Delineation of the Topoclimates of Eastern Cuba by Local Weather Patterns and Unsupervised Machine Learning

Abstract

Abstract In this study, the principles of complex climatology were applied to delineate topoclimates in the mountains of eastern Cuba. A regional numerical weather model, driven by reanalysis, was used to obtain temperature patterns at a resolution of 0.6 km. Unsupervised machine learning techniques were then utilized to identify weather types based on temperature during rainy and less rainy periods, as well as geographical location. This analysis was supplemented with historical precipitation and geographical data to identify 23 topoclimates in the study area. The methods allowed for the estimation of topoclimate identification errors and provided insights into the representativeness of surface meteorological stations for the study area. The results showed that the rainfall distribution of the identified topoclimates was consistent with that of local climate-forming factors and previous research. Furthermore, new insights into the climatological rainfall characteristics of the lower and middle heights of the mountains in eastern Cuba were identified.

Keywords

• topoclimates
• tropical islands
• complex climatology
• machine learning
• autoorganized networks