Dynamic Phenology and Seasonal Forecast Systems
Monday, 15 December 2014
Dynamic carbon processes within a seasonal forecast system serve as an additional degree of freedom in the system and thus may have important impacts on forecast skill. Skill may be enhanced, for example, though knowledge of vegetation state (e.g., LAI) at the start of the forecast; vegetation that starts off in a relatively lush or relatively barren state (compared to other years) may stay in that state into the forecast period, with concommitant effects on forecasted meteorology. In addition, the ability of soil moisture and vegetation phenology to evolve together may amplify the impacts of soil moisture initialization, known to be a contributor to skill. We examine these effects with three separate ensembles of forecasts, each covering 63 boreal summers: (1) an ensemble of forecasts in which carbon states (including LAI) are initialized realistically ; (2) an ensemble of forecasts in which the dynamic phenology model operates as in (1), but the carbon states are initialized to the same nominal values on each forecast start date; and (3) an ensemble in which phenology is not a degree of freedom at all in the system, i.e., an ensemble with prescribed climatological seasonal cycles of LAI. Evaluating the seasonal forecasts of temperature and precipitation in each ensemble against observations -- as well as quantifying the inherent predictability of these variables within each ensemble -- provides a clear picture of the contribution of dynamic phenology to seasonal prediction.