Laboratoire de Météorologie Dynamique: Model LMD LMD6b (3.6x5.6 L11) 1995

Model Lineage

Model LMD LMD6b (3.6x5.6 L11) 1995 developed from modifications of the AMIP baseline model LMD LMD5 (3.6x5.6 L11) 1991. Differences include:

Introduction of a diurnal cycle (in addition to the seasonal cycle) in solar radiation. This also requires higher-frequency calculation of radiative fluxes, provision for diurnal sun-angle dependence in ocean albedo, and a somewhat different sampling (storage) procedure for surface variables.
Inclusion of a multilayer soil heat transfer model.
A formulation of vertical advection of momentum that ensures angular momentum conservation.
Changes in the numerics of horizontal advection of moisture.
Modifications of the prognostic precipitation scheme, which also affects cloud-formation.
Changes in the optical properties of clouds which impact their interaction with radiation.

In addition, the initial conditions and computational environment of the AMIP integration are different from those of the baseline model.

Documents that are especially relevant to the LMD6b model include a note on angular momentum conservation by Hourdin (1992)[34], a description by Polcher (1994)[35] of the multilayer soil heat transfer model, and a discussion of the cloud-formation and precipitation schemes by Le Treut et al. (1994)[36].

Numerical/Computational Properties

Horizontal Representation

The semi-upstream horizontal advection scheme for moisture in the baseline model is replaced by a simple upstream scheme in order to avoid unrealistic simulation of clouds and precipitation associated with the former scheme.

Computer/Operating System

In contrast to the baseline model, simulations are performed on an 8-processor Cray C90 (but only using a single processor) in a UNICOS operating environment.

Computational Performance

Use of a more powerful computer than for the baseline integration yields a performance improvement > 200%: about 0.8 minutes of C90 computation time per simulated day.

Initialization

The conditions of atmospheric state, soil temperature/moisture, and snow mass/cover for the start of the repeated AMIP integration on January 1, 1979 are obtained from the conditions of the same variables after integrating companion model LMD LMD6s (3.6x5.6 L11) 1995 with the SECHIBA land-surface scheme for one year following its 1 January 1978 initialization.

Time Integration Scheme(s)

Because of the introduction of a diurnal cycle, shortwave and longwave radiative fluxes are recalculated more frequently (every two hours) than in the baseline model. Fluxes in cloudy regions are computed with cloud optical properties that are updated at each 30-minute physics time step.

Sampling Frequency

For the repeated AMIP experiment, daily mean values of model variables are saved once per 24 hours, as in the baseline model. However, the daily maxima and minima of surface variables with large intradiurnal fluctuations (e.g., temperature, precipitation, evaporation) also are saved, together with the times of their maxima.

Dynamical/Physical Properties

Atmospheric Dynamics

In addition to the dynamical framework of the baseline model, vertical advection is formulated so as to ensure conservation of angular momentum (cf. Hourdin 1992[34]).

Solar Constant/Cycles

In contrast to the baseline model, a diurnal cycle in solar forcing is simulated in addition to the seasonal cycle.

Radiation

As in the baseline model, cloud optical thickness and emissivity are determined from the cloud liquid water path W (in kg/m^2), the effective radius r of cloud droplets (in m), and the absorption coefficient k (in m^2/kg); however, the prognostic cloud liquid water content (LWC) (and therefore W) as well as r and k are determined differently.

The LWC changes because of the introduction of a different parameterization of precipitation. The effective radius of cloud droplets r is prescribed as 10 x 10^-6 m for warm clouds (with cloud-top temperatures > -10 deg C) and as 30 x 10^-6 m for cold clouds (with cloud-top temperatures < -10 deg C), instead of r being a linear function of cloud liquid water density as in the baseline model. The absorption constant k is also different for warm and cold clouds: 130 m^2/kg and 70 70 m^2/kg, respectively, rather than a constant 130 m^2/kg, as for all clouds in the baseline model.

These changes in cloud optical properties result in a general improvement of the simulation of long-wave cloud radiative forcing as compared with Earth Radiation Budget Experiment (ERBE) satellite data. Cf. Le Treut et al. (1994)[36] for further details. See also Cloud Formation and Precipitation.

Cloud Formation

The scheme for prognostic cloud formation as a function of liquid water content (LWC) is the same as in the baseline model. However, changes in cloud formation result from a different parameterization of precipitation.

Precipitation

Instead of specifying a sharp distinction between warm and cold clouds for the prediction of precipitation as in the baseline model, functions that provide for a smoother transition between warm and cold regimes are used.

For warm clouds (with cloud-top temperatures > -10 C) the precipitation rate is parameterized after Sundqvist (1981)[37] as the product of a characteristic precipitation time scale T (value = 5.5 x 10^-4 s^-1), the prognostic cloud liquid water mixing ratio m, and an exponential function of (m/C)^2, where C is a prescribed precipitation threshold value = 2 x 10^-4 kg/kg.
For cold clouds (with cloud-top temperatures < -10 deg C), the precipitation rate is determined by the ratio of m to a different timescale t = z/v, where z is the depth of the vertical layer and v is the terminal velocity of the water droplets, which is determined as an empirical function of m after Heymsfield and Donner (1990)[38].
Because precipitation is the main sink term in the budget of cloud liquid water, these parameterization changes also affect cloud formation and cloud-radiative interactions, both of which depend on cloud liquid water content (LWC).

Surface Characteristics

Surface characteristics are treated in the same manner as in the baseline model, except that:

The ocean albedo is dependent on the diurnally and seasonally varying sun angle, which is recalculated at each 30-minute physics time step. The albedo is prescribed so that at each latitude the integral over a 24-hour day yields the same value as in the baseline model (with no diurnal cycle).
The seasonal variation of the vegetation (affecting the seasonal land-surface albedo and roughness length) depends on the prognostic soil temperature, rather than being prescribed as in the baseline model. (However, when vegetation and snow cover are present, the land-surface albedo is computed in the same manner as in the baseline model.) Cf. Polcher (1994)[35] for details.

Land Surface Processes

Soil hydrology is simulated with the same "bucket" scheme as in the baseline model, but soil thermodynamics is instead determined by a 7-layer heat transfer model. The 7 layers are of uneven depths and are spaced between 0.02 m and 3.0 m below the surface, providing for resolution of thermal forcing at periods from 0.5 hour to 2 years. A zero-flux condition is imposed at the model's lower boundary, and the thermal insulation of snow is accounted for at its upper boundary. Introduction of the 7-layer model impacts ground temperature, snow mass/melt, and the seasonal change in prescribed vegetation that is tied to the soil temperature at a depth of 0.4 m. In turn, changes in snow cover and vegetation affect the albedo and roughness length over land. Cf. Polcher (1994)[35] for further details.

UCRL-ID-116384