All Issue

2026 Vol.59, Issue 10 Preview Page
Development of a combined physics and machine learning model for predicting groundwater discharge and groundwater level distribution caused by subsurface structure installation

물리 및 기계학습 결합 모형을 기반의 지하구조물 설치에 따른 지하수 유출 및 지하수위 분포 예측 모형 개발

Yubeen Junga
,
Geonwoo Kwakb
,
Hyeongbin Parkc
,
Jeong-Seok Yangd*
정 유빈a
,
곽 건우b
,
박 형빈c
,
양 정석d*
aEngineer, Korea Engineering Consultants Corp. (KECC), Seoul, Korea
bM.S. Candidate, Department of Civil & Environmental Engineering, Kookmin University, Seoul, Korea
cM.S. Candidate, Department of Civil & Environmental Engineering, Kookmin University, Seoul, Korea
dProfessor, Department of Civil & Environmental Engineering, Kookmin University, Seoul, Korea
a한국종합기술 사원
b국민대학교 건설시스템공학부 석사과정
c국민대학교 건설시스템공학부 석사과정
d국민대학교 건설시스템공학부 교수
*Corresponding Author
31 October 2026. pp. 1183-1194
PDF XML Citation
Abstract

As urban underground development expands, large-scale structures such as subway stations and deep drainage tunnels increasingly interfere with natural groundwater circulation. These alterations can generate seepage and induce fluctuations in local water tables, which in turn may trigger secondary urban issues such as land subsidence, groundwater depletion, and excessive drainage loads. To quantify and anticipate these hydrogeological responses, this study applies a physics-model-based machine learning approach, which utilizes the results of a physics-based groundwater flow model (MODFLOW) for the training of a data-driven machine learning model. The MODFLOW model reproduces groundwater flow by incorporating hydrogeological parameters including hydraulic conductivity, storage coefficients, and boundary conditions, while the machine learning model infers temporal and spatial variations in groundwater levels and seepage using both simulated and observed datasets. While existing groundwater data has limitations in estimating levels in unmeasured areas, the combination of the two models in this study demonstrated improved consistency and stability in groundwater level prediction, even under conditions with sparse observations. The results of this study provide practical foundational data that can be used to evaluate the hydrogeological effects of underground structure installation and to review drainage planning and design in urban areas. As such, this machine learning model, linked with a physical model, is expected to be utilized as a useful decision-support tool for future urban groundwater management and sustainable underground space development.

Keywords
Subsurface structure
Groundwater discharge
Groundwater level distribution
Combined physics and machine learning model
MODFLOW
Convolutional Neural Network (CNN)

최근 도시 지하공간 개발이 확대됨에 따라 지하철역이나 대심도 빗물터널과 같은 대규모 지하구조물은 자연적인 지하수 유동 체계를 교란할 수 있다. 이러한 변화는 용출수 발생과 주변 지하수위 변동을 초래하며 결과적으로 지반 침하, 지하수위 저하, 배수 부하 증가와 같은 2차 도시 환경 문제를 유발할 가능성이 있다. 본 연구에서는 이러한 지하수 영향 변화를 정량적으로 예측하기 위해, 물리기반의 지하수 유동해석 모형(MODFLOW)의 결과를 데이터 기반 기계학습 모형의 학습에 활용하는 물리 모델 기반 기계학습 접근법을 적용하였다. MODFLOW는 연구지역의 투수계수, 저류계수, 경계조건 등 수리·지질학적 매개변수를 반영하여 지하수 흐름을 재현하였으며, 기계학습 모형은 지하수위 시뮬레이션 및 관측 자료를 활용해 지하수위 및 용출 특성을 추정하였다. 기존 지하수위 데이터의 경우 미계측 지역의 지하수위 추정이 어려운 단점이 있었는데, 이번 연구에서 두 모형의 결합을 통해 관측 자료가 부족한 조건에서도 지하수위 예측의 일관성 및 안정성이 향상되는 것으로 나타났다. 본 연구의 결과는 지하구조물 설치에 따른 수리지질학적 영향을 평가하고 도시 지역의 배수계획 및 설계 검토에 활용할 수 있는 실질적인 기초 자료를 제공한다. 이처럼 물리 모형과 연계된 기계학습 모형은 향후 도시 지하수 관리 및 지속가능한 지하공간 개발을 위한 유용한 의사결정 지원 도구로 활용될 수 있을 것으로 기대된다.

키워드
지하구조물
지하수 유출
물리 및 기계학습 모형 결합
MODFLOW
CNN
References
1

Chae, B.G. (2004). “Application of homogenization method for estimating hydraulic conductivity.” Journal of Soil and Groundwater Environment, KoSSGE, Vol. 9, No. 1, pp. 79-86.

