Application of Clustering Algorithms to Observed and Simulated Daily Precipitation over the Tropical and Southern Pacific Ocean

We apply two data organization methods, self-organizing maps (SOMs) and k-means clustering with linear unidimensional scaling (k-means+LUS), to identify and organize the spatial patterns inherent in daily austral summer (December-January-February or DJF) rainfall over the tropical and southern Pacific Ocean basins from Tropical Rainfall Measuring Mission (TRMM) satellite observations.  For either a 2x2 SOM or k = 4 clustering of all available DJFs from 1998-2013, we find an El Niño/Southern Oscillation (ENSO) signature, with pairs of maps reflecting either El Niño or La Niña phase conditions.  Within each of the ENSO-phase pairs, one map favors Intertropical Convergence Zone (ITCZ)-active conditions, in which precipitation is more intense over the ITCZ region compared to the South Pacific Convergence Zone (SPCZ) region, while the remaining one is SPCZ-active.  The SPCZ-active maps show a spatial translation of the principal SPCZ diagonal consistent with the impacts of El Niño/Southern Oscillation (ENSO) or analogous low-frequency modes of variability on the SPCZ as shown in prior studies.  Because of the dominant impact of ENSO, we further apply these methods separately on subsets of rainfall data for each ENSO phase.  While the overall position of the SPCZ is sensitive to the phase of ENSO, within each phase, more- or less-steeply sloped SPCZ diagonals may occur.  Thus, while the mean position of the SPCZ is largely controlled by ENSO phase, the distinct orientations of the SPCZ within the same ENSO phase point to higher-frequency modulation of SPCZ slope.  To investigate the nature of these further, we construct composites of pressure-level winds and specific humidity from the Climate Forecast System Reanalysis (CFSR) associated with the rainfall patterns.  For either SOM or kmeans-based composites, we find large-scale dynamics and moisture signatures that are consistent with the rainfall patterns and which we interpret in terms of previously described mechanisms of SPCZ variability.   By progressively increasing the number of clusters, patterns reminiscent of Rossby wave propagation begin to emerge. To further investigate the connection to propagation, we examine upper air vorticity composites in relationship to the periodic enhancements of SPCZ precipitation which appear to be independent of ENSO.