AMIP II Diagnostic Subproject 34

The knowledge of the water vapor distribution and its transport is essential for the understanding of the global water cycle of the atmosphere.  Water vapor is the most important greenhouse gas and in the context of a global climate change, it is usually considered to introduce a strong positive feedback [Manabe and Wetherald, 1967].  Despite the decrease of specific humidity wiht height, the upper level moisture can play an important role in the climate system through radiative processes [Schmetz et al., 1995]. Hence a good description of the upper level humidity is an important challenge for GCM's [Harries, 1996].

A net effect of convection appears to be a moistening of the upper atmosphere [Soden and Fu, 1995], so that, knowledge of the relationship between convective activity and water vapor distribution is also one of the key issues for modelling the atmosphere. Indeed, a GCM should not only reproduce the observed mean climatology, but also the correlation between different parameters which are likely to induce feedbacks. In the present frame, the GCM should reproduce a correct distribution of both cloudiness and upper level moisture and, also the response of subtropical dry subsidence areas to a modification of convective activity in the ITCZ.

Using a conceptual model, [Pierrehumbert, 1995] shows that the relationship between convective tropical area and subtropical subsiding area is an important parameter controlling the energy balance of the tropics, and [Picon et al., 1995] deduced from Meteosat water vapor channel data a negative relationship between the extension of the ITCZ and the size of the subtropical subsiding areas.

Recent work at LMD based on direct comparison between observed, Meteosat water vapor channel equivalent brightness temperature (WVEBT) and simulated WVEBT from the LMD GCM outputs, assessed the capacity of the LMD GCM to reproduce the observed link, on a seasonal scale, between the size of the ITCZ and the extension of subtropical subsiding areas [Roca et al., 1997].

Objectives

The objectives of this subproject are two fold :

• Evaluate the quality of the models climatologies of WVEBT versus the observations.
• Check the GCMs ability to reproduce  the large scale relationship between the extension of the ITCZ and the extension of subtropical subsiding zones on seasonal and interannual scales.

II.1 METEOSAT water vapor channel data

METEOSAT is a geostationary plateform located at the equator at 0 degrees of longitude. The first one the serie was launched in late 70's and METEOSAT-7 has been launched on the 3rd of September 97. In this subproject, we will use the data from its so-called water vapor channel, centered on the vibration-rotation band of the water vapor around 6.3 microns. In clear sky conditions, it is mainly influenced by the water vapor between 300 and 600 millibars, whereas in cloudy atmosphere, it is sensible to the cloud top temperature modulated by the water vapor above. METEOSAT water vapor channel has a vicarious calibration that makes use of a radiatve transfer model and of rawinsondes [Schmetz and Turpeinen, 1988].

The monthly mean water vapor equivalent brightness temperatures patterns have been linked to large-scale dynamics features : the coldest brightness temperatures (CBTs) are associated to upwards vertical movement and the warmest brightness temperatures (WBTs) are linked to subsiding motions [Picon and Desbois, 1990].

On a monthly mean basis, it has been shown using ISCCP cloud products, that the CBTs from that channel are strongly correlated with high and thick clouds and that the WBTs are associated with low levels clouds or clear sky [Picon et al., 1995]. Hence the CBTs are related to the ITCZ and the WBTs to the dry subsiding zones.

II.2   The model-to-satellite approach

We will use a narrow band radiative transfer model to simulate the WVEBT from the temperature, specific humidity profiles, introducing the cloudiness profile consistently in the calculation. This model has 10 spectral intervals covering the METEOSAT water vapor spectrum and was designed by Jean-Jacques Morcrette, ECMWF, from an improved version of the code described in [Morcrette and Fouquart, 1985]. The response function of the satellite captor is taken into account explicitely by convoluting the spectral radiances with the filter function of the instrument. The viewing angle is also taken into account in the computation.

This narrow band has been validated in the water vapor band against the operational code of EUMETSAT and gives similar results within 1 Kelvin in clear sky [Roca and Picon, 1997].

The clouds are treated as grey bodies as in the original code and classical overlap assumptions are available for the total sky radiance computation. Using a set of temperature, humidity profiles from the ECMWF analysis together with clouds profiles derived from EUMETSAT operational products, comparaisons with observations yield a good agreement between observed and simulated total sky brightness temperatures,  within 3K for the maximum overlap assumption [Roca, 1999].

In this way, we can make direct comparison with Meteosat water vapor images from both cloudy and clear part of the atmosphere without applying any cloud-clearing algorithm to the satellite data.

II.3   Simulated climatology assessment
Before studying in deep , the relationship between convection and upper-level moisture in the GCMs, one needs to characterize the main behavior of the models. This part of the proposed work accounts as much as the second point. Precisely, by comparing the observed and simulated WVEBT fields, we evaluate the ability of the GCMs to reproduce an observed important field over differents time scale.

