Understanding and characterizing the intermittency of the river network in Africa : methodological developments and hydrological applications

Abstract

Intermittent rivers are rivers that stop flowing at some point in space and time. However, the understanding of the spatio-temporal variability of intermittent rivers as well as their precise location in the hydrographic network remains limited due to a lack of data (hydrometric, hydrographic, etc..). These difficulties are accentuated in data-limited regions such as Africa and this thesis aims to develop methodological approaches that rely on Random Forest models and statistical analyses to characterize the spatial distribution of intermittent rivers and to better understand the controlling factors of intermittency in Africa in order to overcome the observation gaps. Firstly, this thesis focused on the regional scale in Burkina Faso where 49 stream gauging stations with at least four years of data over the period 1955-1985 were examined. The mean number of months with zero flow per year ((Ndry) ̅) was used as a predictor to define four increasing classes of flow intermittency namely: permanent (0-1 months with zero flow), weakly intermittent (2-4), highly intermittent (5-7), ephemeral (8-12). A principal component analysis (PCA) performed on 49 gauging stations showed that, although the Strahler order and the average annual precipitation influence the geographical distribution of the different intermittency classes in Burkina Faso, the average permeability and the upstream catchment area mainly explain this distribution. This study suggests that the seasonality of precipitation in Burkina Faso also makes streamflow seasonal regardless of mean annual precipitation unless hydrogeological processes are involved, notably through the contribution of aquifers to baseflow during low-flow periods. The Random Forest model estimated that 88% of the total length of the rivers in Burkina Faso are intermittent compared to 98% in the national reference hydrographic databases (IGB-BNDT). Secondly, at the African scale, 1125 gauging stations with at least 4 years of