Princeton University
Princeton, NJ
This research illustrates the utility of combining satellite observations and operational analysis for the evaluation of parameterizations. A parameterization based on ice water path (IWP) captures the observed spatial patterns of tropical cirrus optical depth. The strong temperature dependence of cirrus ice water path in both the observations and the parameterization is probably responsible for the good correlation where it exists. Poorer agreement is found in Southern Hemisphere mid-latitudes where the temperature dependence breaks down. Uncertainties in effective radius limit quantitative validation of the parameterization (and its inclusion into GCMs). Also, it is found that monthly mean cloud cover can be predicted within an RMS error of 10% using ECMWF relative humidity corrected by TOVS Upper Troposphere Humidity.
Parameterization of cirrus optical properties
A parameterization is developed to predict the optical depth of cirrus clouds formed by large scale lifting. The routine is used in locations where cirrus is a priori known to exist. IWP is calculated as an equilibrium between deposition (from large scale lifting) and sedimentation of crystals. By reconstructing parcel trajectory the IWP can be specified as a function of only four parameters: temperature, pressure, vertical velocity, and lapse rate.
Operationally, ECMWF analyses are used to provide these four input parameters, and the ISCCP retrievals are used to find the occurrence of cirrus and its optical depth. In this way, the input to the parameterization and the optical depth against which it is validated are both obtained from observational sources. This allows the parameterization to be validated in isolation. Exact quantitative validation of the parameterization is not possible due to the large uncertainty in cirrus ice crystal size. Instead, optical depth patterns are studied. Spatial correlations between the observed and predicted optical depths are typically greater than 0.7 for the tropics and Northern Hemisphere mid-latitudes (Figure 18.1). The good spatial agreement largely stems from the strong dependence of the ice water path upon the temperature of the environment in which the clouds form.
Poorer correlations (r~0.3) are noted over the Southern Hemisphere mid-latitudes, suggesting that additional processes not accounted for by the parameterization may be important there. One potential source of error is a possible cold bias in the ECMWF analysis for the Southern Hemisphere upper troposphere. Improved correlations in the Southern Hemisphere and tropical regions are obtained by using global mean "cirrus" values of pressure, vertical velocity, and lapse rate, and allowing regional variations in temperature only (Figure 18.2). This may be due to errors in ECMWF vertical velocity, sub-grid variability in vertical velocity, and oversimplifications in the parameterization.
Quantitative evaluation of the parameterization is hindered by the present uncertainty in the size distribution of cirrus ice particles. Consequently, it is difficult to determine if discrepancies between the observed and predicted optical properties are attributable to errors in the parameterized ice water path or to geographic variations in effective radii.
Parameterizability of cirrus cloud cover
An empirical relationship between cloud cover and relative humidity is produced for high clouds. ECMWF relative humidity is used as input and output cloud cover is validated against ISCCP cloud cover. The relationship is optimized for one day of data and used to test monthly average agreement. While the agreement is generally good, there is a systematic underprediction of cloud fraction in the Inter Topical Convergence Zone and a systematic overprediction in subtropical descending regions. These discrepancies are probably due to errors in the ECMWF upper troposphere and can be partially corrected using the TOVS satellite Upper Troposphere Humidity. With this correction, monthly mean high cloud cover can be predicted within an RMS error of 10%. The agreement is poorer on a daily basis because there is no strong correlation between cloud cover and relative humidity on this time scale.
Reference
Soden, B.J. and L.J. Donner, Evaluation of a GCM Cirrus Parameterization Using Satellite Observations, J. Geo. Res., July 1994.
