AMIP II Diagnostic Subproject 36

Water Vapor and Cloud Feedback Processes

Project coordinators:
Ming-Dah Chou1,  Curt Covey2, Arthur Y. Hou1, Richard S. Lindzen3 , De-Zheng Sun4

1NASA GSFC, Greenbelt MD
2Program for Climate Model Diagnosis and Intercomparison
3Masacusetts Institude of Technology
4Cooperative Institute for Research in Environmental Sciences (CIRES)






Background
Objectives
Methodology
Data Requirements
References

Background

Feedback processes within the climate system that involve clouds and water vapor exert a dominant influence on the way the climate responds to external forcing. Coupled ocean-atmosphere models typically exhibit strong positive water vapor feedback, amplifying the warming effect of enhanced greenhouse gas concentrations (e.g., anthropogenic CO2). Such models also exhibit a range of positive and negative cloud feedbacks depending on details of the models' parameterizations of convection and microphysics. Consequently, model-based predictions of global warming include both a strong amplifying effect of water vapor feedback and a significant uncertainty range due to different possible cloud feedback processes. It has proved difficult to narrow this uncertainty range, to confirm (or refute) positive water vapor feedback in the real-world climate sysem, or even to evaluate the extent to which the two types of feedback are separable. Progress toward these objectives requires a wide variety of model diagnostics, especially those involving comparison of model-simulated quantities with observational data.

We propose to examine AMIP II model output using two new diagnostics related to cloud and water vapor feedback. Two existing AMIP II subprojects are examining the general characteristics of model-simulated clouds and water vapor: No. 13, "Evaluation of Global Cloudiness" (B. Weare), and No. 27, "Tropospheric Humidity and Meridional Moisture Fluxes" (D. Gaffen, R. Rosen D. Salstein, J. Boyle, B. Soden). We do not anticipate that these subprojects will employ the diagnostics proposed below, but we will maintain close contact with Weare and Boyle to prevent duplication of effort. An additional AMIP II subproject -- No. 34, "Evaluation of Convection and Upper Level Moisture and their Links Using Meteosat Water Vapor Channel Data" (R. Roca and L. Picon) -- will use specialized diagnostics that appear to be largely unrelated to ours.

Objectives

We propose to compare AMIP II model simulations with observations during the AMIP II period (1979-1996) of the following tropical phenomena:
  1. Correlations involving interannual variations of water vapor and temperature
  2. Correlations between cloud area and underlying sea surface temperature
Item (1) has been investigated in several studies, but typically results from only one or two models have been considered in each study. Item (2) has not previously been investigated with the particular method discussed below.

Methodology

For Item (1) we will repeat the major diagnostics employed by Sun and Held (1996) on the complete set of AMIP II model simulations. For example, Figure 9 in Sun and Held's paper indicates that correlations between interannual variations of water vapor at different atmospheric levels are much stronger in their model than in the real world. Our preliminary results from the public-domain AMIP I database reveal similar results from all of these models (see calc_r.ps, available on the anonymous FTP server sprite.llnl.gov in subdirectory pub/covey/IPCC). The discrepancy between models and observations appears to be independent of model characteristics such as vertical resolution or convective parameterization. On the other hand, the AMIP I model family is several years old, and it will be interesting to see if the more recent models from AMIP II give significantly different results when this diagnostic (and others from Sun and Held) are applied to their output.

It should be noted that the observational data employed by Sun and Held is open to criticism. The observations were obtained from radiosondes whose high-altitude measurements are questionable, and whose coverage of the tropics is sparse. In our subproject we will employ satellite and reanalysis data where possible, in addition to updated radiosonde data. When comparing with the radiosonde data, we will use a subset of model output that matches the space-time coverage of the observations. Del Genio (personal communication) has found that this technique reduces the disagreement between the radiosondes and the GISS GCM.

For Item (2) we will extend our current work, which compares recent satellite observations of the tropical Pacific (30S-30N, 130E-170W; Jan 1998 - Aug 1999) with simulations by a GCM including the NCAR CCM3 subgridscale physics. In the observations we find a large, robust negative correlation between high elevation cloud cover and cloud-weighted SST. Our use of cloud-weighted rather than actual SST was motivated by our assumption that cloud area is proportional to ice detrainment by cumulus, which depends on the temperature dependent stochastic coalescence in the cumuli. In that case, the only temperature relevant to the process is the temperature as seen by the clouds. For the periods studied, most of the change in cloud weighted SST comes from movement of the clouds rather than actual changes in SST.

We have found that our model simulations completely fail to replicate the observed cloud area dependence. This result is not surprising since CCM3 physics uses diagnostic clouds. However, several AMIP II models (and indeed several AMIP I models) use prognostic cloud schemes. We therefore wish to repeat our cloud area / cloud-weighted SST study with the complete AMIP II results. Although the AMIP II period does not overlap the period of our observational data, we believe that for a given model, any 1-2 year period from the full AMIP II time span will give results similar to those from any other 1-2 year period. We will test this assumption by examining several 1-2 year periods for each model.

Data requirements

For Item (1) we require monthly mean air temperature and specific humidity (ta and hus). For Item (2) we require monthly mean SST (ts) and high-time-frequency cloud amounts. If the optional 6-hourly total cloud cover (clt) has been supplied by a sufficient number of AMIP II models, then we will use this quantity. Otherwise we will deduce cloud amounts as best we can from 6-hourly OLR (rlut). We will also examine the analogous AMIP I output for Item (2), as we already have for Item (1).

References

D.-Z. Sun and I. M. Held, 1996: "A Comparison of Modeled and Observed Relationships between Interannual Variations of Water Vapor and Temperature", Journal of Climate, 9: 665-675