Comparison of meteorological and plant hydraulic based water stress functions for modeling canopy phenology

Background/Question/Methods

A number of phenological models have been developed; however, the majority of research has been directed towards predicting phenology based solely on meteorological variables.  To develop more generalized phenology models it is preferable to utilize physiological responses of vegetation instead of relying solely on meteorological factors.  Additionally, while modeling the phenology of temperate and boreal ecosystems has received considerable attention, less has been studied in water-limited systems.  Our objective is to compare the use of plant hydraulics and meteorological variables for representing the role of water stress in phenological models.  Two contrasting AmeriFlux sites, a temperate deciduous broadleaf forest and a Mediterranean oak savanna, were selected to test the models under different environmental conditions.  We used the Terrestrial Regional Ecosystem Exchange Simulator (TREES), an ecosystem process model that incorporates plant hydraulics and a phenology submodel, to model the physiological response of vegetation to water stress.  Vapor pressure deficit (VPD) and predawn leaf water potential (ψ_pd) were used as the meteorological and hydraulic representation of water stress respectively.  Meteorologically and hydraulically driven water stress functions were then compared within a shared phenological modeling framework.  Both phenology model formulations were validated against eddy covariance data and ancillary variables at each site.

Results/Conclusions

At the woody savanna both models had a tendency to underestimate annual GPP but the hydraulically driven model reduced underestimation of annual GPP by an average of 86 g C m^-2 Year^-1 over the meteorologically driven model.  We then rescaled modeled GPP based on observed GPP to partition out errors due to the magnitude of modeled GPP and errors due to misrepresentation of seasonality.  Errors attributed to the seasonality of modeled GPP only differed by .66 g C m^-2 Year^-1 between the two models at the temperate deciduous site.  However, at the Mediterranean oak savanna the hydraulically informed phenology model reduced bias by 43.95 g C m^-2 Year^-1 over a purely meteorologically driven model.  The two models performed nearly identically at a well-watered site and the phenology model with a hydraulically driven water stress function performed better at a water limited site.  These results demonstrate the importance of accurately representing the influence of water stress on mediating phenology in water-limited systems.  Additionally, they suggest that phenological models which include plant physiology should be more generalizable across sites with varying conditions.  By including physiology we can begin to bridge the gap between empirical and mechanistic phenological models.