Improving forecasting of biome shifts with data assimilation of paleoecological data

Biomes are shifting their distributions in response to both sudden disturbance and gradual climate changes. We use models to predict biome shifts, but it can be difficult to assess the importance of sudden disturbance because observations of past biome shifts are smooth typically do not show the effects of individual events. Combing models directly with observations of past biome change will help us determine how important sudden disturbance verses gradual climate change is for long term biome change and improve forecasts of biome changes. Here we use state data assimilation within the predictive ecosystem analyzer (PEcAn) to constrain a simple forest gap model (LINKAGES) with mechanisms for both sudden and gradual changes with paleoecological data from Harvard Forest that contains a biome change over 1100 years from a northern hardwood to southern oak hickory forest. We found that sudden climate changes caused by extreme events had large effects on stand structure which catalyzed long term change, ultimately driving the long term shift in vegetation. This finding has direct implications for improving forecasts of biome shifts caused by current and future climate changes and points to including better representations of the long term effects of extreme events on vegetation.