Search for: All records

Abstract. This study proposes a novel structural self-organizingmap (S-SOM) algorithm for synoptic weather typing. A novel feature of theS-SOM compared with traditional SOMs is its ability to deal with input datawith spatial or temporal structures. In detail, the search scheme for thebest matching unit (BMU) in a S-SOM is built based on a structuralsimilarity (S-SIM) index rather than by using the traditional Euclideandistance (ED). S-SIM enables the BMU search to consider the correlation inspace between weather states, such as the locations of highs or lows, that is impossible when using ED. The S-SOM performance is evaluated by multipledemo simulations of clustering weather patterns over Japan using theERA-Interim sea-level pressure data. The results show the S-SOM'ssuperiority compared with a standard SOM with ED (or ED-SOM) in tworespects: clustering quality based on silhouette analysis and topologicalpreservation based on topological error. Better performance of S-SOM versusED is consistent with results from different tests and node-sizeconfigurations. S-SOM performs better than a SOM using the Pearsoncorrelation coefficient (or COR-SOM), though the difference is not as clear as it is compared to ED-SOM.