A paradigm shift in drought forecasting: first application of Bayesian Concept Drift-Enhanced Hidden Markov Models for probabilistic climate risk mapping in Northwestern Algeria
Italian Journal of Agrometeorology · 2026
DOI: 10.36253/ijam-3944 ↗Kurz gefasst
Fachartikel in der Zeitschrift Italian Journal of Agrometeorology, 2026. Verfasst von Malik Kaci und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Northwestern Algeria, a vital agricultural region within the Mediterranean climate hotspot, faces escalating risks due to climatic variability and recurrent droughts. Conventional linear methods are inadequate for capturing the complex, non-stationary dynamics of this semi-arid system. This study introduces an innovative framework integrating a Non-Stationary Hidden Markov Model (NSHMM), Bayesian Concept Drift detection, and CUSUM changepoint analysis. Analyzing 35 years (1990-2024) of daily meteorological data, we identified four distinct climate regimes. Results reveal profound non-stationarity: the frequency of the extreme ‘Hot & Very Dry’ regime increased by 45% since the 1990s, while its persistence rose by 14%. The Bayesian analysis quantifies three major structural drifts with scores reaching 0.80, indicating abrupt climate reorganization. These findings are corroborated by 113 changepoints and a significant teleconnection with the North Atlantic Oscillation (r = 0.58, p < 0.01). Critically, these shifts remain undetected by conventional trend analysis (Mann-Kendall p = 0.16), underscoring the necessity of non-stationary methodologies. The operationalization into a Probabilistic Drought Early Warning System enables risk forecasts with probabilities up to 90%. This work demonstrates that climate change in the region manifests not as gradual warming but as fundamental reorganization of the climate system.
Zitieren
APA
Kaci, M., Martínez-Ruiz, E., & O’Connel, J. T. (2026). A paradigm shift in drought forecasting: first application of Bayesian Concept Drift-Enhanced Hidden Markov Models for probabilistic climate risk mapping in Northwestern Algeria. Italian Journal of Agrometeorology. https://doi.org/10.36253/ijam-3944
Vancouver
Kaci M, Martínez-Ruiz E, O’Connel JT. A paradigm shift in drought forecasting: first application of Bayesian Concept Drift-Enhanced Hidden Markov Models for probabilistic climate risk mapping in Northwestern Algeria. Italian Journal of Agrometeorology. 2026. doi:10.36253/ijam-3944
BibTeX
@article{Kaci2026,
author = {Kaci, Malik and Martínez-Ruiz, Elena and O’Connel, James T.},
title = {A paradigm shift in drought forecasting: first application of Bayesian Concept Drift-Enhanced Hidden Markov Models for probabilistic climate risk mapping in Northwestern Algeria},
journal = {Italian Journal of Agrometeorology},
year = {2026},
doi = {10.36253/ijam-3944}
}
EUVIMED