Parametrization of Lakes

(more details, Sandor Kertesz, HMS)

1. Introduction

At present, in ARPEGE/ALADIN lakes are treated in the same way as sea, so their surface temperature (Ts) is kept constant during the integration and they are characterized with their monthly climatological values. The main problem with this approach is that the initial values of surface temperature of lakes are not reliable in the model. Recent modifications in the code made it possible to use the correct monthly climatological values as initial values for the lakes. The first task of this study was the investigation of the effect of these modifications to the forecast of ALADIN. While, the second task was connected to the design of a simple lake parametrization model with the modification of the ISBA scheme in order to describe the temporal variation of Ts and the ice conditions on lakes.

2. Investigation the effect of changing the Ts on lakes in ALADIN

There are two problems with the initial values of Ts of lakes in ARPEGE/ALADIN. The first problem is that lakes are usually not described in the low resolution input datasets of configuration 923. So in the absence of data the monthly climatological values of Ts will be extrapolated from the SST values of the nearest sea gridpoints. The second problem appears in the creation of the initial conditions for ALADIN. Since, in configuration 927 the difference between Ts and its monthly climatological value is interpolated to the target geometry using the land-sea masks and then it is added to the target climatological value. So if the lake is represented only in the target domain the result could be quite unrealistic (Figure 2).

In the experiments the case of Lake Balaton in Hungary (Figure 1) was investigated and cycle AL11T2.05 was used with the operational ALADIN/HUNGARY domain. The coupling was based on ARPEGE. For each investigated date, two different initial lake temperatures were used: the original one that comes from the operational settings (in what follows EORI) and the correct monthly mean based on observations (in what follows ENEW). Mainly those dates has been selected between 1999/06/01 and 2000/01/31, when the surface temperature in EORI was extremely warm.

The 2m temperature (T2) and 2m relative humidity (RHU2) forecasts were verified with the observations of the four synoptic stations around Lake Balaton. The biggest impact in T2 was found at Siofok: the mean error in ENEW was smaller by 1 C on average, which improved the forecast in most cases. For the other stations the forecasts were nearly the same for both experiments. The cooling effect of the colder lake temperature strongly depended on the advection (Figure 3). When the wind was weak the cooling effect stayed local even if the lake was colder by 20°C. The colder lake temperature significantly reduced the evaporation on the lake. There were two cases with dry weather, high pressure and slight wind when the evoporation in the daytime decreased to almost zero. This phenomenon was caused by an extremely strong inversion over the lake in ENEW, even in the beginning of the integration, which blocked the turbulent transport between the surface and the lowest model level.

3. Lake parametrization

The ISBA scheme (Noilhan and Planton, 1989) with the SWF2L (Two Layer Soil Water Freezing) scheme (Bazile, 1999) was modified to simulate a lake. At first, the following basic modifications were made:

• In ISBA the time evolution equations for Ts and Tp (the mean of Ts for one day) and the formulation of the thermal coefficients are based on physical properties of the soil such as heat conductivity. In the case of a lake the turbulent mixing is a more intensive transport process near the surface so this approach cannot be applied. For the sake of simplicity it was supposed that these equations are formally valid for lakes, but Ts is the mean temperature for the whole water column and the characterisctic time for Tp is a properly set parameter. So the lake was supposed to be a well mixed water column, with uniform temperature.
• Both the surface and the deep reservoirs were forced to be always saturated.
• The surface ice content was taken into account in the formula of albedo and emissivity.

The experiments were carried out with the simplified 1D model which describes only the soil and surface processes and requires atmospheric forcings. Because the necessary observations were not available for a real la,ke the dataset measured at Col de Porte (in the Alps) between 1995/08/13 and 1996/08/12 was used. The best simulation was able to describe the changes of the surface temperature and the freeze/thaw cycle of a 3m deep lake (Figure 4), but to achieve this result further modifications were necessary:

• Snow was eliminated from all the equations.
• Ice formation was allowed only in the superficial reservoir.
• A new flux was introduced in the time evoulution equation of Tp in order to describe the heat conservation properties of the lake during the ice period.

4. References

• Bazile, E. (1999). The soil water freezing in ISBA. In HIRLAM Newsletter, Number 33, pp. 92-95.
• Nolihan, J. and S. Planton (1989). A simple parametrization of land surface processes for meteorological models. Monthly Weather Review 117, 536-549.

Figure 3: Effect of changed lake temperature to the forecast of ALADIN/HUNGARY (Run:1999-10-03 12UT). (The lake in ENEW forecast is colder by 21 C than in EORI forecast)

Figure 4: Superficial water content and temperature of the simulated 3m deep lake at Col de Porte from 1995/08/13 to 1996/08/07

Home