Towards a global leaf phenology model

Leaf phenology, often measured by the seasonal dynamics of leaf area index (LAI), is a key control on the exchanges of CO₂ and energy between land ecosystems and the atmosphere. It is therefore also a key target process for dynamic vegetation models. However, there is no agreement on how leaf phenology should be modelled. Much research has focused on the specific triggers for budburst– and, to a lesser extent, leaf senescence– in biomes characterized by distinct cold or dry seasons. Recent theoretical developments however suggest the existence of a more general, global relationship between leaf phenology and the seasonal time course of “steady-state LAI”: the LAI would be in equilibrium with GPP if weather conditions were held constant. This can be predicted from the time course of gross primary production (GPP) because LAI and GPP are mutually related, via the Beer’s law dependence of GPP on LAI, and the requirement for GPP to support LAI development. In our current research we are developing a new global phenology model, by combining this new theoretical approach with a terrestrial photosynthesis model (the P-model) that avoids the multiplicity of parameters required by more complex models, while achieving good fit to GPP derived from flux towers in all biomes. But whereas P-model applications to date have exploited satellite-derived green vegetation cover indices as input, our current research aims to predict the seasonal time course of both LAI and GPP. This is done in two steps. First, we predict seasonal maximum LAI as the lesser of an energy-limited value that maximizes GPP, and a water-limited value that allows vegetation to transpire a fraction of annual precipitation. Second, we model the time-course of LAI assuming that its derivative tracks the difference between current and steady-state LAI with some lag. We are testing this approach with data from a global phenocam network and using remotely sensed LAI. Results so far are promising, but point to challenges, especially in representing interannual variability and trends.