Characterizing the synoptic timescale ITCZ in the eastern to central Pacific

The Intertropical Convergence Zone (ITCZ) in the eastern to central Pacific is a feature that is frequently observed daily in the summer half year (May through October). It exhibits strong modes of variability throughout its synoptic time-scale life cycle. It may undulate, break down, and re-form at highly variable time intervals within a ½ – 3 week period. The morphic nature of the ITCZ makes it difficult to define life stages objectively, and track its constitutive disturbances automatically. In addition, several fundamental properties of the synoptic ITCZ are poorly understood, for example, it is not clear if there is a ``typical'' ITCZ life cycle, it is not known if there is (strong) interaction with extratropical synoptic variability such as Rossby wave breaking, or if there is interaction with climate patterns (or spatial patterns of low-frequency variability) within and outside the tropics. Interannual variability of the phenomenon is not known.

In this talk, recent and current work on the E Pacific ITCZ at UC Irvine will be described. A composite picture obtained by spectral methods applied to 850hPa vorticity in 23 years of daily ERA-40 reanalysis data will be presented. The obvious drawback to this approach is that it does not describe the dynamic nature of the ITCZ (its variability) and the disturbances that form upon breakdown of the ITCZ cannot be separated out. Our aim is to develop an objective method to learn and model temporal evolution of the ITCZ. An approach using a spatiotemporal Markov Random Field (MRF) is being developed using four different meteorological fields as input data and producing an ITCZ classification (present vs. not present) as output. The meteorological fields are +25-year timeseries of GOES visible (one full scene per day) and infrared (3-hourly) images, along with satellite derived total precipitable water (once per day) and ERA-40 850hPa vorticity (6-hourly). The MRF is trained on labeled images that were obtained in two different seasons and is structured to model the different temporal resolution of the input fields as well as imperfect observations.