Characterization of daily profiles of wind patterns in a subtropical (34° S) coastal zone using polar coordinate reconstruction

Resumen: Understanding wind patterns in the coastal regions are relevant for oceanographic, infrastructure and ecological processes. Wind patterns are vectorial quantities that require special mathematical treatment. Here, using an algorithm based on a polar coordinate reconstruction based on a projection on the euclidean space, we developed a novel algorithm to cluster daily wind patterns. Wind intensity in knots and directions in degrees were measured every 5 minutes with a meteorological station located in the Southweastern Atlantic in La Paloma harbour by the National Hydrographic and Oceanographic Service (SOHMA) from 1st January 2021 to 2nd September 2023. We analyzed 220572 records of wind data from 770 days. We projected directional data in the cylinder and estimated dissimilarity among days for each 5-minutal wind measure and then we averaged the dissimilarity between each pair of days. In doing so we estimated an average daily distance matrix of wind patterns. Using the dissimilarity matrix, we conducted a hierarchical cluster algorithm using the Ward metric as implemented in the stats package under the hclust function using “ward.D2” method in R. We selected ten groups characterizing daily profiles of wind patterns. North East (NE) winds and South West (SW) winds were predominant with intensities ranging from calm days (no wind) to a maximum of 44 knots in the passage of an extratropical cyclone. Daily patterns showed variation: for example, one cluster included the typical sea breeze regime, winds blowing from the North (N) in the morning and intensifying in the afternoon and rotating to the NE. Another cluster included the passage of a cold front (SW) in the noon, with the rotation of wind from the N to the SW and intensifying. Another cluster with intense winds from the South all day long were also common. The transition matrix showed that two clusters persist from one day to another (~36 to 50% of times respectively), two never repeat from one day to another and most of the clusters (N=5) presented a ca. 20% of persistence. Results provide a detailed description of wind patterns and to our knowledge the first implementation of the algorithm for wind classification that allows to characterize wind profiles integrating intensity and speed. Implications for oceanographic processes, biological, sailing and coastal infrastructure and predictability of weather forecast are discussed.