MISTRAL project: trajectory-based classification of wind fields

摘要

To investigate the influence of different regional flow patterns on the dispersion and transport of air pollutants, a cluster analysis of atmospheric trajectories was performed to identify regional wind fields of similar transport. The goal of a pilot study was to test the sensitivity of the cluster analysis results to changes in factors as method of clustering or distance measure. Five hierarchical clustering techniques (Ward's, Single Linkage, Complete Linkage, Average Linkage Between and Average Linkage Within) were evaluated for three different distance functions to determine which combination yielded the most meaningful classification. The ratio of between- to within-cluster variances was considered the representative number for each clustering procedure and each distance measure. The mass-consistent flow field was deduced by using measurements from the diagnostic wind model CONDOR. The clustering procedure was repeated for a 4 month study using an improved version of CONDOR and terrain-following trajectories for the MISTRAL area. The results were interpreted based on the defined ratio of between-cluster to within-cluster variance.