Determination of new national groundwater monitoring sites using artificial neural network model in South Korea.

South Koreas National Groundwater Monitoring Network (NGMN) was installed during the early 1990s to detect the trends in groundwater levels; however, a revised NGMN plan is required in consideration of climate change and human activities. Linear trend analyses of the groundwater levels, Cl concentration, and NO₃-N concentration of 521 monitoring wells were conducted to define the downward and upward trends. The attributes of 14 items relating to topography, stream proximity, land use, soil, hydrogeology, and well density were extracted from thematic maps for each monitoring site. An artificial neural network (ANN) model for groundwater level trends was developed using these 14 input variables to predict the downward (output variable: 1) and non-downward (output variable: 0) trends at any grid point in a standard watershed. Candidate sites for new groundwater monitoring wells were suggested based on the probability of the existence of two trends for groundwater levels. Candidate sites were excluded if they showed upward trends of Cl and NO₃-N because the primary objective of the NGMN was not to observe changes in water quality but to observe the background conditions of water quality. It was proposed to install a total of 1475 groundwater monitoring wells (existing plus new wells) by 2045, and the percentage contributions of non-downward and downward trends of groundwater levels to the total number of trends (i.e., wells) were projected to be 61.5% and 38.5%, respectively. The NGMN will play an important role in recognizing climate change, observing groundwater level declines caused by human activities, and assessing the relationship between surface water and groundwater in standard watersheds. [ABSTRACT FROM AUTHOR]