Figure1Zonally averaged temperature in January. Left: model result. Contour interval is 10K. Right: deviation from GANAL climatology. Contour interval is 2.5K. Areas of less than -2.5K are shaded.
An atmospheric general circulation model for climate studies (CCSR/NIES AGCM) is developed. Although the model has not been participated in AMIP comparisons, a ten-year AMIP SST integration is performed and presented here briefly. The model reasonably reproduces the observed climatology. The simulation of hydrological cycles and cloud-radiation interactions seems reasonable, but leaves several problems to be further improved. For the interannual variability, temporal correlations between simulated and observed anomalies are reasonably good in the tropics but not significant in the mid-latitudes. Now we are preparing longer-period integrations, higher resolution (T42) ten-year AMIP SST integrations, and parametric sensitivity experiments.
The model is based on the primitive equation in global domain and uses spectral transformation method in horizontal and grid differentiation on sigma coordinate in vertical. The semi-implicit leap-frog time integration scheme is used. The physical parameterization includes a sophisticated radiation scheme, simplified Arakawa-Schubert cumulus scheme, prognostic cloud water scheme, Yamada-Mellor level 2 turbulence closure scheme with cloud effect, orographic gravity wave drag, and a simple land-surface submodel.
Radiative transfer scheme is based on the two-stream discrete ordinate method and the k-distribution method (Nakajima and Tanaka, 1986). The radiative fluxes at each level interface is calculated considering solar incidence, absorption, emission and scattering by gases, clouds and aerosols. The calculation of the flux is done in 18 separate bands. Band absorption by H2O, CO2, O3, N2O, CH4 are considered by k-distribution method with one to six sub-channels in each band. As for cloud, randomly overlapped partial cloudiness is included.
The cumulus parameterization scheme is based on Arakawa and Schubert (1974) with a few simplifications. One simplification is based on Moorthi and Suarez(1992) and that the vertical mass flux is assumed as a linear function of height. Other simplifications are that the precipitation efficiency is specified as a function of height, and that the bottom mass flux is determined by a relaxation of cloud work function to zero in a specified time scale.
The prognostic cloud water scheme with large-scale condensation is developed based on the scheme of Le Treut and Li (1991). The actual prognostic variable is the total water mixing ratio and it is diagnostically divided to water vapor and liquid water assuming a subgrid distribution of total water mixing ratio.
The model integrated with T21 (600km transform grid) resolution and with 20 vertically levels (top). The AMIP ten-year observed SSTs have been used as the bottom boundary condition. In this section, ten-year average of monthly climatologies (mainly January) are shown.
The zonally averaged temperature and its deviation from climatology of objective analysis data (JMA-GANAL) in January are shown in in Figure 1. The deviation is generally within 2.5K in the troposphere but there are rather large cold biases in the stratosphere, especially in high latitudes.
Zonally averaged temperature in January. Left: model result. Contour interval is 10K. Right: deviation from GANAL climatology. Contour interval is 2.5K. Areas of less than -2.5K are shaded.
The zonally averaged zonal wind (not shown) is realistic, but the center of the southern hemisphere jet is shifted equatorward by about 10 degrees and the easterly in the tropical and summer hemispheric stratosphere is rather weak.
The relative humidity (Figure 2) is generally low compared with GANAL climatology except for the baroclinic zone. Among others, dryness of the lower troposphere in the summer hemispheric subtropics is significant. The tropical upper tropospheric air is rather moist, which is created by the detrainment from cumulus clouds.
Figure2
Zonally averaged relative humidity in January. Left: model result. Right: deviation from GANAL climatology. Contour intervals are 0.1. Areas of less than -0.2 are shaded.
The precipitation (Figure 3) reproduces the estimated climatology reasonably well. However, there are several regions of insufficient precipitation: eastern Pacific ITCZ region and Amazon basin in January. In July, the precipitation is smaller than observation in the equatorial Africa and Indonesia and excessive near Philippine.
Figure3
Precipitation in January. Left: model result. Right: Shea climatology. Contour intervals are 60mm/month. Areas of larger than 120mm/month are shaded.
The zonally averaged radiative fluxes and the cloud radiative forcing are shown in Figure 4. These corresponds well to the observed value, except near 60S latitude. The shortwave forcing in the subtropical region is a little too strong, suggesting too large amount of the cloud there.
Figure4
Zonal averaged radiative fluxes at top of atmosphere and surface(left), and cloud radiative forcing(right). Solid line: longwave at top(left), net forcing(right); broken line: shortwave at top(left), longwave forcing(right); dotted line: shortwave at surface(left), shortwave forcing(right). Thick lines are observation (ISCCP,ERBE)and thin lines are model results. Unit is W/m^2.
In the low-level cloudiness field in July (Figure 5), the high cloudiness over the subtropical ocean just west of the continents are reproduced well albeit with a reduced magnitude.
Figure5
Low cloud amount in July. Left: model result (100hPa to 700hPa). Right: ISCCP climatology. Contour intervals are 0.2. Areas of larger than 0.4 are shaded.
In summary, the model reproduces the observed climatology well but several problems remain especially in the hydrological cycles and in the stratosphere.
Interannual variabilities simulated by the model are compared with observed circulation patterns. Overall magnitude of interannual variance in the troposphere is comparable, but somewhat smaller than, the observations. Temporal correlations between simulated and observed monthly-mean anomalies are significant (0.64 for 850hPa zonal wind) in eastern equatorial Pacific (Figure 6), but the correlation drops to insignificant values outside the tropical Pacific.
Figure6 Time-series of anomalies averaged over 5S-5N, 180W-120W for (a) 850hPa zonal wind , (b) 200hPa zonal wind , (c) OLR , (d) observational Nino-3 SST anomalies. 3-month running means are applied and thick lines for observations, thin lines for simulated results , "cor" indicates the simultaneous correlation coefficients between observations and simulations
Figure 7 compares correlations between simulated and observed OLR fields with observed NINO3 SST index. The response of eastern Pacific is reasonable but the signal in Indonesian region is weak and is shifted eastward. There is a moderate signal in Indian region in the model whereas the signal is very weak in the observation. It appears important to simulate accurately the spatial distributions of convective anomaly not only in the immediate neighborhood of the largest SST anomalies but also in some key regions such as the western Pacific and Indian monsoon regions.
Figure7
Mean correlation of the NINO3 SST with OLR (a) model result, (b) observation.