Drought is among the most impactful climate-related hazards affecting water resources, agriculture, and food security, particularly in semi-arid regions. Across the Mediterranean and Africa, the increasing impacts of climate change, particularly recurrent droughts, are compounded by limited hydrometeorological observation capacity, which constrains effective drought monitoring, modelling, and forecasting. Addressing these challenges, this dissertation develops and applies a multi-scale framework to improve drought characterization, hydrological simulation, and seasonal streamflow forecasting, with a focus on Morocco and the African continent. At the regional scale, drought monitoring is investigated in the Tensift River Basin (Morocco) through a comprehensive evaluation of satellite-based (CHIRPS v2) and reanalysis datasets (ERA5-Land) against in situ observations. The performance of drought indices, the Standardized Precipitation Index (SPI) and the Standardized Precipitation and Evapotranspiration Index (SPEI), is assessed by evaluating their ability to reproduce documented drought events using receiver operating characteristic (ROC) analysis. Results show that CHIRPS v2 precipitation and ERA5-Land temperature datasets reproduce hydroclimatic variability in the basin with satisfactory agreement. The Pearson Type III distribution is identified as the most suitable for SPI computation, while the log-logistic distribution is confirmed for SPEI. SPI at the 12-month timescale provides the highest skill in detecting reported drought events. This research also contributes to the evaluation of the Climate Hazards Center Infrared Precipitation with Stations, Version 3 (CHIRPS v3), an updated high-resolution precipitation dataset developed by the Climate Hazards Center at the University of California, Santa Barbara. Improvements in climatology, station coverage, satellite estimation, and gauge-undercatch correction enhance precipitation representation in semi-arid and mountainous environments, including Morocco, reinforcing its suitability for drought monitoring in data-limited regions. At the local scale, a seasonal streamflow forecasting framework is developed for the N’fis Basin, a drought-prone tributary of the Tensift River Basin. Statistical and machine-learning models predict streamflow at short lead times relevant to reservoir operations. Forecasts can anticipate inflow deficits 1–3 months before peak irrigation demand, supporting proactive reservoir management and illustrating the potential of forecast-informed reservoir operations for drought preparedness. At the continental scale, the performance of the Famine Early Warning Systems Network Land Data Assimilation System Forecast (FLDAS-Forecast), developed by NASA and FEWS NET to estimate hydroclimatic conditions in data-scarce regions, is evaluated against extensive in situ streamflow observations across Africa. FLDAS-Forecast streamflow outputs reproduce major hydrological signals and seasonal variability, showing useful predictive skill at short to medium lead times despite uncertainties related to land-surface modeling, meteorological forcing, and sparse gauge coverage. Finally, this dissertation demonstrates that integrating satellite observations, reanalysis products, land-surface modeling, and machine-learning approaches can substantially enhance drought monitoring and forecasting across spatial scales. By improving the capacity to detect, characterize, and anticipate drought impacts on water resources in data-scarce regions, the findings support drought early warning and adaptive water-resources planning across Africa, contributing to SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 13 (Climate Action), and SDG 15 (Life on Land).
La siccità è, tra gli eventi climatici estremi, quello che produce gli impatti negativi più rilevanti sulle risorse idriche, sull’agricoltura e, di conseguenza, sulla sicurezza alimentare, in particolare nelle regioni semi-aride. Nel Mediterraneo e in Africa, i crescenti impatti del cambiamento climatico, in particolare le siccità ricorrenti, sono aggravati dalla limitata capacità di osservazione idrometeorologica, che ostacola un efficace monitoraggio, modellazione e previsione della siccità. Per affrontare queste sfide, la presente tesi propone un approccio metodologico multi-scala volto a migliorare la caratterizzazione delle siccità, la modellazione idrologica e la previsione delle portate fluviali, con particolare riferimento al Marocco e al continente africano. Alla scala regionale, l’identificazione e la caratterizzazione delle siccità sono condotte nel bacino del fiume Tensift (Marocco) attraverso la valutazione di dati satellitari (CHIRPS v2) e di rianalisi (ERA5-Land) confrontati con osservazioni in situ. Le prestazioni degli indici di siccità SPI e SPEI sono valutate in termini di capacità di riprodurre eventi siccitosi documentati mediante analisi ROC. I risultati mostrano che CHIRPS v2 ed ERA5-Land riproducono la variabilità idroclimatica osservata nel bacino del Tensift con accordo soddisfacente. La distribuzione Pearson di Tipo III risulta la più idonea per il calcolo dello SPI, mentre la distribuzione log-logistica è confermata per lo SPEI. Inoltre, lo SPI alla scala temporale di 12 mesi presenta la maggiore capacità nel rilevare gli eventi siccitosi documentati. Questa ricerca contribuisce anche alla valutazione del dataset di precipitazione CHIRPS v3, versione aggiornata di CHIRPS v2 sviluppata dal Climate Hazards Center (University of California, Santa Barbara). I risultati relativi al Marocco mostrano che CHIRPS v3 migliora la rappresentazione delle precipitazioni in ambienti semi-aridi e montuosi, rafforzandone l’utilità per il monitoraggio della siccità e per applicazioni idrologiche in regioni con scarsità di dati. Alla scala locale, viene sviluppato un sistema di previsione delle portate per il bacino dell’N’fis, tributario soggetto a frequenti condizioni di siccità del bacino del Tensift. Modelli statistici e di machine learning sono applicati per prevedere le portate a brevi tempi di anticipo rilevanti per la gestione degli invasi. I risultati mostrano che è possibile anticipare eventuali deficit di portata da 1 a 3 mesi prima del picco della domanda irrigua, fornendo informazioni operative per una gestione proattiva delle risorse idriche in condizioni di siccità. Alla scala continentale, le prestazioni del Famine Early Warning Systems Network Land Data Assimilation System Forecast (FLDAS-Forecast), sistema di modellistica idrologica sviluppato da NASA e FEWS NET per stimare condizioni idroclimatiche in regioni con scarsa disponibilità di dati, sono valutate mediante confronto con osservazioni in situ delle portate fluviali in Africa. Le portate simulate da FLDAS-Forecast riproducono i principali segnali idrologici e la variabilità stagionale, mostrando una capacità predittiva utile a brevi e medi tempi di anticipo nonostante le incertezze legate alla modellistica di superficie, alle forzanti meteorologiche e alla limitata copertura delle stazioni idrometriche. In conclusione, la presente tesi dimostra come l’uso integrato di osservazioni satellitari, prodotti di rianalisi, modellistica di superficie terrestre e approcci statistici e di machine learning possa migliorare significativamente il monitoraggio e la previsione della siccità a diverse scale spaziali in un contesto di cambiamento climatico. I risultati supportano sistemi di allerta precoce più efficaci e una pianificazione adattiva delle risorse idriche in Africa, contribuendo agli SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 13 (Climate Action) e SDG 15 (Life on Land).
