Abstract. Traditional Budyko analysis is predicated on the assumption that the watershed of interest is in dynamic equilibrium over the period of study, and thus surface water partitioning will not be influenced by changes in storage. However, previous work has demonstrated that groundwater–surface water interactions will shift Budyko relationships. While modified Budyko approaches have been proposed to account for storage changes, given the limited ability to quantify groundwater fluxes and storage across spatial scales, additional research is needed to understand the implications of these approximations. This study evaluates the impact of storage changes on Budyko relationships given three common approaches to estimating evapotranspiration fractions: (1) determining evapotranspiration from observations, (2) calculating evapotranspiration from precipitation and surface water outflow, and (3) adjusting precipitation to account for storage changes. We show conceptually that groundwater storage changes will shift the Budyko relationship differently depending on the way evapotranspiration is estimated. A 1-year transient simulation is used to mimic all three approaches within a numerical framework in which groundwater–surface water exchanges are prevalent and can be fully quantified. The model domain spans the majority of the continental US and encompasses 25 000 nested watersheds ranging in size from 100 km² to over 3 000 000 km². Model results illustrate that storage changes can generate different spatial patterns in Budyko relationships depending on the approach used. This shows the potential for systematic bias when comparing studies that use different approaches to estimating evapotranspiration. Comparisons between watersheds are also relevant for studies that seek to characterize variability in the Budyko space using other watershed characteristics. Our results demonstrate that within large complex domains the correlation between storage changes and other relevant watershed properties, such as aridity, makes it difficult to easily isolate storage changes as an independent predictor of behavior. However, we suggest that, using the conceptual models presented here, comparative studies could still easily evaluate a range of spatially heterogeneous storage changes by perturbing individual points to better incorporate uncertain storage changes into analysis.