Background/Question/Methods Climate change is expected to bring greater variability in the amount and frequency of rainfall in future. Grassland productivity is strongly affected by rainfall variability, mainly via the timing of foliage green-up (i.e., when new growth exceeds senescence) and brown-down (i.e.,when senescence exceeds new growth). New ecosystem monitoring techniques provide opportunities to examine the processes governing grassland phenology, thereby facilitating the development of predictive ecosystem models. Here, we used fixed cameras (phenocams) to monitor grassland “greenness”, as a proxy for foliage phenology. Greenness, along with meteorological data, was measured at the plot scale in rainfall manipulation field experiments (n=52) and pasture monitoring sites (n=1) in western Sydney, Australia. These sites include both monospecies and multispecies plots. The data were used to evaluate an empirical model linking phenology to soil water content (SWC) and temperature. Within a model data assimilation framework, we quantified the sensitivities of leaf growth and senescence to SWC.
Results/Conclusions Previous models have assumed a linear sensitivity of growth to SWC. In contrast, here we found that the sensitivity of growth to SWC varied by species, with the majority of species showing a non-linear, concave-down response. Similarly, models commonly assume a constant senescence rate, whereas we found a strong nonlinear increase in senescence rate with declining SWC. Incorporating the sensitivities of growth and senescence to SWC helped predict the thresholds of greenness and SWC of the onset of brown-down and was important to reproduce the observations. Model evaluation against data under drought treatments indicated that these sensitivities to SWC enabled the prediction of drought impacts on greenness, and helped explain differences among species’ responses to variable rainfall. Our results suggested the modelling framework as a useful way to improve the representation of soil moisture impacts on grassland phenology in land surface models.