Predicting the hydrological consequences following changes in grassland vegetation type (i.e., woody encroachment) requires an understanding of water flux dynamics at high spatiotemporal resolution for predominant species within grassland communities. However, grassland fluxes are typically measured at the leaf or landscape scale, which inhibits our ability to predict how individual species contribute to changing ecosystem fluxes. We used external heat balance sap flow sensors and a hierarchical Bayesian state-space modeling approach to bridge this “flux-gap” and estimate continuous species-level water flux in common tallgrass prairie species. Specifically, we asked: 1) How do diurnal and nocturnal water fluxes differ among woody and herbaceous plants? (2) How sensitive are woody and herbaceous species to environmental drivers of diurnal and nocturnal water flux? We highlight three results: (1) Cornus drummondii, the primary woody encroacher in this grassland, exhibited the greatest canopy-level water loss, (2) nocturnal transpiration was a large component of the water lost in this ecosystem and was driven primarily by C₄ grasses and C. drummondii, and (3) the sensitivity of canopy transpiration to environmental drivers varies among plant functional types and throughout a 24-hour period. Our data reveal important insights regarding the water-use strategies of woody versus herbaceous species in tallgrass prairies, and about the potential hydrological consequences of ongoing woody encroachment. We suggest that the high, static flux rates observed in woody species will likely deplete deep water stores over time, potentially creating hydrological deficits in grasslands experiencing woody encroachment and concomitantly increasing the vulnerability of these ecosystems to drought.

Please see manuscript for a description of data collection and processing.

The readme file describes the variables within the .csv data files accompanying the above publication. Within each file, each row indicates an individual 30-minute time interval, and missing data indicated by “NA.”