Predicting water level fluctuations in glacier-fed lakes by ensembling individual models into a quad-meta model
Predicting water levels in glacier-fed lakes is vital for water resource management, flood forecasting, and ecological balance. This study examines the predictive capacity of multiple climate factors affecting Blue Moon Lake Valley, fed by the Baishui River glacier on Yulong Snow Mountain. The study...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Engineering Applications of Computational Fluid Mechanics |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/19942060.2024.2449124 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Predicting water levels in glacier-fed lakes is vital for water resource management, flood forecasting, and ecological balance. This study examines the predictive capacity of multiple climate factors affecting Blue Moon Lake Valley, fed by the Baishui River glacier on Yulong Snow Mountain. The study introduces a novel quad-meta (QM) ensemble model that integrates outputs from four machine learning models – extreme gradient boosting (XGB), random forest (RF), gradient boosting machine (GBM), and decision tree (DT) – through meta-learning to improve the prediction accuracy for water levels under complex environmental conditions. High-frequency water depth data, recorded every five minutes using an RBR logger, alongside climate variables such as temperature, wind speed, humidity, evaporation, solar radiation, and rainfall, were analyzed. Temperature was identified as the most significant factor influencing water levels, with an importance score of 15.69, followed by atmospheric pressure (14.08) and solar radiation (12.89), which impacted surface conditions and evaporation. Relative humidity (10.24) and wind speed (8.71) influenced lake stability and mixing. The QM model outperformed individual models, and achieved RMSE values of 0.003 m (climate data) and 0.001 m (water depth data), with R2 values of 0.994 and 0.999, respectively. In comparison, XGB and GBM exhibited higher RMSE and lower R2 scores. RF struggled with an RMSE of 0.008 and R2 of 0.962, while DT performed better (RMSE: 0.006 m for climate data) but remained inferior to the proposed model. These findings demonstrate the robustness of the QM ensemble approach in handling complex environmental data, particularly where individual models fall short. This study highlights the potential of ensemble learning for enhanced prediction accuracy in glacier-fed lake systems, recommending future research directions that incorporate the QM model with deep learning and long-term forecasting to expand predictive capabilities on a global scale. |
|---|---|
| ISSN: | 1994-2060 1997-003X |