All Issue

2025 Vol.12, Issue 2 Preview Page

Original Article

Drought Risk Prediction Method Using Machin Learning Based on Emergency Water Supply Data

비상급수 자료 기반의 기계학습 기법을 활용한 가뭄 위험도 예측 기법 개발

Junyeong Gil1
,
Kyeungmin Moon1
,
Seohyeon Kim1
,
Ju-Young Shin2*
길 준영1
,
문 경민1
,
김 서현1
,
신 주영2*
1Undergraduate Student, Department of Civil and Environmental Engineering, Kookmin University, Seoul 02707, Korea
2Assistant Professor, Department of Civil and Environmental Engineering, Kookmin University, Seoul 02707, Korea
1국민대학교 건설시스템공학과 학사과정
2국민대학교 건설시스템공학과 교수
*Corresponding Author
30 June 2025. pp. 57-68
PDF XML Citation
Abstract

This study developed a methodology for predicting drought risk using machine learning techniques. While previous drought prediction studies have used various indices, few have incorporated actual impact or damage resulting from drought events. Hence, this research attempted to use emergency water supply data to quantitatively evaluate and predict drought risk. Data were collected for the Jeolla region (Yeosu, Wando, Imsil, Jeongeup, Jangsu, Jangheung, and Goheung), including meteorological data (wind speed, temperature, relative humidity, and sunshine hours), hydrological data (dam water levels and discharge), the Standardized Precipitation Index (SPI), and emergency water supply data. RandomForest, XGBoost, and LightGBM were applied as drought prediction models, and their performance was evaluated in three scenarios. In Case 1, where data from all regions were combined for training and evaluation, all three models showed high accuracy. In Case 2, the data were divided by region for prediction, and most regions showed low performance except Wando. In Case 3, a model optimized for Wando was applied, resulting in high predictive accuracy. Wando island’s superior performance is presumed to be due to the availability of sufficient data from that region. Unlike previous research, this study confirms that using actual damage data allows for more practical and effective drought risk prediction.

Keywords
Drought
Emergency water supply
LightGBM
RandomForest
XGBoost

본 연구는 인공지능 기법을 활용하여 가뭄 위험도를 예측하는 방법론을 개발하였다. 기존의 가뭄 예측 연구들은 다양한 지수를 활용했지만, 실제 피해 사례를 반영한 연구는 부족하여 본 연구는 비상급수 자료를 활용하여 가뭄 피해의 위험도를 정량적으로 평가하고 예측하는 기법을 개발하였다. 전라도 지역(여수, 완도, 임실, 정읍, 장수, 장흥, 고흥)을 대상으로 기상 데이터(풍속, 기온, 습도, 일조량), 수문 데이터(댐 수위 및 방류량), 표준강수지수, 비상급수 발생 데이터를 수집하였다. 가뭄 예측 모델로는 RandomForest, XGBoost, LightGBM을 적용하였으며, 3가지 사례(Case 1, Case 2, Case 3)로 성능 평가를 진행하였다. Case 1에서는 모든 지역의 데이터를 통합하여 학습 및 평가를 수행한 결과, 세 모델 모두 높은 정확도를 보였다. Case 2에서는 지역별로 데이터를 나누어 예측을 수행한 결과, 완도 지역을 제외한 대부분의 지역에서 성능이 저조하였다. Case 3에서는 완도 지역을 대상으로 최적화한 모델을 적용한 결과, 높은 예측 정확도를 확인하였다. 완도 지역에서의 예측 성능이 가장 우수하였으며, 이는 해당 지역의 데이터가 충분히 확보되었기 때문으로 판단된다. 본 연구는 기존 연구들과 달리 실제 피해 데이터를 활용하여 보다 실질적인 가뭄 위험 예측이 가능하다는 것을 확인하였다.

키워드
가뭄
비상급수자료
LightGBM
RandomForest
XGBoost
References
1

Bates, B.C., Kundzewicz, Z.W., Wu, S., and Palutikof, J.P. 2008. Climate Change and Water, Technical Paper of the Intergovernmental Panel on Climate Change, Intergovernmental Panel on Climate Change Secretariat, Geneva. p. 210.

2

Breiman, L. 2001. Random forests. Machine Learning. 45(1): 5-32.

10.1023/A:1010933404324
3

Byun, H.R. and Wilhite, D.A. 1999. Objective quantification of drought severity and duration. Journal of Climate 12(9): 2747-2756.

10.1175/1520-0442(1999)012<2747:OQODSA>2.0.CO;2
4

Chen, T. and Guestrin, C. 2016. XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery, San Francisco, pp. 785-794.

