 9.6 km simulation, as expected. In the South-North direction, the forecasted precipitation pattern is strongly tied to the gradient of the model orography. It is found quite often that the strongest precipitation coincides with the steepest upslope. This correlation is not as strong in the real atmosphere (Figure 1). Use of the 4 km (vs. 9.6 km) resolution gives no significant improvement. In connection with this problem, some work on E923 will be done, for example using a linear grid. The semi-prognostic method is used to validate the precipitation parameterization. Since NWP models have a high degree of complexity, it is difficult to isolate errors caused by the parameterization from errors caused by other components of the model (e.g. numerical methods, other parameterizations). The semi-prognostic method can to some degree alleviate this problem. It is free of errors other than those caused by the parameterizations and observations. In the semi-prognostic mode the convection scheme gives weaker activity than in reality, and the large-scale precipitation is absent due to the lack of saturated layers in the model analysis (spin-up problem).
Figure 1: North-South profiles of topography and precipitation rate at 9° E longitude for 20-09-1999. Grey: topography, red: large-scale precipitation, black: convective precipitation. Shown are results from a 4 km run (2 uppermost panels), a 9.6 km run (middle panels), and radar as well as surface observations (thin black line).
 |
Home |