The project is aimed at developing statistical learning techniques that can be applied to satellite data in order to characterize the time-dependent structure and dynamics of an atmospheric feature known as the Inter-Tropical Convergence Zone (ITCZ), an important part of the large-scale atmospheric circulation. The information gathered by applying the statistical learning techniques will be used to improve the scientific understanding of ITCZ dynamics and breakdown.
One goal of the project is to develop statistical methods that identify dynamical features in satellite data for the tropics. These data include visible and infra-red imagery, scatterometer winds (QuikSCAT data), rainfall estimates from the Tropical Rainfall Monitoring Mission satellite, and liquid cloud water content from microwave measurements, together with derived quantities such as relative vorticity fields. Of particular interest are the shape, location and intensity of relative vorticity pools within the ITCZ, thought to be related to ITCZ breakdown. These parameters will first be estimated from instantaneous two-dimensional data fields, primarily the combination of the QuickSCAT data and visible and infra-red data from the GEOS satellite. The project will then go on to identify these parameters in time sequences of imagery using several statistical techniques. Parameter estimation within the ITCZ will be extended to include the estimation of non-ITCZ disturbances which may affect ITCZ dynamics, such as "westward-propagating disturbances". The algorithms developed will be tested by applying them to unseen data and comparing the results to human-produced annotation of the images. Initial development will use fields generated by numerical model simulations of ITCZ dynamics, including simulations of breakdown by internal vorticity-mode instabilities as well as ITCZ breakdowns induced by interaction with westward-propagating disturbances. The sequences of maps of ITCZ structure will be made available via the web as a data product.
Once developed, the statistical techniques just described will be used to produce sequences of ITCZ state that will then be used to determine the different modes of breakdown exhibited by the ITCZ and their relative frequency. There will also be an examination of the relative frequency of shallow and deep ITCZs, and the influence of tropical disturbances on meridional mixing, as well as a search for evidence of whether Madden-Julien oscillations influence ITCZ dynamics, including breakdown. In the later stages of the project, the statistical learning techniques will be applied to high-resolution atmospheric model output. The statistics on ITCZ states derived from the general circulation model will be compared with those derived from the satellite data to examine quantitatively whether there is any systematic bias in the behavior of the ITCZ in the atmospheric circulation model. The researchers will also attempt to derive state sequences of the ITCZ from GOES data alone, comparing the results with those obtained from the combination of GOES satellite data and scatterometer wind data. If this is successful, the evolution of ITCZ dynamics will be studied over the period 1979-2005, using the raw GOES data as input to the analysis. This will permit a search for correlations between variations in ITCZ dynamics and decadal climate variability.
This project includes education and training opportunities for a post-doctoral investigator and a graduate student. The techniques developed may be applicable to other types of satellite and non-satellite imagery. Dissemination of results will be by the standard methods of presentations at scientific meetings and in peer-reviewed journal articles; however data products developed by the project will be made publicly available via a web-site.
