Recent work with cumulus ensemble models by Tompkins and Craig, Zeng and Raymond, and others shows that the adjustment of the tropical atmosphere to radiative-convective equilibrium has two time scales. First, surface evaporation comes into approximate balance with precipitation within a few days. Second, the total surface entropy flux comes into balance with radiation over ten days or so. The short moisture time scale relative to the entropy time scale is a consequence of the small scale height of water vapor in the atmosphere relative to that of mass.
The short moisture time scale gives rise to a large scale dynamical instability which starts with the development of a tropospheric moisture anomaly. This (positive) anomaly immediately results in increased precipitation and convective heating. The excess heating produces upward motion, which in turn stimulates more moistening, precipitation, and heating. In dry regions subsidence caused by the corresponding heating deficit results in further drying. The net effect is an atmosphere in which ascent, deep convection, and precipitation quickly become confined to a small fractional area, with dry, suppressed conditions occurring elsewhere. This, of course, is close to the observed state of the tropical atmosphere. I call this instability ``moisture instability''.
When a numerical model with a cumulus parameterization which mimics this two-time-scale behavior is run in two-dimensional Hadley cell mode, the above-hypothesized instability occurs. In the two-dimensional case, a single, equatorially asymmetric convective cell develops, with ascent in a narrow region about 10 degrees north or south of the equator and descent elsewhere in the equatorial belt. This behavior occurs even when the sea surface temperature (SST) distribution is flat. The region of ascent resembles the intertropical convergence zone (ITCZ). Moisture instability thus explains why there is typically only one ITCZ and why it is generally located off the equator, even in regions of nearly flat SST, such as the western Pacific.