| Title |
Characteristics Detection of Hydrological and Water Quality Datain Jangseong Reservoir by Application of Pattern Classification Method |
| Authors |
박성천 ( Sung Chun Park ) ; 진영훈 ( Young Hoon Jin ) ; 노경범 ( Kyong Bum Roh ) ; 김종오 ( Jong O Kim ) ; 유호규 ( Ho Gyu Yu ) |
| Keywords |
Data characteristics; Hydrological data; Pattern classification; Reservoir; Self Organizing Map (SOM); Water quality |
| Abstract |
Self Organizing Map (SOM) was applied for pattern classification of hydrological and water quality data measured at Jangseong Reservoir on a monthly basis. The primary objective of the present study is to understand better data characteristics and relationship between the data. For the purpose, two SOMs were configured by a methodologically systematic approach with appropriate methods for data transformation, determination of map size and side lengths of the map. The SOMs constructed at the respective measurement stations for water quality data (JSD1 and JSD2) commonly classified the respective datasets into five clusters by Davies-Bouldin Index (DBI). The trained SOMs were fine-tuned by Ward`s method of a hierarchical cluster analysis. On the one hand, the patterns with high values of standardized reference vectors for hydrological variables revealed the high possibility of eutrophication by TN or TP in the reservoir, in general. On the other hand, the clusters with low values of standardized reference vectors for hydrological variables showed the patterns with high COD concentration. In particular, Clsuter1 at JSD1 and Cluster5 at JSD2 represented the worst condition of water quality with high reference vectors for rainfall and storage in the reservoir. Consequently, SOM is applicable to identify the patterns of potential eutrophication in reservoirs according to the better understanding of data characteristics and their relationship. |