Weighted Ensemble of GRU, LSTM and XGBoost for Multi-Horizon Temperature Forecasting at a Tropical Highland Station

Kalimi Kalimi; Ahmad Musyafa; Taswanda Taryo; Marzuki Sinambela; Tonny Wahyu Aji

doi:10.30871/jaic.v10i4.13593

Authors

Kalimi Kalimi Universitas Pamulang
Ahmad Musyafa Universitas Pamulang
Taswanda Taryo Universitas Pamulang
Marzuki Sinambela Sekolah Tinggi Meteorologi, Klimatologi dan Geofisika
Tonny Wahyu Aji Sekolah Tinggi Meteorologi, Klimatologi dan Geofisika

DOI:

https://doi.org/10.30871/jaic.v10i4.13593

Keywords:

Deep Learning, Ensemble Learning, Multi-horizon Forecasting, Air Temperature, Multivariate Variables

Abstract

This research evaluates multi-horizon forecasting of daily mean temperature (TAVG) and maximum temperature (TMAX) at the Malang/Karangploso Climatology Station WMO 96943 using daily meteorological observations from 2014–2023. The predictors include humidity, rainfall, atmospheric pressure, wind variables, weather conditions, and sunshine duration. Forecasts are generated for H+1, H+3, H+7, and H+14 using a chronological train, validation, and test split. The study compares GRU, LSTM, Hybrid Gated LSTM-GRU, XGBoost, and Weighted Ensemble models against Persistence and Climatology baselines. To prevent data leakage, preprocessing includes missing-date handling, placeholder correction, training-set-based imputation, lag and rolling feature construction, and input normalization. Optuna is used to tune recurrent models, while ensemble weights are optimized on the validation set. Model performance is assessed using MAE, RMSE, and R², with the lowest test RMSE as the main selection criterion. Results show that the Weighted Ensemble achieves the best aggregate performance, with mean MAE of 0.796, mean RMSE of 1.007, and mean R² of 0.472. GRU is the strongest individual model, with mean RMSE of 1.022. However, the best model varies by target and horizon. Weighted Ensemble leads in five of eight scenarios for TAVG H+1, TAVG H+3, TAVG H+14, TMAX H+1 and TMAX H+14, Hybrid Gated performs best for TAVG H+7, and GRU is superior for TMAX H+3 and H+7.

Downloads

Download data is not yet available.

References

[1] R. Gernowo, C. E. Widodo, and A. P. Widodo, “The Atmospheric Dynamic Extreme of Tropical as an Analysis of Climate Change in Indonesia,” Geogr. Tech., vol. 15, no. 2, pp. 138–146, 2020, doi: 10.21163/gt_2020.152.14.

[2] Suhadi, I. Iskandar, Supari, M. Irfan, and H. Akhsan, “Extreme Drought Assessment in Sumatra-Indonesia Using SPI and EDI,” Sci. Technol. Indones., vol. 8, no. 4, pp. 691–700, 2023, doi: 10.26554/sti.2023.8.4.691-700.

[3] S. Siswanto, G. v. d. Schrier, and B. v. d. Hurk, “Observed Increase of Urban Extreme Rainfall as Surface Temperature Rise: The Jakarta Case,” J. Meteorol. Soc. Japan, vol. 100, no. 3, pp. 475–492, 2022, doi: 10.2151/jmsj.2022-023.

[4] G. G. Sudarman, T. W. Hadi, N. S. Ningsih, A. Sopaheluwakan, M. R. Syahputra, and F. Marpaung, “Nonstationary Changes in Annual Rainfall over Indonesia’s Maritime Continent,” Adv. Meteorol., vol. 2024, 2024, doi: 10.1155/2024/9392844.

[5] A. Widi Lestari, T. Yuniningsih, E. Larasati, and M. Okto Adhitama, “Sustainable Development Based Tourism in Batu City,” KnE Soc. Sci., vol. 2023, pp. 447–458, 2023, doi: 10.18502/kss.v8i11.13565.

[6] S. Herho and G. Firdaus, “Time-Series Analysis and Statistical Forecasting of Daily Rainfall in Kupang, East Nusa Tenggara, Indonesia: A Pilot Study,” 2022, doi: 10.31223/x5pw5g.

[7] F. Subiyantoro, F. Abdullah, N. Ismail, I. D. Gede Arya Putra, and T. Grefyolin Simbolon, “Projection of Climate Extreme Indices Based on Representative Concentration Pathways (RCP) in Aceh Province,” J. Ilmu Fis. | Univ. Andalas, vol. 17, no. 1, pp. 1–8, 2024, doi: 10.25077/jif.17.1.1-8.2025.

[8] R. Pahlepi, S. Soekirno, and H. S. Wicaksana, “Solar Radiation Intensity Imputation in Pyranometer of Automatic Weather Station Based on Long Short Term Memory,” Ultim. Comput. J. Sist. Komput., vol. 15, no. 2, pp. 35–40, 2023, doi: 10.31937/sk.v15i2.3348.

[9] O. A. Wani et al., “Predicting rainfall using machine learning, deep learning, and time series models across an altitudinal gradient in the North-Western Himalayas,” Sci. Rep., vol. 14, no. 1, 2024, doi: 10.1038/s41598-024-77687-x.

[10] A. M. Assaf, H. Haron, H. N. Abdull Hamed, F. A. Ghaleb, S. N. Qasem, and A. M. Albarrak, “A Review on Neural Network Based Models for Short Term Solar Irradiance Forecasting,” Appl. Sci., vol. 13, no. 14, 2023, doi: 10.3390/app13148332.

[11] A. Bekkar, B. Hssina, S. Douzi, and K. Douzi, “Air-pollution prediction in smart city, deep learning approach,” J. Big Data, vol. 8, no. 1, pp. 1–21, 2021, doi: 10.1186/s40537-021-00548-1.

[12] S. Boubaker, M. Benghanem, A. Mellit, A. Lefza, O. Kahouli, and L. Kolsi, “Deep Neural Networks for Predicting Solar Radiation at Hail Region, Saudi Arabia,” IEEE Access, vol. 9, pp. 36719–36729, 2021, doi: 10.1109/ACCESS.2021.3062205.

[13] M. G. Schultz et al., “Can deep learning beat numerical weather prediction?,” Philos. Trans. R. Soc. A Math. Phys. Eng. Sci., vol. 379, no. 2194, 2021, doi: 10.1098/rsta.2020.0097.

[14] J. Xie, H. Zhang, L. Liu, M. Li, and Y. Su, “Decomposition-Based Multistep Sea Wind Speed Forecasting Using Stacked Gated Recurrent Unit Improved by Residual Connections,” Complexity, vol. 2021, 2021, doi: 10.1155/2021/2727218.

[15] X. Zhao et al., “A Comprehensive Review of Methods for Hydrological Forecasting Based on Deep Learning,” Water, vol. 16, no. 10, p. 1407, 2024, doi: 10.3390/w16101407.

[16] R. Chandrasekaran and S. K. Paramasivan, “A State-of-the-Art Review of Time Series Forecasting Using Deep Learning Approaches,” Int. J. Recent Innov. Trends Comput. Commun., vol. 10, no. 12, pp. 92–105, 2022, doi: 10.17762/ijritcc.v10i12.5890.

[17] J. F. Torres, D. Hadjout, A. Sebaa, F. Martínez‐Álvarez, and A. Troncoso, “Deep Learning for Time Series Forecasting: A Survey,” Big Data, vol. 9, no. 1, pp. 3–21, 2021, doi: 10.1089/big.2020.0159.