Evaluation of GLA-GCM upper-tropospheric moisture using TOVS radiance observations

Eric P. Salathé, Jr., Dennis Chesters, and Y. C. Sud

NASA Goddard Laboratory for Atmospheres Greenbelt, Maryland

This document is abreviated from Salathé, E. P., D. Chesters, and Y. C. Sud, 1995: Evaluation of the upper-tropospheric moisture climatology in a general circulation model using TOVS radiance observations, J. Clim., 8, 2404-2414.

1. Introduction

General circulation models (GCMs) are valuable devices for addressing global change issues. However, a GCM and its physical processes need to be evaluated for the realism of the simulated atmospheric dynamics and thermodynamics. In particular, accurate simulation of the vertical moisture profiles is central to the accurate modeling of the diabatic heating of the atmosphere by radiative and condensation heating processes. For example, the accuracy of simulated global warming, in response to increased greenhouse gases in the atmosphere, crucially depends upon the influence of water vapor and clouds. Therefore, a model projecting such scenarios must predict water vapor and clouds consistent with observations. In this paper, we use TOVS radiance data to evaluate the ability of the GLA model to depict the climatology of upper tropospheric moisture and its interannual variability. We also examine the effect of including downdrafts in the convective parameterization on the simulated moisture.

2. TOVS Data

TOVS radiance data are archived by NOAA as idealized nadir- viewing clear-sky brightness temperatures after applying corrections to the raw observations for satellite view angle and cloud contamination (see Wu et al. ,1993 for details). The TOVS data set used in this study is described by Chesters and Sharma (1992). The brightness temperatures presented here are the linear combination of the TOVS 6.7 and 7.2 µm channels used to enhance the sensitivity to the upper-most levels of the water vapor profile. To create daily images, the data along the satellite sub-orbital swaths were placed on a polar stereographic grid using the NOAA objective analysis, and then interpolated to a 5°x5° rectangular grid from 40S to 40N. 6% of the daily images were unrecoverable or discarded due to excessive missing observations; Seasonal averages are obtained from the remaining data

3. The GLA GCM

The version of the Goddard Laboratory for Atmospheres (GLA) GCM has a coarse resolution (4° x 5° x 17-layers) and is fully documented in Sud and Walker (1993, Table 3). The model is driven by observed SST. The simulated fields are interpolated to the 5x5 degree grid of the TOVS data, and daily images of TOVS- like radiances are computed from the temperature and moisture profiles using a narrow band radiation model (Salathé and Smith, 1994).

4. Results

4.1. Climatology

To compare the climatological seasonal cycle of brightness temperature as observed by TOVS and simulated by the GLA-GCM, seasonal averages were taken over the 10 year period 1979-1988. The correspondence between the observed (Fig 1) and computed (Fig 2) fields is generally quite good; the GCM captures most of the main regional features and their seasonal cycle. The brightness temperatures computed from the GCM are on average 2.8 K higher than observed. This bias is consistent with the bias found by Salathé and Smith (1994) in comparing brightness temperatures computed from precisely measured atmospheric soundings to GOES observations. The difference between the model and observations tends to be largest where the observed brightness temperatures are close to maximum or minimum. The GLA-GCM does not produce as strong a contrast between moist and dry regions as is observed. This effect is most pronounced JJA. The GCM indicates a dryer upper troposphere than the observations over the moist convective regions (e.g. over Central America and along the equator) and a moister upper troposphere in the dry subtropical subsidence regions just north and south of the equator (e.g. over the Arabia)

Fig 1. Seasonal cycle of observed brightness temperatures in K.

Fig 2. Seasonal cycle of computed brightness temperatures in K.

The region of high brightness temperatures across the southern Indian Ocean in JJA is an exception to the trend of the model¹s inability to capture extreme conditions well. In the observations, this region exhibits lower values than are found over N. Africa and the Middle East and comparable values to those in the East Pacific. In the GCM, this region gives greater brightness temperatures than are found over the Middle East or the East Pacific. In JJA, the contrast between the high values over the S. Indian ocean and the low values in the moist convective region over the bay of Bengal are similar for both the model and the observations. The gradient is weaker in the model, as the model becomes too dry in the middle of the Indian Ocean. The linear correlations of the spatial patterns of 10 year monthly average brightness temperature are nearly uniform throughout the seasonal cycle, varying from 82% in June to 78% in December. The correlation of year-by-year monthly averages of computed and observed brightness temperatures over the 10 year period is shown in Fig 3 (top line). The correlations are high during northern summer (JJA), but low during DJF. An exception is the beginning of 1983 during the El Niño, when the correlation is high throughout the year. The statistical significance of the lowest correlations is marginal, the highest values (80%) are significant. Thus the model can capture the patterns observed in JJA of individual years, but not those of individual DJF months. The 10 year model climatology for DJF months, however, does correspond well to the observed climatology. In JJA, convective centers are strongly centered on the N. Hemisphere land masses, whereas in DJF the convection is more sensitive to SST anomalies. Thus the weaker correlation in DJF may be attributable to the SST anomalies only weakly influencing the upper tropospheric moisture in the GCM.

Fig 3. Correlation of monthly brightness temperature patterns. Upper solid line: Observed and computed brightness temperatures. Lower solid line: Same but with seasonal cycle removed. Dashed line: Same as solid but using simulation with downdrafts.

4.2. Interannual Variability

The ability of the GLA model to capture the interannual variability in upper tropospheric moisture as indicated by the TOVS data can be examined by comparing their departures from the seasonal cycles. The monthly mean computed and observed brightness temperature anomaly fields are uncorrelated for most of the 10 year period (Fig 3 lower line). The correlations may be significant only for the1983 El Niño, indicating that the model is sensitive to this large SST anomaly. Otherwise, the GCM does not capture the interannual variability of upper tropospheric moisture.

4.3. Convective-scale Downdrafts

The convective parameterization used in the GLA GCM for the simulation discussed above does not include the influence of convective scale downdrafts. Sud and Walker (1993) recently introduced in the convective parameterization a new mass-flux parameterization for downdrafts that brings colder and drier air near the surface, displacing the warm and moist air upwards, and thereby moistening upper levels. The scheme was found to produce significant changes in the tropical precipitation and even the Hadley circulation (Sud and Walker, 1993). To evaluate the effect of the downdrafts on upper-tropospheric moisture in the model, we compare the brightness temperatures computed from a 1979-1980 run of the GLA GCM to the TOVS data. Including downdrafts has no significantly significant effect on the upper tropospheric moisture.

5. Conclusion

This study not only shows the GLA-CGM¹s strengths and deficiencies, but it gives us confidence in the value of the TOVS data for model evaluation and validation.

References

Return to Eric Salathé,'s WWW Page