2

Chen, J., Zhang, X., Chen, X., Wang, H., Cui, J., and Liu, C. (2023). “Shallow groundwater responses to rainfall based on correlation and spectral analyses in the heilonggang region, China.” Water, Vol. 15, No. 6, 1100.

10.3390/w15061100
3

Freeze, R.A., and Cherry, J.A. (1979). Groundwater. Prentice Hall, Englewood Cliffs, NJ, U.S.

4

Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep learning. MIT Press, Cambridge, MA, U.S.

10.23000/TRKO201600014868.169
5

He, S., Wu, J., Wang, H., and Xin, P. (2021). “Quantifying the time lags of groundwater level response to precipitation at multiple timescales.” Journal of Hydrology, Vol. 598, 126257.

6

Kim, J.H., Lee, J.Y., and Park, E. (2018). “Uncertainty analysis of spatial interpolation for groundwater level mapping in sparse observation networks.” Journal of the Korean Society of Groundwater Environment, Vol. 23, No. 4, pp. 34-45.

7

Lam, L., Yu, X., and Han, D. (2020). “A review of machine learning applications in groundwater science and engineering.” Hydrogeology Journal, Springer, Vol. 28, No. 2, pp. 639-655.

8

LeCun, Y., Bottou, L., Bengio, Y., and Haffner, P. (1998). “Gradient-based learning applied to document recognition.” Proceedings of the IEEE, Vol. 86, No. 11, pp. 2278-2324.

10.1109/5.726791
9

Ministry of Land, Infrastructure and Transport (MOLIT) (2020). Manual for underground safety impact assessment. p. 123.

10

Montgomery, D.C., Peck, E.A., and Vining, G.G. (2012). Introduction to linear regression analysis. John Wiley & Sons, Hoboken, NJ, U.S.

11

Moriasi, D. N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., and Veith, T.L. (2007). “Model evaluation guidelines for systematic quantification of accuracy in watershed simulations.” Transactions of the ASABE, Vol. 50, No. 3, pp. 885-900.

10.13031/2013.23153
12

Mosavi, A., Ozturk, P., and Chau, K.W. (2021). “Flood prediction using machine learning models: Literature review.” Water, Vol. 10, No. 11, 1536.

10.3390/w10111536
13

Park, E. (2007). “A development of groundwater level fluctuations due to precipitations and infiltrations.” Journal of the Korean Society of Soil and Groundwater Environment (KoSSGE), KoSSGE, Vol. 12, No. 4, pp. 54-59.

14

Rasouli, K., Hsieh, W.W., and Schnorbus, M.A. (2022). “Evaluating the performance of machine learning models for streamflow forecasting using information criteria.” Water Resources Research, Vol. 58, No. 7, e2021WR031665.

15

Seoul Metropolitan Government (SMG) (2024). Statistics of groundwater discharge in subway stations. Open Data Portal, accessed 10 September 2025, <https://data.seoul.go.kr>.

16

U.S. Geological Survey (USGS) (2005). MODFLOW-2005, the U.S. Geological Survey modular ground-water model — the Ground-Water flow process. Techniques and Methods 6-A16, Reston, VA, U.S.

17

Van Gerven, M., and Bohte, S. (2017). “Artificial neural networks as models of neural information processing.” Frontiers in Computational Neuroscience, Frontiers Media, Vol. 11, 114.

10.3389/fncom.2017.0011429311884PMC5742181
18

Willmott, C.J., and Matsuura, K. (2005). “Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in model performance evaluation.” Climate Research, Vol. 30, No. 1, pp. 79-82.

10.3354/cr030079
19

Yang, J.S., and Ahn, T.Y. (2008). “Analysis of correlation between groundwater level and the moving average of precipitation in Kum River watershed.” Journal of Engineering Geology, KSEG, Vol. 18, No. 1, pp. 1-6.

Information
  • Publisher :KOREA WATER RESOURECES ASSOCIATION
  • Publisher(Ko) :한국수자원학회
  • Journal Title :Journal of Korea Water Resources Association
  • Journal Title(Ko) :한국수자원학회 논문집
  • Volume : 59
  • No :10
  • Pages :1183-1194
  • Received Date : 2025-10-21
  • Revised Date : 2026-07-24
  • Accepted Date : 2026-07-28
  • DOI :https://doi.org/10.3741/JKWRA.2026.59.10.1183
Journal Informaiton Journal of Korea Water Resources Association Journal of Korea Water Resources Association
Online Submission
Archives
  • scopus
  • KCI-수자원
  • NRF
  • KOFST
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close