Department of Numerical Mathematics (of the Russian Academy of Sciences): Model DNM A5407.V2 (4x5 L7) 1995

In AMIP baseline model DNM A5407.V1 (4x5 L7) 1991, a numerical error in implementing the Matsuno time integration scheme resulted in unrealistic simulation of the state variables (cf. Galin et al. (1995)[26]). In the present model, a different time integration scheme is used. In addition, horizontal diffusion, penetrative convection, the surface characteristics and moisture flux, and land surface processes are parameterized differently than in the baseline model.

Model Documentation

Galin et al. (1995)[26] discuss the nature of the numerical error in the baseline model and its impact on the first AMIP simulation. They also demonstrate the improved simulation of several AMIP standard output variables by the present model.

Numerical/Computational Properties

Computer/Operating System

In contrast to the baseline model, the repeated AMIP simulation was run on an IBM RISC computer using a single processor in a UNIX environment.

Computational Performance

For the repeated AMIP experiment, about 4 minutes of IBM RISC computer time per simulated day.

Time Integration Scheme(s)

Instead of the baseline model's Matsuno scheme, time integration is by a semi-implicit leapfrog method with time step of 20 minutes for dynamics, 3 hours for radiation, and 1 hour for all other model physics.

Smoothing/Filling

Atmospheric temperature, specific humidity, and u-v winds are filtered at latitudes poleward of 69 degrees, at somewhat higher latitudes than in the baseline model.

Dynamical/Physical Properties

Diffusion

As in the baseline model, horizontal diffusion of winds, temperature, and specific humidity on constant-sigma surfaces is accounted for, but the second-order nonlinear formulation of diffusion is replaced by a fourth-order linear scheme.

Convection

A dry and moist convective adjustment scheme is used in place of the Kuo (1974)[13] parameterization in the baseline model. If the vertical lapse rate exceeds dry adiabatic, the lapse rate is restored to dry adiabatic while conserving dry static energy in the vertical column. A moist convective adjustment is applied if the lapse rate exceeds a critical value that is a linear function of relative humidity. (The critical value is equal to a dry adiabatic lapse rate if the relative humidity is 70 percent, and is equal to a moist adiabatic lapse rate if the relative humidity is 100 percent). The moist adjustment restores the lapse rate to the critical value while conserving the total moist energy of the vertical column. After application of the moist adjustment, the relative humidity of the column is set at 70 percent.

Surface Characteristics

In addition to the surface characteristics of the baseline model, the fractional area of vegetation (undistinguished by type) is specified for each grid square from 1x1-degree data of Matthews (1983)[27], as modified by Wilson and Henderson-Sellers (1985)[28]. The vegetation canopy affects both the surface moisture flux and the land surface hydrology.

Surface Fluxes

The baseline model's formulation of the surface moisture flux from land grid boxes is modified to include the effects of a vegetation canopy. The total surface evaporative flux in a grid box therefore includes contributions from snow cover, bare soil, the moisture intercepted by the vegetation canopy, and the transpiration through the dry foliage. See also Surface Characteristics and Land Surface Processes).
Over oceans, the surface heat and moisture fluxes are no longer augmented by the evaporaton of sea spray in stormy conditions, as in the baseline model.

Land Surface Processes

The representation of land surface hydrology is modified to include the effects of a vegetation canopy (see Surface Characteristics and Surface Fluxes). That part of the precipitation not intercepted by the canopy (whose capacity varies from 2 x 10^-3 to 8 x 10^-3 m, depending on location) is assumed to infiltrate the soil. Variations in soil moisture are simulated by the same modified "bucket" scheme as in the baseline model, but runoff is partitioned into fast (surface) vs slow (deep) components.

UCRL-ID-116384

AMIP Home