Development of a machine learning-based reconstruction method for anomalous and missing water-level data using nearby gauge observations

Jaehwan Yoo; Je yun Lee; Hyuntak Jo; Byunghyun Kim

doi:10.3741/JKWRA.2026.59.8.827

All Issue

2026 Vol.59, Issue 8 Next Page

Development of a machine learning-based reconstruction method for anomalous and missing water-level data using nearby gauge observations 인근 관측자료와 기계학습을 활용한 수위자료의 오·결측 보완 기법 개발

31 August 2026. pp. 827-841

PDF XML

Abstract

As the risk of extreme rainfall and flooding increases under climate change, the need for reliable water-level data and effective data correction has become more important. This study investigated the applicability of machine learning methods for reconstructing anomalous and missing water-level data using nearby gauge observations and rainfall data at Gyeongho Bridge and Susan Bridge in the Nam River basin. Long Short-Term Memory (LSTM), Deep Neural Network (DNN), and Random Forest-based prediction models were developed using data from 2019 to 2023 for training and 2024 data for validation. The results showed that the LSTM model had poor predictive performance and was excluded from the final comparison. Between the DNN and Random Forest models, the Random Forest model showed the best overall performance. In addition, areal mean rainfall generally provided more stable prediction results than cumulative rainfall. In anomalous cases, the Random Forest model reconstructed the rising and recession limbs of the hydrograph more naturally and continuously. Multiple regression analysis based on the 2023 observed data also showed high explanatory power, although it had limitations in reproducing abrupt nonlinear water-level variations. These findings demonstrate that reconstruction based on hydrologically connected nearby gauges can be effectively applied in practice.

Keywords

Water-level data

Anomalous and missing data

Nearby gauge observations

DNN

Random forest

기후변화에 따른 극한 강우와 홍수위험 증가로 인해 신뢰도 높은 수위자료의 확보와 보정 필요성이 커지고 있다. 본 연구에서는 남강 유역의 경호교와 수산교를 대상으로 인근 관측소 수위자료와 강우자료를 활용하여 수위자료의 오·결측 구간을 보완하는 기계학습 기법의 적용성을 검토하였다. 이를 위해 장단기기억 신경망(Long Short-Term Memory, LSTM), 완전연결형 다층 퍼셉트론 구조의 DNN (Deep Neural Network), Random Forest 기반 예측모형을 구축하고 2019~2023년 자료를 학습에, 2024년 자료를 검증에 사용하였다. 분석 결과, LSTM은 예측 성능이 낮아 최종 비교대상에서 제외되었으며, DNN과 Random Forest 중에서는 Random Forest가 가장 우수한 성능을 보였다. 또한 강우 입력자료로는 누적강우량보다 면적평균강우량이 대체로 더 안정적인 예측 결과를 나타냈다. 오측 의심 사례 비교에서도 Random Forest는 수위 상승부와 감수부를 보다 자연스럽고 연속적으로 복원하였다. 한편, 2023년 관측자료를 이용한 다중회귀분석은 높은 설명력을 보였으나 급격한 비선형 수위 변동 재현에는 한계가 있었다. 본 연구는 인근 관측소 간 수문학적 연계성을 활용한 자료 복원이 실무적으로 유효함을 보여준다.

키워드

수위자료

오·결측 자료

인근 관측소

DNN

Random forest

References

Breiman, L. (2001). “Random forests.” Machine Learning, Vol. 45, No. 1, pp. 5-32.

10.1023/A:1010933404324

Cho, M., Kim, C., Jung, K., and Jung, H. (2022). “Water level prediction model applying a long short-term memory (LSTM)-gated recurrent unit (GRU) method for flood prediction.” Water, Vol. 14, No. 14, 2221.

10.3390/w14142221

Han River Flood Control Office (HRFCO) (2026a). Water Resources Management Information System (WAMIS), accessed 26 March 2026, <https://www.wamis.go.kr/>.

Han River Flood Control Office (HRFCO) (2026b). Help for hydrologic terms and definitions, accessed 26 March 2026, <https://www.wamis.go.kr/dic/dic_main.do>.

Han River Flood Control Office (HRFCO) (2026c). Water level correlation data by river, accessed 26 March 2026, <https://www.wamis.go.kr/wkw/wl_dubwlobs.do>.

Hochreiter, S., and Schmidhuber, J. (1997). “Long short-term memory.” Neural Computation, Vol. 9, No. 8, pp. 1735-1780.

