ScholarWorks@경상국립대학교

초록

This research aims to identify spatial and temporal variations in groundwater quality, spanning inland and coastal areas in western Gyeongnam province. The groundwater quality samples were collected randomly at 1,466 sites between January 2013 and December 2018, and were characterized using eight parameters. The resulting data were then input into a self-organizing map (SOM), which was an algorithm that excelled in data visualization and clustering. We did not find a significant relationship between the observed variables, which might be attributed to differences in local pollutant sources and geological characteristics in these large study areas. An output map with 15 rows and 13 columns was divided into six homogeneous subgroups, which were based on the similarity between their spatial and temporal profiles rather than only their temporal profiles. The SOM also showed good prediction performance for testing data sets that contained 2.5%, 5%, and 10% of unknown samples, although its performance was degraded progressively in response to increasing the number of samples used in the prediction. Overall, this study demonstrates that a SOM is used to effectively account for complex patterns of groundwater quality as well as to establish an effective groundwater management plan based on the areas of concern addressed in this study.

키워드