10.1145/2939672.2939785
5

Choi, C.H., Park, K.H., Park, H.K., Lee, M.J., Kim, J.S. and Kim, H.S. 2017. Development of Heavy Rain Damage Prediction Fnction for Public Facility Using Machine Learning. Journal of the Korean Society of Hazard Mitigation 17(6): 443-450. (in Korean)

10.9798/KOSHAM.2017.17.6.443
6

Han, Y.J. and Joe, I.W. 2022. Imbalanced Data Improvement Techniques Based on SMOTE and Light GBM. KIPS Transactions on Computer and Communication Systems 11(12): 445-452. (in Korean)

7

Jang, K.M., Lee, M.B., Lim, J.H., Oh, H.B., Shin, C.S. and Park, J.W. 2023. Comparison of Machine Learning- Based Greenhouse VPD Prediction Models. KIPS Transactions on Software and Data Engineering 12(3): 125-132. (in Korean)

8

Kim, S.M., Park, S.W. and Kim, H.J. 1998. Drought Analyses of 1994 Using Standardized Precipitation Index. Journal of the Korean Society of Agricultural Engineers 1988: 92-98 (in Korean)

9

Kim, Y.E., Lee, K.E. and Kim, G.S. 2020. Forecast of drought index using decision tree based methods. The Korean Data & Information Science Society 31(2): 273-288. (in Korean)

10.7465/jkdi.2020.31.2.273
10

Kim, Y.S., Jung, M.K., Kim, T.W., Jeoung, S.M. and Kwon, H.H. 2022. Assessment of hydrological drought risk in the southern region in 2022: based on bivariate regional drought frequency analysis. Journal of Korea Water Resources Association 56(2): 151-163. (in Korean)

11

Lee, Y.J. and Sun, J.W. 2020. Predicting Highway Concrete Pavement Damage using XGBoost. Korean Journal of Construction Engineering and Management 21(6): 46-55. (in Korean)

12

Park, B.S., Lee, J.H., Kim, C.J. and Jang, H.W. 2013. Projection of Puture Drought of Korea Based on Probabilistic Approach Using Multi-model and Multi Climate Change Scenarios. Journal of Civil and Environmental Engineering Research 33(5): 1871-1885. (in Korean)

10.12652/Ksce.2013.33.5.1871
13

Park, J.Y., Yoo, J.Y., Lee, M.W. and Kim, T.W. 2012. Assessment of Drought Risk in Korea: Focused on Data-based Drought Risk Map. Journal of Civil and Environmental Engineering Research 32(4B): 203-211. (in Korean)

10.12652/Ksce.2012.32.4B.203
14

Seo, J.H., Chi, H.W., Kim, H.J. and Kim, Y.J. 2022. Hydrological drought risk assessment for climate change adaptation in South Korea. Journal of Korea Water Resources Association 55(6): 421-435. (in Korean)

15

Seo, M.S. and Hwang, H.G. 2022. Darknet Traffic Classification based on LightGBM. Journal of Korean Institute of Communication and Information Sciences 2022(6): 1606-1607. (in Korean)

16

Sim, D,C., Lee, J.Y., Jang, J.W. and Lee, M.W. 2022. Prediction of chloride concentration in groundwater on Jeju Island using XGBoost regression machine learning. Journal of the Geological Society of Korea 58(2): 243-255. (in Korean)

10.14770/jgsk.2022.58.2.243
17

Yoon, D.H., Nam, W.H., Lee, H.J., Hong, E.M. and Kim, T.G. 2020. Drought Hazard Assessment using MODIS- based Evaporative Stress Index (ESI) and ROC Analysis. Journal of the Korean Society of Agricultural Engineers 62(3): 51-61. (in Korean)

Information
  • Publisher :Korean Society of Ecology and Infrastructure Engineering
  • Publisher(Ko) :응용생태공학회
  • Journal Title :Ecology and Resilient Infrastructure
  • Journal Title(Ko) :응용생태공학회 논문집
  • Volume : 12
  • No :2
  • Pages :57-68
  • Received Date : 2025-03-25
  • Revised Date : 2025-03-27
  • Accepted Date : 2025-04-02
  • DOI :https://doi.org/10.17820/eri.2025.12.2.057
Journal Informaiton Ecology and Resilient Infrastructure Ecology and Resilient Infrastructure
Online Submission
  • NRF
  • KISTI Current Status
  • KISTI Cited-by
  • crosscheck
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close