10.1162/neco.1997.9.8.1735

Hong, S.M., and Jang, A. (2017). “The development study on the integrated management system for water information based on ICT.” Journal of Korean Society of Environmental Engineers, Vol. 39, No. 12, pp. 723-732.

10.4491/KSEE.2017.39.12.723

Hyndman, R.J., and Koehler, A.B. (2006). “Another look at measures of forecast accuracy.” International Journal of Forecasting, Vol. 22, No. 4, pp. 679-688.

10.1016/j.ijforecast.2006.03.001

Intergovernmental Panel on Climate Change (IPCC) (2023). Climate change 2023: Synthesis report. Geneva, Switzerland.

Jang, H., Yoo, J., Lee, C., and Kim, B. (2025). “HEC-RAS 2D-based compound flood modeling in the Naengcheon Basin, a coastal urban area.” Journal of Korea Water Resources Association, Vol. 58, No. S-4, pp. 1617-1627. doi: 10.3741/JKWRA.2025.58.S-4.1617.

10.3741/JKWRA.2025.58.S-4.1617

Kim, B.J., Kim, M., Yoo, J., and Kim, B. (2025). “Rapid simulation for real-time flood depth prediction using support vector machine.” Scientific Reports, Vol. 15, 31818. doi: 10.1038/s41598-025-17090-2.

10.1038/s41598-025-17090-240883470PMC12397246

Kim, D., Han, H., Wang, W., and Kim, H.S. (2022). “Improvement of deep learning models for river water level prediction using complex network method.” Water, Vol. 14, No. 3, 466.

10.3390/w14030466

Korea Water Resources Corporation (K-water) (2025a). Operation information of rainfall and water level stations, accessed 26 March 2026, <https://www.data.go.kr/data/15099115/openapi.do>.

Korea Water Resources Corporation (K-water) (2025b). Specification information of water level stations, accessed 26 March 2026, <https://www.data.go.kr/data/15049847/fileData.do>.

Korea Water Resources Corporation (K-water) (2025c). Hourly stage and discharge data by station, accessed 26 March 2026, <https://opendata.kwater.or.kr/open/data/list/list.do>.

Korea Water Resources Corporation (K-water) (2026). Namgang Dam. Nakdong River basin headquarters, accessed 26 March 2026, <https://www.kwater.or.kr/busi/sub02/facilitiespresent03Page.do?s_mid=1515>.

Kvålseth, T.O. (1985). “Cautionary note about R².” The American Statistician, Vol. 39, No. 4, pp. 279-285.

10.1080/00031305.1985.10479448

LeCun, Y., Bengio, Y., and Hinton, G. (2015). “Deep learning.” Nature, Vol. 521, pp. 436-444.

10.1038/nature14539

Ouma, Y.O., Moalafhi, D.B., Anderson, G., Nkwae, B., Odirile, P., Parida, B.P., and Qi, J. (2022). “Dam water level prediction using vector autoregression, random forest regression and MLP-ANN models based on land-use and climate factors.” Sustainability, Vol. 14, No. 22, 14934.

10.3390/su142214934

Willmott, C.J., and Matsuura, K. (2005). “Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance.” Climate Research, Vol. 30, pp. 79-82.

10.3354/cr030079

Yi, J., Kim, M.M., and Park, S.J. (2009). “The study on the integrated monitoring of water quantity and quality data.” Journal of the Korean Society of Hazard Mitigation, Vol. 9, No. 5, pp. 115-123.

Yoo, J., Kim, J.J., Kim, C.H., and Kim, B. (2024a). “Analysis of flood damage mitigation effects based on design rainfall before and after sewer improvement.” Journal of the Korean Society of Safety, Vol. 39, No. 6, pp. 61-69. doi: 10.14346/JKOSOS.2024.39.6.61.

10.14346/JKOSOS.2024.39.6.61

Yoo, J., Lim, B., Shin, J., and Kim, B. (2024b). “Case study on urban water cycle improvement in Andong City using the SWMM-LID model.” Journal of Korea Water Resources Association, Vol. 57, No. 11, pp. 897-910. doi: 10.3741/JKWRA.2024.57.11.897.

10.3741/JKWRA.2024.57.11.897

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 :8
Pages :827-841
Received Date : 2026-04-07
Revised Date : 2026-06-03
Accepted Date : 2026-06-05
DOI :https://doi.org/10.3741/JKWRA.2026.59.8.827

Journal of Korea Water Resources Association ISSN:2799-8746(Print) 2799-8754(Online) 한국수자원학회 논문집

All Issue