All Issue

2025 Vol.58, Issue 12S-4 Preview Page

Special Issue: 지능형 도시홍수 예측

Lead-time-dependent predictive performance of a MIMO-LSTM model for urban river flood forecasting

도시하천 홍수위 예측을 위한 MIMO-LSTM 모형의 리드타임별 예측 성능 평가

Yunji Lima
,
Boosik Kangb
,
Hyung Ki Kimc*
임 윤지a
,
강 부식b
,
김 형기c*
aMaster’s Course, Department of Civil & Environmental Engineering, Dankook University, Yongin, Korea
bProfessor, Department of Civil & Environmental Engineering, Dankook University, Yongin, Korea
cExecutive Director, MagpieSoft Corp., Daejeon, Korea
a단국대학교 토목환경공학과 석사수료
b단국대학교 토목환경공학과 교수
c(주)매그파이소프트 대표이사
*Corresponding Author
31 December 2025. pp. 1629-1643
PDF XML Citation
Abstract

The increasing uncertainty in rainfall patterns and the growing frequency of flash flood events driven by climate change have further emphasized the importance of accurate river stage forecasting. In this study, a Multi-Input Multi-Output Long Short-Term Memory (MIMO-LSTM) model was applied to predict river stage at 10-minute intervals for up to 36 steps (360 minutes) at the Daegokgyo and Gungnaegyo stations within the Tancheon River basin. The input dataset included past rainfall and water level observations, as well as time-shifted rainfall data assumed as forecast rainfall. This configuration enabled an assessment of continuous prediction performance across both short-term (10 minutes) and long-term (360 minutes) lead times. Model performance was evaluated using statistical metrics (NSE, RMSE, MAE) and hydrological metrics (QER, PBIAS), with particular emphasis on the reproducibility of peak stages. The results demonstrated high predictive accuracy in the short-term range (1–6 steps), while a progressive degradation in performance was observed as the forecast lead time increased. Notably, stable reproduction of peak stages was achievable only within approximately 1–2 hours of lead time. This study empirically identified the performance limitations of the multi-input multi-output forecasting structure according to lead time, thereby clarifying both the applicability and constraints of AI-based flood forecasting models. Future work should incorporate actual meteorological forecast rainfall and expand the analysis to diverse catchments and flood events to evaluate uncertainty under realistic conditions.

Keywords
Multi-Input Multi-Output (MIMO) LSTM
Machine learning
Flood forecast model
Lead-time
Peak flood stage

기후변화로 인한 강우 패턴의 불확실성과 돌발홍수 발생 빈도 증가는 하천 홍수위 예측의 중요성을 더욱 부각시키고 있다. 본 연구에서는 탄천 유역 내 대곡교와 궁내교 지점을 대상으로 Multi-Input Multi-Output LSTM 모델을 적용하여 10분 간격으로 최대 36 step (360분)까지 수위를 예측하였다. 입력 자료에는 과거 수위·강우와 함께 시차 이동된 강우를 예보 강우로 가정하여 포함하였으며, 이를 통해 단기(10분)에서 장기(360분)까지 연속적 예측선행시간별 성능 변화를 분석하였다. 모델의 예측 성능은 통계적 지표(NSE, RMSE, MAE)와 수문학적 지표(QER, PBIAS)를 이용해 평가하였으며, 특히 첨두 구간의 재현성을 중점적으로 검토하였다. 연구 결과, 단기(1–6 step)에서는 두 지점 모두 우수한 예측 성능을 보였으나, 예측선행시간이 증가할수록 성능 저하가 뚜렷하게 나타났다. 특히 첨두 구간 예측은 약 1~2시간 이내에서만 안정적인 재현이 가능하였다. 본 연구는 다중입출력 예측 구조에서 예측선행시간별 성능 한계를 실증적으로 규명함으로써, 인공지능 기반 홍수 예측 모델의 적용 가능성과 제약점을 명확히 제시한다. 향후 연구에서는 실제 기상 예보 강우를 반영한 불확실성 평가와 다양한 유역·홍수사상에 대한 확장이 필요하다.

