Vegetation as a source of predictability for subseasonal-to-seasonal forecasts
In recent years there has been substantial interest in improving the skill of subseasonal-to-seasonal (S2S; 2–8 weeks) forecasts for the atmosphere, which provide useful information for decision-makers across sectors such as agriculture and health. Accurately representing the interaction between the land surface and the atmosphere is key to developing skilful S2S forecasts, because the land surface state varies more slowly than the atmospheric state, and can thus provide predictability at longer lead times. Previous work has mostly focused on the predictability gained from realistic soil moisture initialisations, with little understanding of how vegetation behaves at these timescales or the predictability that could be gained from correctly representing it in models. In this talk, I will show how new Earth Observation datasets can allow us to understand the response of vegetation to subseasonal modes of rainfall variability, and demonstrate that in some regions vegetation feedbacks can provide a source of predictability for near-surface air temperatures out to 30 days. I will also show how a focus on the subseasonal behaviour of vegetation can provide useful diagnostics for evaluating water-carbon cycle coupling in Earth System Models.