Using FLUXNET Data to Improve Models of Springtime Vegetation Activity Onset in Forest Ecosystems

Despite recent efforts, accurate representation of vegetation phenology in land surface models has been elusive. As a result, current schemes exhibit both unrealistic variability in phenology and biased predictions for leaf area dynamics and the seasonality of surface-atmosphere fluxes. Improvements in model representation of phenology, especially the timing of key transition dates such the spring onset of leaf growth and photosynthetic activity, are urgently needed to support predictions of how climate change will affect ecosystem function and regional carbon budgets. Using meteorological and eddy covariance data provided by the FLUXNET ‘La Thuile' database, we evaluated 11 different models driven by air temperature and photoperiod to simulate spatial and temporal variability in spring phenology. To do this, we used a subset of sites in the La Thuile database located in ecosystems with distinct seasons (active/dormant) and where temperature is considered its primary driver of seasonality. We then compared the results from these models with other prediction schemes, including phenology sub-routines currently implemented in the land surface parameterizations of several large-scale (ecosystem to earth system) models. The root mean square errors produced by our modified growing degree-day models range from 9-19 days and provide substantial improvement over standard approaches. Results from this study demonstrate that relatively simple variants of widely used phenology models can be used to predict the timing of key growing season events with good realism. By extension, these models should provide an improved basis for making predictions about how the phenology and carbon budgets of these ecosystems may change in the coming decades.