Hysteresis metrics reveal land-use-dependent stormflow constituent dynamics in tropical Afromontane catchments

Abstract. High-frequency monitoring can resolve storm-event transport processes obscured by mean concentrations and low-frequency sampling, yet these dynamics remain poorly understood in tropical African headwaters. We analysed a six-year (2015–2020), 10-min record of rainfall, discharge, nitrate, dissolved organic carbon, electrical conductivity, and total suspended solids from natural forest, smallholder agriculture, and tea/tree plantation catchments in the South-West Mau, Kenya. Across 695 rainfall–runoff events, 2,506 event–constituent combinations were classifiable. Chemostasis dominated overall (70 %), followed by mobilization (27 %) and dilution (3 %). Electrical conductivity (98 %), nitrate (89 %), and dissolved organic carbon (66 %) were predominantly chemostatic, whereas total suspended solids were predominantly mobilized (69 %). Of 1,472 responsive combinations, 1,323 yielded valid hysteresis geometry and 1,319 also had finite flushing index values for clustering. Joint analysis of the Lloyd hysteresis index, normalized loop area, and flushing index identified three stable response clusters of similar prevalence (33–34 %): delayed flushing (Cluster 1: 33 %), rapid mixed response (Cluster 2; 34 %), and rapid flushing (Cluster 3; 33 %).

Catchment–constituent setting strongly structured membership: Cluster 1 was dominated by dissolved organic carbon responses in natural forests and tea/tree plantations. Cluster 2 was characterized by dissolved constituents in smallholder agriculture and nitrate in the tea/tree plantation catchments. Finally, Cluster 3 was controlled by total suspended sediments, with highest occurrences in the natural forest and tea/tree plantation catchment. Antecedent dry period, discharge-peak count, and event rainfall differed among response groups, but their distributions overlapped substantially. Land use and constituent type therefore structured storm-event hysteresis and flushing responses more strongly than individual hydrological conditions.