For example,  Figure 1 shows this kind of comparison for one integration of the LMD-GCM cycle 6 with prescribed SST for an averaged july from 1983 to 1986. Fig 1(top) presents the observed monthly mean of the WVEBT. The coldest patterns clearly identify the ITCZ, and the subsidence areas, both side of the equator are also well characterized. The simulated field (fig 1 middle) shows up some relevant patterns like the ITCZ over Africa and the southern hemisphere subsiding zone shape is in agreement with the observations. Some discrepancies also appear: the nearly absent ITCZ over the Atlantic ocean and the strong dry zone over the Azores which is not so develloped in the observations. The extrema in the simulation are stretched compare to the observations.
Finally fig 1(bottom) is the difference between simulated and observed WVEBT. The GCM has a very wet (cold) zone within the ITCZ over Ethiopia which related to a well know cloud cover and precipitation overestimation in the LMD-GCM over these moutains.

But overall, this difference exhibits a double bias in the GCM: the simulated WVEBT are warmer over the ITCZ and are colder over the observed subsiding areas. In terms of moisture, it means that the GCM has a dry (resp. wet) bias over the ITCZ (resp. upper level of the subsiding zones). This seems to be a characteristic of many GCMs [Schmetz and Van de Berg, 1994; Soden and Bretherton, 1994; Roca et al., 1997] and may be important to be checked over the AMIP2 participating models.

Such a diagnostic can be related to problems in the simulation as diagnosticed by the WGNE standard diagnostic and help to characterize them.

II.4   Relationship between convection and upper-level moisture
Aware of the eventual bias of the GCMs, one can now go deeper in the analysis since models should reproduce climatology but also relationships that implies important variables.  This second part of the proposed work aims to check the ability of model to reproduce an observed relationship between the extension of the ITCZ and the extension of the subtropical subsidence areas, which indicates that a broader ITCZ is associated with smaller dry subsidence zones, i.e. an increase of convective activity in the ITCZ leads to wetter subtropical subsiding zones [Picon et al., 1995; Roca et al., 1997].

The size of the ITCZ is defined using an arbitrary threshold and is a first order indicator of the convective activity in this region. Similarly, a second threshold allows to discriminate the subsiding zone. The prodecude is fully detailled in [Picon et al., 1995]. Then, time series of these indices will be computed, first over a mean seasonal cycle and then over the anomalie fields computed from the 12 years of data and of simulations, in order to assess the GCMs simulation of this large-scale moistening.

Fig 2 is taken from the simulation described in Roca et al., 1997 and shows this relationship over a mean seasonal cycle computed from a 5 years period for both data and the LMD-GCM. The negative correlation indicates that a broader ITCZ is associated with smaller dry subsidence zones in the observations as well as in the GCM.

In the present proposed diagnostic, we will consider both seasonal and interannual variabilities of the upper tropospheric humidity and ITCZ size to encompass differents modifications of the convective activity and related dry zones.

III Computation strategy and data

II.1   Methodological bias characterization
Monthly means of Meteosat WVEBT are constructed from the imagery available every 3 hours and are compared to synthetic WVEBT computed from the monthly means of the models outputs, since the air temperature, specific humidity and cloudiness profiles are not available on a six-hourly basis. We expecte a slight bias induced by the use of monthly mean outputs[Chen et al., 1997].

We have investigated this bias in details with the cycle 6 of the LMD-GCM. This model is an improved version of the LMD model described in [Sadourny and Laval, 1984] and has a diurnal cycle and a advanced surface scheme. More details about this version of the model are avaliable in [Polcher and Laval, 95].

To assess the weak sensitivity of the simulated WVEBT monthly mean to time sampling of the GCM outputs and to confirm the feasibility of this diagnostic using monthly means of temperature, humidity and cloud profiles provided by AMIP2, the LMD-GCM was run for July 1993 and outputs where sampled as follow:

a. One single monthly mean
b. 30 daily means
c. 120 4xdaily means (6 hours averages)
d. 120 4xdaily snapshots (every 6 hours, no averaging)

WVEBT simulations were performed for each of these runs and averaged in monthly means. The comparison is done with respect to the first simulation here considered as the reference.

The figure 3(top) shows zonal mean differences between monthly mean simulations of the METEOSAT-4 WVEBT in July 1993 for the runs b.(plain line), c.(dashed line), d.(dotted line) and the reference (run a.). The maximum difference occurs in the southern hemisphere  and concerns the 6 hourly snapshot run. This maximum reaches 1.5 Kelvin which is around the uncertainties of the whole prodecure. For the others runs, this difference is even smaller. Notice the small discrepancies between 30 daily averaged inputs and 6 hourly averaged inputs.

