Identifying patterns of climate variability from principal component analysis - PCA, Fourier y k-means clustering

Context: Is achieved a research through Principal Component Analysis (PCA) for determining the variability and climate patterns of two important cities in the Colombia Caribbean. Method: This research used satellite data with three hourly resolution contained in a 35 year data set (1980 to 2014), and a spatial scaling was performed using information related to Cartagena and Barranquilla cities, located in the north of Colombia. Results: The correlation results, above 80 %, show an appropriate adjustment for the information analysis of wind speed and temperature. Time lag matrixes were built for the time series with the aim of applying the Principal Component Analysis (PCA), known as Singular Spectrum Analysis. The main components were identified, which represent more than 70% of the temperature and the wind data in both cities. A Fourier analysis for the wind speed and the temperature allowed identifying similar oscillation modes (main components) detected by the PCA. Conclusions: Sea and land breezes explain the identified diurnal temperature and wind speed variability in Cartagena. Additionally were observe a quarterly variability associated with fluctuations Maden Julian, semiannual, annual, and 6-year variability associated with ENSO. Finally, the cluster analysis allowed the identification of two-climate pattern in the study area.