All Issue

2025 Vol.12, Issue 4 Preview Page

Original Article

Convolutional Neural Network-Based Weather Forecasting Post-Processing Method Using Short-Term Neighborhood Forecast Data

단기예보 데이터를 활용한 Convolutional Neural Network 기반 다지점 단기강수예측 정보 보정 기법 개발

Sangbeom Jang1
,
Ju-Young Shin2*
,
Sunghun Kim3
,
Yongseok Kim4
장 상범1
,
신 주영2*
,
김 성훈3
,
김 용석4
1Master’s 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
3Postdoctoral Researcher, Department of Civil and Environmental Engineering, Kookmin University, Seoul 02707, Korea
4Research Scientist, National Institute Agricultural Sciences, Jeonbuk Special Self-Governing Province, Jeonbuk 55365, Korea
1국민대학교 건설시스템공학과 석사과정
2국민대학교 건설시스템공학과 조교수
3국민대학교 건설시스템공학과 박사후 연구원
4국립농업과학원 연구사
*Corresponding Author
31 December 2025. pp. 252-265
PDF XML Citation
Abstract

This study developed a short-term precipitation forecasting correction technique using a Convolutional Neural Network (CNN) based on the Korea Meteorological Administration’s Local Forecast (Dongne Yeobo) data. The Local Forecast system has limitations, such as systemic bias in certain regions and errors induced by complex terrain. It is often effective to approach the correction of these errors by leveraging artificial intelligence-based models. Accordingly, this research utilized the spatial pattern learning capability of CNNs to learn the spatiotemporal errors of the Local Forecast and combined it with synoptic weather observation data to improve correction accuracy. The model was trained, validated, and evaluated using precipitation data from 78 stations nationwide, covering the period from July 2021 to December 2024, with independent CNN models constructed for each prediction lead time (T=1 to 24). As a result, the proposed model demonstrated performance improvement compared to the original Local Forecast, achieving an average reduction of 16.5% in Root Mean Squared Error and an average increase of 53.9% in the Correlation Coefficient. Notably, the prediction accuracy was significantly improved in the short lead-time range of T < 6. Furthermore, even during extreme precipitation events (exceeding 50 mm in 1 hour and 90 mm in 3 hours), the CNN-based multi-point short-term precipitation correction model showed results closer to the observed values than the original forecast. A comparison with the traditional statistical correction method, Quantile Mapping, also showed that the CNN-based multi-point short-term precipitation correction model exhibited consistent superiority across all prediction lead-time intervals, confirming the effectiveness of CNN-based correction in comprehensively learning spatial information. The results of this study are expected to enhance the reliability of short-term forecasts, which can be utilized for disaster response and water resource management.

Keywords
Convolutional neural network
Deep learning
Dongne yeobo forecast
Natural disaster
Post-processing
Precipitation correction

본 연구는 기상청의 동네예보 자료를 대상으로 합성곱 신경망(Convolutional Neural Network, CNN)을 활용하여 단기 강수 예측 정보 보정 기법을 개발하였다. 동네예보는 복잡한 지형과 일부 지역에서 구조적 편향(Systemic bias)이 발생하는 한계가 있어, 이를 보정하기 위해 인공지능을 활용하는 방식으로 접근하는 것이 효과적일 수 있다. 이에 본 연구에서는 CNN의 공간 패턴 학습 능력을 이용하여 동네예보의 시공간적 오차를 학습하고, 종관기상관측 자료를 결합하여 보정 정확도를 향상시켰다. 모델은 2021년 7월부터 2024년 12월까지의 전국 78개 지점 강수 자료를 이용하여 예측 리드타임별(T=1~24) 독립 모델을 구성하여 학습 및 검증, 평가하였다. 그 결과, 제안된 모델은 원 동네예보 대비 평균 제곱근 오차를 평균 16.5% 감소시키고, 상관계수를 평균53.9% 향상시켰으며, 특히 리드타임 T<6 구간에서의 예측 정확도를 크게 개선시켰다. 또한 극한 강수(1시간 50 mm 이상, 3시간 90 mm 이상) 상황에서도 본 연구의 CNN 기반 다지점 단기강수예측 정보 보정 모델이 원 예보보다 관측값에 더 근접한 결과를 보였다. 전통적 통계 보정기법인 Quantile Mapping과의 비교에서도 CNN 기반 다지점 단기강수예측 정보 보정 모델이 전 예측시간 구간에서 일관된 우수성을 나타내어, 공간 정보를 통합적으로 학습하는 CNN기반 보정의 효용성을 확인하였다. 본 연구의 결과는 단기 예보의 신뢰성을 향상시켜 재해 대응 및 수자원 관리에 기여할 수 있을 것으로 기대된다.

키워드
CNN
딥러닝
동네예보
자연재해
후처리(Post-Processing)
강수 보정
References
1

Al-Samrraie, L.A.A., Abdalla, A.M., Alrawashdeh, K.A.B., Al Bsoul, A., Awad, M.A., Alzboon, K., and Al-Taani, A.A. 2025. Deep Learning Models Based on CNN, RNN, and LSTM for Rainfall Forecasting: Jordan as a Case Study. Mathematical Modelling of Engineering Problems 12(7): 2456-2466.