Overall, the frequency of GCM outputs have a weak impact on the final WVEBT monthly mean and has to be related to difference between GCM WVEBT and observations. Figure 3(bottom) represents the difference between run a. and the observed WVEBT monthly mean build from the 3 hourly imagery. This difference is large and may reach up to -15 Kelvin over a large part of the considered region imagery, indicating that time sampling bias in the comparison can be considered as negligible.

These results imply that we can use the monthly mean outputs from GCM as defined by the AMIP Panel to process our diagnostic without any important methodological bias.

III.2   Satellite data
We will consider the data from July 1983 to July 1995, i.e. from METEOSAT-2 to METEOSAT-5. These data are avalaible through the ISCCP program, in the climatological studies adapted B3 format (1 pixel over 6 is retained from the full resolution). Over the 1983-1995 period, many changes occured in the METEOSAT time serie (changes of filter function of the captor, operational calibration modifications). Hence the long term observations of METEOSAT cannot be used as is. An intercalibration technique has been developped at LMD [Picon et al., 1998] and has been applied to homogeneise the long term serie of water vapor data [Picon et al., 1999] and is ready for use in this diagnostic subproject.

The observed monthly means images will be computed from this intercalibrated database using the imagery every three hours.

III.3   AMIP variables needed
In order to complete the proposed diagnositc, the following outputs from the GCMs are needed:

Air temperature , Specific humidity , cloud fraction
at the 17 WMO pressure level from July 83 to July 95
monthly mean

References

Chen C-T., Roeckner E. and Soden B.J., A comparison of satellite observations and model simulation of column-integrated moisture and upper troposheric humidity, J. Climate, 9, 1561-1585, 1996.

Harries J.E., The greenhouse earth : a view from space, Q.J.R.Meterol.Soc., 122, 799-818, 1996.

Manabe, S., and R.T. Wetherald, Thermal equilibrium of the atmosphere with a given distribution of relative humidity, J. Atmos. Sci., 24, 241-259, 1967.

Morcrette J-J. and Fouquart Y., On systematic errors in parametrized claculations of longwave radiation transfer, Q.J.R.Meterol.Soc., 111, 691-708, 1985.

Picon L. and Desbois M., Relation between METEOSAT water vapor radiance fields and large-scale tropical circulation features, J. Climate, 3, 865-876, 1990.

Picon L, S. Fongang, G. Seze, and M. Desbois, African and atlantic short-term climatic variations described from Meteosat water vapor channel, Ann. Geophysicae, 13, 768-781, 1995.

Picon L., J-L. Monge, R. Roca, M. Desbois, J-L. Reynes and V. Saudemon, Investigation of the homogeneity of water vapor data for long-term climatic studies, Contract EUM/CO/97/512/HW, EUMETSAT Publications, Darmstadt, 1998.

Picon L., S. Serrar, M. Desbois, J-L. Monge, and R. Roca, Homogeneity of METEOSAT water vapor data from 1983 to 1994, Contract EUM/CO/98/606/HW, EUMETSAT Publications, Darmstadt, 1999.

Pierrehumbert R.T., Thermostats, radiators and the local runaway greenhouse, J. Atmos.Sci., 52, 1784-1806, 1995.

Polcher J. and K. Laval, A statistical study of regional impact of deforestation on climate of the LMD-GCM, Climate Dyn., 10, 205-219, 1994.

Roca R., Contributions \`a l'\'etude de la vapeur d'eau et de la convection et de leurs interactions dans les Tropiques, Thése de doctorat de l'Universit\'e Paris VII Denis Diderot, To be defended in Nov. 1999.

Roca R., L. Picon, M. Desbois, H. Letreut and J-J Morcrette, Direct comparison between Meteosat water vapor channel and GCM results, Geophys. Res. Lett., 24, 147-150, 1997.

Roca R. and L. Picon, GCM validation and climatic study using Meteosat water vapor data, Proceedings of the 1997 EUMETSAT Meteorological satellite data users's conference, Brussels,  1997.

Soden B.J. and R. Fu, A satellite analysis of deep convection, upper tropospheric humidity and the greenhouse effect, J. Climate, 8, 2333-2351, 1995.

Soden B.J and Bretherton F.P., Evaluation of water vapor distribution in general circulation models using satellite observations, J. Geophys. Res., 99, d1, 1187-1210, 1994.

Schmetz J. and Turpeinen O.M., Estimation of the upper tropospheric relative humidity from Meteosat water vapor image data, J. Appl. Meteorol., 27, 889-899, 1988.

Schmetz J. and Van de Berg L., Upper troposheric humidity observations from Meteosat compared with short-term forecast fields, Geophys. Res. Lett., 21, 573-576, 1994.

Schmetz, J., and coauthors, Satellite observations of upper tropospheric relative humidity, clouds and wind fields divergence, Beitr. Phys. Atmosph., 68, 345-357, 1995.

LLNL Disclaimers

UCRL-MI-127350