Spatial Patterns and Controlling Factors of River Flow Intermittency in Africa

Abstract Understanding river flow intermittency remains a major challenge in data‐scarce regions such as Africa. This study presents the first high‐resolution, four‐class intermittency map of African rivers using a sequential modeling approach based on Random Forest. A binary model (BC‐UNS) distinguishes perennial from non‐perennial reaches across 15.5 million river reaches, while a multiclass model (MC‐WOR) further categorizes non‐perennial reaches as weakly intermittent, highly intermittent, or ephemeral. A reach was considered non‐perennial when the long‐term mean annual number of zero‐flow months exceeded one; otherwise, it was considered perennial. Models were trained on data from 1,269 gauging stations. BC‐UNS demonstrated substantial performance, correctly identifying 92% of perennial stations and 71% of non‐perennial stations. MC‐WOR achieved fair performance, correctly identifying 73% of weakly intermittent, 45% of highly intermittent, and 58% of ephemeral stations. Climate variables, particularly the aridity index, emerged as the dominant controls of intermittency, with catchment area, potential evapotranspiration, and permeability also playing important roles. Human influences, represented by indices such as the Human Footprint Index and dam storage, further affected intermittency, especially among non‐perennial subclasses. Results indicate that 69% of Africa's river length is non‐perennial, comprising 7% weakly intermittent, 16% highly intermittent, and 46% ephemeral reaches. Differences in definitions between this study and previous global assessments limit direct comparisons, although all studies consistently indicate that most African rivers are non‐perennial. This study advances previous global efforts by producing a fine‐scale, reach‐level classification map that distinguishes four flow intermittency classes, then offering actionable insights for water resource planning and ecosystem conservation.