Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa
Reliable, daily estimates of reservoir storage are pivotal for water allocation and drought response decisions in semiarid regions. Conventional rating curves at Loskop Dam, the primary storage on South Africa’s Olifants River, have become increasingly uncertain owing to sedimentation and episodic d...
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Lenguaje: | Inglés |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | https://hdl.handle.net/10568/179498 |
| _version_ | 1855517721241845760 |
|---|---|
| author | Retief, H. Vigneswaran, Kayathri Ghosh, Surajit Garcia Andarcia, Mariangel Dickens, Chris |
| author_browse | Dickens, Chris Garcia Andarcia, Mariangel Ghosh, Surajit Retief, H. Vigneswaran, Kayathri |
| author_facet | Retief, H. Vigneswaran, Kayathri Ghosh, Surajit Garcia Andarcia, Mariangel Dickens, Chris |
| author_sort | Retief, H. |
| collection | Repository of Agricultural Research Outputs (CGSpace) |
| description | Reliable, daily estimates of reservoir storage are pivotal for water allocation and drought response decisions in semiarid regions. Conventional rating curves at Loskop Dam, the primary storage on South Africa’s Olifants River, have become increasingly uncertain owing to sedimentation and episodic drawdown. A 40 year Digital Earth Africa (DEA) surface area archive (1984–2024) fused with gauged water levels to develop data driven volume predictors that operate under a maximum 9.14% 90 day drawdown constraint. Four nested feature sets were examined: (i) raw water area, (ii) + a power law “calculated volume” proxy, (iii) + six river geometry metrics, and (iv) + fullsupply elevation. Five candidate algorithms, Gradient Boosting (GB), Random Forest (RF), Ridge (RI), Lasso (LA) and Elastic Net (EN), were tuned using a 20 draw random search and assessed with a five fold Timeseries Split to eliminate look ahead bias. Prediction errors were decomposed into Low (< 250 × 106 m³) and High (≥ 250 × 106 m³) storage regimes. Ridge regression achieved the lowest cross validated RMSE (12.3 × 106 m³), outperforming GB by 16% and RF by 7%. In regime terms, Ridge was superior in the Low band (18.0 versus 22.7 MCM for GB) and tied RF in the High band (≈ 12 MCM). In sample diagnostics showed GB’s apparent dominance (6.8–5.4 MCM) to be an artefact of overfitting. A Ridge meta stacked ensemble combining GB, RF, and Ridge reduced full series RMSE to ≈ 11 MCM (≈ 3% of live capacity). We recommend (i) GB retrained daily for routine operations, (ii) Ridge for drought early warning, and (iii) the stacked blend for all weather dashboards. Quarterly rolling retraining and regime specific metrics are advised to maintain operational accuracy below the 5% threshold mandated by the Department of Water and Sanitation. The open source workflow provides a transferable blueprint for satellite enabled reservoir monitoring across data scarce African basins. |
| format | Preprint |
| id | CGSpace179498 |
| institution | CGIAR Consortium |
| language | Inglés |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| record_format | dspace |
| spelling | CGSpace1794982026-01-08T09:06:10Z Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa Retief, H. Vigneswaran, Kayathri Ghosh, Surajit Garcia Andarcia, Mariangel Dickens, Chris water reservoirs water storage machine learning models satellite observation dams Reliable, daily estimates of reservoir storage are pivotal for water allocation and drought response decisions in semiarid regions. Conventional rating curves at Loskop Dam, the primary storage on South Africa’s Olifants River, have become increasingly uncertain owing to sedimentation and episodic drawdown. A 40 year Digital Earth Africa (DEA) surface area archive (1984–2024) fused with gauged water levels to develop data driven volume predictors that operate under a maximum 9.14% 90 day drawdown constraint. Four nested feature sets were examined: (i) raw water area, (ii) + a power law “calculated volume” proxy, (iii) + six river geometry metrics, and (iv) + fullsupply elevation. Five candidate algorithms, Gradient Boosting (GB), Random Forest (RF), Ridge (RI), Lasso (LA) and Elastic Net (EN), were tuned using a 20 draw random search and assessed with a five fold Timeseries Split to eliminate look ahead bias. Prediction errors were decomposed into Low (< 250 × 106 m³) and High (≥ 250 × 106 m³) storage regimes. Ridge regression achieved the lowest cross validated RMSE (12.3 × 106 m³), outperforming GB by 16% and RF by 7%. In regime terms, Ridge was superior in the Low band (18.0 versus 22.7 MCM for GB) and tied RF in the High band (≈ 12 MCM). In sample diagnostics showed GB’s apparent dominance (6.8–5.4 MCM) to be an artefact of overfitting. A Ridge meta stacked ensemble combining GB, RF, and Ridge reduced full series RMSE to ≈ 11 MCM (≈ 3% of live capacity). We recommend (i) GB retrained daily for routine operations, (ii) Ridge for drought early warning, and (iii) the stacked blend for all weather dashboards. Quarterly rolling retraining and regime specific metrics are advised to maintain operational accuracy below the 5% threshold mandated by the Department of Water and Sanitation. The open source workflow provides a transferable blueprint for satellite enabled reservoir monitoring across data scarce African basins. 2025-02-27 2026-01-08T08:00:10Z 2026-01-08T08:00:10Z Preprint https://hdl.handle.net/10568/179498 en Open Access Retief, H.; Vigneswaran, K.; Ghosh, S.; Garcia Andarcia, M.; Dickens, C. 2025. Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa. arXiv, 20p. doi: https://doi.org/10.48550/arXiv.2502.19989 |
| spellingShingle | water reservoirs water storage machine learning models satellite observation dams Retief, H. Vigneswaran, Kayathri Ghosh, Surajit Garcia Andarcia, Mariangel Dickens, Chris Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title | Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title_full | Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title_fullStr | Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title_full_unstemmed | Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title_short | Satellite-surface-area machine-learning models for reservoir storage estimation: regime-sensitive evaluation and operational deployment at Loskop Dam, South Africa |
| title_sort | satellite surface area machine learning models for reservoir storage estimation regime sensitive evaluation and operational deployment at loskop dam south africa |
| topic | water reservoirs water storage machine learning models satellite observation dams |
| url | https://hdl.handle.net/10568/179498 |
| work_keys_str_mv | AT retiefh satellitesurfaceareamachinelearningmodelsforreservoirstorageestimationregimesensitiveevaluationandoperationaldeploymentatloskopdamsouthafrica AT vigneswarankayathri satellitesurfaceareamachinelearningmodelsforreservoirstorageestimationregimesensitiveevaluationandoperationaldeploymentatloskopdamsouthafrica AT ghoshsurajit satellitesurfaceareamachinelearningmodelsforreservoirstorageestimationregimesensitiveevaluationandoperationaldeploymentatloskopdamsouthafrica AT garciaandarciamariangel satellitesurfaceareamachinelearningmodelsforreservoirstorageestimationregimesensitiveevaluationandoperationaldeploymentatloskopdamsouthafrica AT dickenschris satellitesurfaceareamachinelearningmodelsforreservoirstorageestimationregimesensitiveevaluationandoperationaldeploymentatloskopdamsouthafrica |