Development of an Urban-Groundwater-Level Prediction Model Using Public Big Data and Nonlinear Regression Analysis

Groundwater level fluctuations can lead to urban disasters, such as ground subsidence, necessitating the development of predictive measures for proactive disaster prevention. The predicted normal groundwater level can serve as a critical baseline for comparison against actual observed values, enabling the detection of sensor anomalies or analysis of impacts from nearby construction activities. In this study, a practical groundwater-level prediction model was constructed using public big data, acquired from sources such as the National Groundwater Information Center, and precipitation data. A hydrologically distinct area in Seoul was selected, and seven years of data were processed to remove outliers. A systematic statistical approach, involving time-series analyses (ACF, PACF, and CCF), was employed to identify the temporal characteristics and lagged correlations among the variables. Based on these findings, a multiple non-linear regression model was developed, whose reliability and high accuracy were verified using recent data; additionally, a minimum data requirement was established for its broader applications. This model offers a practical tool for predicting variations in groundwater level, supporting urban disaster prevention and sustainable water management.