Le mardi 3 mars 2026,  Deogratias NAHAYO présentera l'examen en vue de l’obtention du grade académique de Docteur en Sciences (Collège de doctorat en Sciences et gestion de l'environnement) sous la direction de Bernard TYCHON et Matthias VANMAERCKE.

Cette épreuve consistera en la défense publique d’une dissertation intitulée :

« Hydrogeomorphological approaches to understanding flood hazards and flood-risk management in northwest Rwanda ».

Le Jury sera composé de :

M. P.  OZER (Président), MM. O. DEWITTE (Musée royal d’Afrique centrale), E. RUKUNDO (Rwanda Water Resources Board), B. TYCHON (Promoteur), M. VANMAERCKE (KULeuven) (Co-promoteur), U.G. WALI (Université du Rwanda), J. WELLENS (Secrétaire).

Rejoindre la défense en visio

Abstract

Limited and unreliable hydrometeorological data often constrain flood hazard assessment and risk management in tropical mountainous regions. As a result, it is frequently difficult to clearly understand how rainfall is translated into runoff and flood response. This thesis tackles this challenge by investigating how topography, soil characteristics, and land use/land cover influence flood behavior in two contiguous agricultural catchments in the Mukungwa watershed of northwest Rwanda: Nyamutera (44 km²) and Gaseke (109 km²). Drawing on short-term but high-resolution rainfall and streamflow observations, combined with statistical and numerical modeling approaches, the study develops a more process-based understanding of flood dynamics in mountainous, data-scarce environments.
Low-cost monitoring systems were deployed to generate high-resolution rainfall and discharge datasets, enabling detailed analysis of storm events. The Rational method was also examined, from its empirical formulation to its physical interpretation. The results reveal clear contrasts between the steep Nyamutera catchment and the floodplain-connected Gaseke catchment. Nyamutera responds rapidly and produces more pronounced flood peaks, whereas Gaseke exhibits more moderate flows due to floodplain storage and hydrological connectivity. These differences underscore the key role of hydrogeomorphic controls in shaping flood response and support a shift from purely empirical approaches toward process-informed flood analysis.
The performance of two additional hydrological models was also evaluated. A custom Python-based framework using the NRCS-CN method produced mixed results, performing better in Gaseke than in Nyamutera and showing strong event-to-event variability and limited transferability. The GR4J rainfall-runoff model successfully reconstructed historical flows and helped define early-warning thresholds. Critical antecedent rainfall windows for flood prediction were identified as 5–6 days for Nyamutera and 7–11 days for Gaseke.
Overall, the study shows that even short-term monitoring, when paired with tailored modeling approaches, can substantially improve flood understanding and early-warning capacity in data-limited mountainous regions.

Lien Orbi