Controls over Amazonian transpiration in the Community Land Model, v5.

Abstract:

The estimation of evapotranspiration in tropical forest ecosystems has long proved difficult for land surface models (LSM's). Typically, flux observations suggest that transpiration is primarily demand-limited over much of Amazonia, but LSM's often struggle to represent the unstressed transpiration during the dry season. We investigate the model features which lead to this pathological behaviour within the context of the in-development Community Land Model, version 5. This model is notable from a hydrological perspective on account of its representation of variable soil depths (a novel feature in the context of Earth System Models) and also on account of the novel representation of the Nitrogen cycle. The default version of the model, however, continues to represent a 'wrong-way round' response to seasonal rainfall, which in turn produces biases representations of land-atmosphere interaction in the context of the coupled system. In this study we probe the high dimensional parameter space of the model in single point simulations to understand which aspects of the representation of the water, carbon and nitrogen cycles exerts the greatest control over the capacity of the model to realistically represent tropical transpiration, assess the physiolgical realism of changes which might improve model biases, and indicate paths towards improved Amazonian ecohydrology in the Community Earth System Model.