Geographical and synoptic drivers of climate and bioclimate in the Carpathian region

The Carpathian region exhibits distinctive climate and bioclimate patterns that are influenced by various geographical and synoptic drivers. This study aims to explore the spatial and altitudinal distribution of climate and bioclimate in the region while examining their relationship with atmospheric circulation. By employing a combination of clustering, regression, machine learning, correlation, and interpolating anomalies on transects techniques, I investigate the climate and bioclimate dynamics in this mountainous region. In conducting this research, gridded climate variables are utilised from the CarpatClim project dataset, including daily meteorological data such as cloud cover, precipitation, and air temperature. Universal Thermal Climate Index dataset is created and applied to assess thermal stress. Geographical variables, namely latitude, longitude, and elevation, are obtained from the CarpatClim model, while circulation indices (strength, vorticity, and direction of airflow) and types serve as independent variables for describing atmospheric circulation patterns. To objectively classify the climate within the Carpathian region, dynamic time warping and partitioning around medoids are employed based on multivariate time series of air temperature, diurnal temperature range, and precipitation. The resulting clustering is evaluated using the Caliński-Harabasz Index and Silhouette Coefficient. Principal Component Analysis is used to reduce the dimensionality of the data and visualise it on 2D plots. Locally weighted polynomial regression and correlation analyses are conducted to examine the relationships between circulation indices and climate variables. Furthermore, the machine learning ensemble algorithm, boosting regression, is applied to predict the generalised extreme values distribution of climate and bioclimate extremes and determine the relative importance of geographical and synoptic drivers. Climate and bioclimate anomalies are visualised on gridded maps and interpolated transects due to circulation types. The findings reveal the existence of eight distinct local climate types in the Carpathian region. The study demonstrates that dividing the region into 5-6 clusters yields similar results to the well-known Köppen-Geiger climate classification. Accurate clustering outcomes depend on careful considerations of clustering algorithms, distance measures, and variable normalisation. Acknowledging the limitations of climate clustering, this study proposes the potential use of segmentation techniques to create continuous spatial units. The ensemble boosting regression model achieves root mean squared error values ranging from 0.66 to 2.67. Each driver’s relative importance (weight, gain, and cover feature importance), including month, latitude, longitude, strength, vorticity, and direction of airflow is presented. The significance of geographical factors such as elevation and latitude and synoptic drivers like airflow strength, vorticity, and direction are analysed and discussed in relation to climate variables. The study highlights the importance of spatial and temporal factors and atmospheric circulation patterns in predicting extreme climate events. The findings contribute to an improved understanding of climate dynamics and hold the potential to enhance forecasting and climate modelling efforts. Furthermore, this study offers insights into the relationships between circulation types and climate patterns, modifications to the atmospheric classification scheme, and emphasises the importance of considering airflow indices in future research. Overall, this study enhances our understanding of the complex interactions between the Carpathian region’s geography, atmospheric circulation, climate, and bioclimate. It provides valuable insights that can guide further research and contribute to improved regional climate assessments and predictions.