Analisis categorico de la predictabilidad de la precipitacion con base en la interaccion oceanoatmosfera.

In dealing with the effect of the oceaSin informaciontmosphere interaction over the clime of the americas, most statistical studies focus on modeling time series, fitting univariate or multivariate aRMa models in time domain or using Fourier or wavelet analysis for frequency domain. However, common practice has shown that the primary interest is to determine the probability of occurrence of El Niño or La Niña (see for example http://portal.iri.columbia.edu/portal/server.pt?open=512&objID=945&PageID=0&cached=true&mode=2&user ID=2). This interest summarizes in determining if a given year or season will or will be not an “El Niño (La Niña) year” and from there define strategies according to the consequences that such events carry over the climate. In lieu of such a binary alternative, we focus on the analysis on categorized time series related to the El Niño, including, sea surface temperature (SST), sea level pressure (SLP) and rainfall anomalies. For the categorical analysis, we also categorize the series into terciles considering a transition between El Niño and La Niña events and, for rainfall, the three alternatives dry/normal/wet. The objective of the study is to find or develop an adequate statistical methodology (for discrete valued observations) to model the relationships between the occurrence of El Niño (La Niña), and the rainfall variability over the west coast of Central and South america, which also have been categorized according to their intensity given by any particular group of percentiles. Data for the study will consist on discrete-valued time series of SST, SLP and rainfall for different latitudes along the west coast of Central and South america, which will be categorized following binary and terciles schemes of categorization, and a third categorization will consider the intensity of the events according with the classification of Quinn et al. (1987), which will provide up to seven different categories, from strong El Niño, to weak La Niña events. Non symmetrical, two and three way contingency tables will be considered, considering as a third variable, time lagged categorizations of the predictor variables. Inclusion of a fourth category will be also evaluated. Data will be analyzed using generalized linear statistical models. We plan to obtain a latitudinal mapping of the association between atmospheric and oceanographic variables and rainfall for the West coast of Central and South america, based on integrated data for rainfall over a 2.5x2.5 degrees grid running along the coast. Contingency tables will also be analyzed using nonsymmetrical correspondence analysis as complementary tool to evaluate the significance of the interactions, which will be primarily tested through likelihood ratio tests. as a result of the analysis, we expect to determine the odds of occurrence of the events, both as a binary response, and considering the intensity of the events, for the different latitudes starting at 10 oN, through 35 oS.