The principal aim of this study was to disentangle hydrochemical influences on primary producers in a pristine, flood-pulse ecosystem. This was undertaken by analysing diatoms from 100 sample points from hydrologically distinct regions in the Okavango Delta, Botswana. Cluster analysis was undertaken using two-way indicator species analysis (TWINSPAN), and groups used to classify sample points in a principal components analysis (PCA) biplot. Linear discriminant analysis (LDA) was performed using hydrological data and diatom guilds as explanatory variables. A series of ordinations using redundancy analysis (RDA) was undertaken to assess which variables significantly accounted for diatom variation across the Delta. Species-response curves for major taxa were generated using generalized additive models (GAMs). Cluster analysis revealed six distinct groups. Groups 5 and 6 consisted mainly of seasonally inundated floodplain sites, which lay at one end of a significant gradient revealed by PCA. Floodplain diatoms were characteristically N-heterotrophs, requiring elevated concentrations of key resources such as total nitrogen (TN) and SiO2. Using forward selection, constrained RDA reveals five variables were significant in explaining diatom distributions across the Delta: hydroperiod class, flood frequency, flow velocity and nutrients SiO2 and TN. Species-response curves show that motile diatoms were most abundant in seasonally inundated floodplains. Species diversity was significantly higher in the upper Panhandle region of the Delta, which may be related to moderate levels of disturbance and increased resource limitation. Species diversity was significantly lower during the period of maximum flood extent, which may in turn be related to fewer limiting resources. Keywords: diatoms, biodiversity, wetlands, multivariate techniques, water level regime, diversity indices, Okavango Delta, Botswana.