키워드
다중입출력 LSTM
기계학습
홍수위 예측모델
예측선행시간
첨두홍수위
References
1

Abolghasemi, M., Hyndman, R.J., Spiliotis, E., and Bergmeir, C. (2022). “Model selection in reconciling hierarchical time series.” Machine Learning, Vol. 111, pp. 739-789.

10.1007/s10994-021-06126-z
2

Atashi, V., Gorji, H.T., Shahabi, S.M., Kardan, R., and Lim, Y.H. (2022). “Water level forecasting using deep learning time-series analysis: A case study of Red River of the North.” Water, Vol. 14, 1971.

10.3390/w14121971
3

Ben Taieb, S., Bontempi, G., Atiya, A.F., and Sorjamaa, A. (2012). “A review and comparison of strategies for multi-step ahead time series forecasting.” Neurocomputing, Vol. 70, No. 16-18, pp. 2322-2335.

4

Bengio, Y., Simard, P., and Frasconi, P. (1994). “Learning long-term dependencies with gradient descent is difficult.” IEEE Transactions on Neural Networks, IEEE, Vol. 5, No. 2, pp. 157-166.

5

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

10.1162/neco.1997.9.8.1735
6

Hu, C., Wu, Q., Li, H., Jian, S., Li, N., and Lou, Z. (2018). “Deep learning with a long short-term memory networks approach for rainfall-runoff simulation.” Water, MDPI AG, Vol. 10, 1543.

10.3390/w10111543
7

Hunt, K.M., Kratzert, F., Klotz, D., Brenner, C., Schulz, K., and Herrnegger, M. (2022). “Hybrid LSTM models for streamflow forecasting across multiple US basins.” Hydrology and Earth System Sciences, Vol. 26, pp. 5449-5468.

8

Jeong, S., Cho, H., Kim, J., and Lee, G. (2018). “Tidal river water level prediction using an LSTM deep learning model.” Journal of Korea Water Resources Association, Vol. 51, No. 12, pp. 1207-1216.

9

Kim, D., Lee, G., Hwangbo, J., Kim, H., and Kim, S. (2022). “Development of flood stage prediction and flood warning techniques using AI-based models.” Journal of the Korean Society of Hazard Mitigation, Vol. 22, No. 4, pp. 145-156.

10

Kratzert, F., Klotz, D., Brenner, C., Schulz, K., and Herrnegger, M. (2018). “Rainfall-runoff modelling using long short-term memory (LSTM) networks.” Hydrology and Earth System Sciences, Vol. 22, No. 11, pp. 6005-6022.

10.5194/hess-22-6005-2018
11

Lee, J. (2021). Development and application of an AI-based model for real-time flood forecasting. Ph.D. Dissertation, Inha University.

12

Li, W., Kiaghadi, A., and Dawson, C. (2020). “Exploring the best sequence LSTM modeling architecture for flood prediction.” Neural Computing and Applications, Vol. 33, No. 11, pp. 5571-5580.

10.1007/s00521-020-05334-3
13

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
14

Tashman, L.J. (2000). “Out-of-sample tests of forecasting accuracy: An analysis and review.” International Journal of Forecasting, Vol. 16, No. 4, pp. 437-450.

10.1016/S0169-2070(00)00065-0
15

Tawalbeh, R., Alasali, F., Ghanem, Z., Alghazzawi, M., Abu-Raideh, A., and Holderbaum, W. (2023). “Innovative characterization and comparative analysis of water level sensors for enhanced early detection and warning of floods.” Journal of Low Power Electronics and Applications, Vol. 13, 26.

10.3390/jlpea13020026
16

Wu, Z., Huang, Y., Huang, K., Yan, K., and Chen, H. (2023). “A review of non-contact water level measurement based on computer vision and radar technology.” Water, Vol. 15, 3233.

10.3390/w15183233
Information
  • Publisher :KOREA WATER RESOURECES ASSOCIATION
  • Publisher(Ko) :한국수자원학회
  • Journal Title :Journal of Korea Water Resources Association
  • Journal Title(Ko) :한국수자원학회 논문집
  • Volume : 58
  • No :12
  • Pages :1629-1643
  • Received Date : 2025-10-24
  • Revised Date : 2025-11-24
  • Accepted Date : 2025-11-27
  • DOI :https://doi.org/10.3741/JKWRA.2025.58.S-4.1629
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