10.18280/mmep.120724
2

Badrinath, A., Delle Monache, L., Hayatbini, N., Chapman, W., Cannon, F., and Ralph, M. 2023. Improving precipitation forecasts with convolutional neural networks. Weather and Forecasting 38(2): 291-306.

10.1175/WAF-D-22-0002.1
3

Buda, M., Maki, A., and Mazurowski, M.A. 2018. A systematic study of the class imbalance problem in convolutional neural networks. Neural Networks 106: 249-259.

10.1016/j.neunet.2018.07.011
4

Cantalejo, M., Cobos, M., Millares, A., and Baquerizo, A. 2024. A non-stationary bias adjustment method for improving the inter-annual variability and persistence of projected precipitation. Scientific Reports 14: 76848.

10.1038/s41598-024-76848-2
5

Ebtehaj, I. and Bonakdari, H. 2024. CNN vs. LSTM: A comparative study of hourly precipitation intensity prediction as a key factor in flood forecasting frameworks. Atmosphere 15(9): 1082.

10.3390/atmos15091082
6

Glahn, H.R. and Lowry, D.A. 1972. The use of Model Output Statistics (MOS) in objective weather forecasting. Journal of Applied Meteorology 11(8): 1203-1211.

10.1175/1520-0450(1972)011<1203:TUOMOS>2.0.CO;2
7

Gneiting, T., Raftery, A.E., Westveld III, A.H., and Goldman, T. 2005. Calibrated probabilistic forecasting using ensemble model output statistics and minimum CRPS estimation. Monthly Weather Review 133(5): 1098-1118.

10.1175/MWR2904.1
8

Hess, P. 2023. Deep learning for improving numerical weather prediction from radar and satellite observations. arXiv preprint arXiv:2309.12345.

9

IPCC. 2022. Climate Change 2022: Impacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press.

10

Jiao, Y., Wang, X., Wang, W., and Li, S. 2022. Image Semantic Segmentation Fusion of Edge Detection and AFF Attention Mechanism. Applied Sciences 12(21): 11248.

10.3390/app122111248
11

Li, W., Pan, B., Xia, J., and Duan, Q. 2022. Convolutional neural network-based statistical post-processing of ensemble precipitation forecasts. Journal of Hydrology 605: 127301.

10.1016/j.jhydrol.2021.127301
12

Li, X., Wu, H., Nanding, N., Chen, S., Hu, Y., and Li, L. 2023. Statistical bias correction of precipitation forecasts based on quantile mapping on the sub-seasonal to seasonal scale. Remote Sensing 15(7): 1743.

10.3390/rs15071743
13

Rosca, M., Weber, T., Mohamed, S., and Turner, R.E. 2023. A prior-based approximate latent Riemannian metric. arXiv preprint arXiv:2302.08235.

14

Sadeghi, M., Nguyen, P., Hsu, K., and Sorooshian, S. 2020. Improving Near Real-Time Precipitation Estimation Using a U-Net Convolutional Neural Network and Geographical Information. Environmental Modelling & Software 134: 104856.

10.1016/j.envsoft.2020.104856
15

Scheuerer, M. and Hamill, T.M. 2015. Statistical postprocessing of ensemble precipitation forecasts by fitting censored, shifted gamma distributions. Monthly Weather Review 143(11): 4578-4596.

10.1175/MWR-D-15-0061.1
16

Steininger, M., Abel, D., Ziegler, K., Krause, A., Paeth, H., and Hotho, A. 2023. ConvMOS: climate model output statistics with deep learning. Data Mining and Knowledge Discovery 37: 136-166

10.1007/s10618-022-00877-6
17

Tefera, G.W. et al. 2023. Evaluating the Impact of Statistical Bias Correction on Climate Change Signal and Extreme Indices in the Jemma Sub-Basin of the Blue Nile Basin. Sustainability 15(13): 10513.

10.3390/su151310513
18

Vaughan, A., Tebbutt, W., Hosking, J.S., and Turner, R.E. 2022. Convolutional conditional neural processes for local climate downscaling. Geoscientific Model Development 15: 251-268.

10.5194/gmd-15-251-2022
19

Wang, J. and Tian, J. 2021. Seasonally-Integrated Autoencoder with Long Short-Term Memory for Precipitation Forecasting. Water 15(11): 2036.

20

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 30: 79-82.

10.3354/cr030079
21

Zhang, T., Liang, Z., Li, W., Wang, J., Hu, Y., Li, B. 2023. Statistical post-processing of precipitation forecasts using circulation classifications and spatiotemporal deep neural networks. Hydrology and Earth System Sciences 27(10): 1945-1962.

10.5194/hess-27-1945-2023
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 :4
  • Pages :252-265
  • Received Date : 2025-10-22
  • Revised Date : 2025-11-07
  • Accepted Date : 2025-11-11
  • DOI :https://doi.org/10.17820/eri.2025.12.4.252
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