Seasonal Spatio-Temporal Machine Learning with Conformal Uncertainty for Forecasting Nitrate, Ammonia, and Orthophosphate in Surface Waters
International Journal of Innovations in Science and Technology · 22.01.2026 · 179
DOI: 10.33411/ijist/1695 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovations in Science and Technology, online erschienen am 22.01.2026. Verfasst von Rabia Tehseen und 6 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Proper prediction of nutrient contents in organizations is critical to safeguarding water bodies as well as sustainable environmental management. This article introduces a seasonal spatio-temporal prediction framework of nutrients based on machine learning and estimation of uncertainty. Based on the large-scale monitoring data in 2010-2025, monthly concentrations of nitrate, ammonia, and orthophosphate were modeled at thousands of sampling sites. Fast features: Feature engineering. The temporal dynamics were modeled using seasonal encoding, lag values, and rolling statistics. Three prediction methods were considered: the Random Forest regression, the XGBoost regression, the feed-forward neural networks, and the hybrid model that concluded with the combination of the Random Forest and the neural network learning. Conformal prediction was used to measure prediction uncertainty to produce statistically sound 90 percent prediction intervals. Mean absolute error, root mean square error, coefficient of determination, and interval coverage were used to determine model performance. It was found that the overall performance of Random Forest with conformal prediction (MAE = 0.1008, RMSE = 0.2294, R2 = 0.7746) is the best, followed by the hybrid residual model, presenting the reported results are averaged across nutrients. XGBoost and neural network models also showed a high predictive power, but the accuracy of the models and the uncertainty confidence limits were lower. Significantly lower performance was exhibited by baseline persistence and climatology methods. ANOVA was required to test whether the performance difference of the models was due to statistically significant variation. The solution offered by the suggested framework is accurate, interpretable, and error-aware in the long-term nutrient forecasting. The methodology provides credible data concerning environmental risks and sound decisions on water quality management through the combination of ensemble learning and conformal uncertainty estimation.
Zitieren
APA
Tehseen, R., Mustaqeem, A., Haq, M. I. U., Awan, M. M., Omer, U., Inayat, W., & Rafaqat, S. (2026). Seasonal Spatio-Temporal Machine Learning with Conformal Uncertainty for Forecasting Nitrate, Ammonia, and Orthophosphate in Surface Waters. International Journal of Innovations in Science and Technology, 179. https://doi.org/10.33411/ijist/1695
Vancouver
Tehseen R, Mustaqeem A, Haq MIU, Awan MM, Omer U, Inayat W, et al. Seasonal Spatio-Temporal Machine Learning with Conformal Uncertainty for Forecasting Nitrate, Ammonia, and Orthophosphate in Surface Waters. International Journal of Innovations in Science and Technology. 2026:179. doi:10.33411/ijist/1695
BibTeX
@article{Tehseen2026,
author = {Tehseen, Rabia and Mustaqeem, Anam and Haq, Muhammad Inam Ul and Awan, Maham Mehr and Omer, Uzma and Inayat, Wasif and Rafaqat, Summaya},
title = {Seasonal Spatio-Temporal Machine Learning with Conformal Uncertainty for Forecasting Nitrate, Ammonia, and Orthophosphate in Surface Waters},
journal = {International Journal of Innovations in Science and Technology},
year = {2026},
pages = {179},
doi = {10.33411/ijist/1695}
}
EUVIMED