Verso un approccio integrato del monitoraggio e della modellazione degli impatti di siccità sulle risorse idriche in Africa in uno scenario di cambiamento climatico / Naim, Mohamed. - (2026 May 14).
Verso un approccio integrato del monitoraggio e della modellazione degli impatti di siccità sulle risorse idriche in Africa in uno scenario di cambiamento climatico
NAIM, MOHAMED
2026-05-14
Abstract
Drought is among the most impactful climate-related hazards affecting water resources, agriculture, and food security, particularly in semi-arid regions. Across the Mediterranean and Africa, the increasing impacts of climate change, particularly recurrent droughts, are compounded by limited hydrometeorological observation capacity, which constrains effective drought monitoring, modelling, and forecasting. Addressing these challenges, this dissertation develops and applies a multi-scale framework to improve drought characterization, hydrological simulation, and seasonal streamflow forecasting, with a focus on Morocco and the African continent. At the regional scale, drought monitoring is investigated in the Tensift River Basin (Morocco) through a comprehensive evaluation of satellite-based (CHIRPS v2) and reanalysis datasets (ERA5-Land) against in situ observations. The performance of drought indices, the Standardized Precipitation Index (SPI) and the Standardized Precipitation and Evapotranspiration Index (SPEI), is assessed by evaluating their ability to reproduce documented drought events using receiver operating characteristic (ROC) analysis. Results show that CHIRPS v2 precipitation and ERA5-Land temperature datasets reproduce hydroclimatic variability in the basin with satisfactory agreement. The Pearson Type III distribution is identified as the most suitable for SPI computation, while the log-logistic distribution is confirmed for SPEI. SPI at the 12-month timescale provides the highest skill in detecting reported drought events. This research also contributes to the evaluation of the Climate Hazards Center Infrared Precipitation with Stations, Version 3 (CHIRPS v3), an updated high-resolution precipitation dataset developed by the Climate Hazards Center at the University of California, Santa Barbara. Improvements in climatology, station coverage, satellite estimation, and gauge-undercatch correction enhance precipitation representation in semi-arid and mountainous environments, including Morocco, reinforcing its suitability for drought monitoring in data-limited regions. At the local scale, a seasonal streamflow forecasting framework is developed for the N’fis Basin, a drought-prone tributary of the Tensift River Basin. Statistical and machine-learning models predict streamflow at short lead times relevant to reservoir operations. Forecasts can anticipate inflow deficits 1–3 months before peak irrigation demand, supporting proactive reservoir management and illustrating the potential of forecast-informed reservoir operations for drought preparedness. At the continental scale, the performance of the Famine Early Warning Systems Network Land Data Assimilation System Forecast (FLDAS-Forecast), developed by NASA and FEWS NET to estimate hydroclimatic conditions in data-scarce regions, is evaluated against extensive in situ streamflow observations across Africa. FLDAS-Forecast streamflow outputs reproduce major hydrological signals and seasonal variability, showing useful predictive skill at short to medium lead times despite uncertainties related to land-surface modeling, meteorological forcing, and sparse gauge coverage. Finally, this dissertation demonstrates that integrating satellite observations, reanalysis products, land-surface modeling, and machine-learning approaches can substantially enhance drought monitoring and forecasting across spatial scales. By improving the capacity to detect, characterize, and anticipate drought impacts on water resources in data-scarce regions, the findings support drought early warning and adaptive water-resources planning across Africa, contributing to SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 13 (Climate Action), and SDG 15 (Life on Land).| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD Thesis MOHAMED NAIM PDF_Final version.pdf
embargo fino al 14/05/2027
Descrizione: Tesi def
Tipologia:
Tesi di dottorato
Dimensione
40.24 MB
Formato
Adobe PDF
|
40.24 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


