When soil moisture is heterogeneous, sap flow from, and ABA status of, different parts of the root system impact on leaf xylem ABA concentration ([X-ABA]leaf). The robustness of a model for predicting [X-ABA]leaf was assessed. 'Two root-one shoot' grafted sunflower (Helianthus annuus L.) plants received either deficit irrigation (DI, each root system received the same irrigation volumes) or partial rootzone drying (PRD, only one root system was watered and the other dried the soil). Irrespective of whether relative sap flow was assessed using sap flow sensors in vivo or by pressurization of de-topped roots, each root system contributed similarly to total sap flow during DI, while sap flow from roots in drying soil declined linearly with soil water potential (Psisoil) during PRD. Although Psisoil of the irrigated pot determined the threshold Psisoil at which sap flow from roots in drying soil decreased, the slope of this decrease was independent of the wet pot Psisoil. Irrespective of whether sap was collected from the wet or dry root system of PRD plants, or a DI plant, root xylem ABA concentration increased as Psisoil declined. The model, which weighted ABA contributions of each root system according to the sap flow from each, almost perfectly explained [X-ABA] immediately above the graft union. That the model overestimated measured [X-ABA]leaf may result from changes in [X-ABA] along the transport pathway or an artefact of collecting xylem sap from detached leaves. The implications of declining sap flow through partially dry roots during PRD for the control of stomatal behaviour and irrigation scheduling are discussed.