Observations of carbon allocation in the world’s forests must match pace with vegetation model development

The data requirements of vegetation models are changing. For more than a decade, the community has been developing “next-generation” models that should be globally applicable and at the same time incorporate great process detail. The individual tree emerges from this development as the finest scale at which carbon, water, and nutrient dynamics can be realistically simulated. As such, precise tree-level observations of the relevant processes would ideally be available from across all forested biomes to inform and evaluate tree-centered vegetation models. This is not the case. Instead, we note a growing discrepancy between the demand for and the availability of highly-resolved measurements of carbon allocation in trees and forests.

To exemplify this discrepancy, we conducted a survey at 90 flux-tower sites from around the world that revealed priorities and deficiencies in existing data collections. We found that forest structure and aboveground carbon stocks have been ubiquitously inventoried, and that tree growth and foliage turnover have also been measured at many sites. By contrast, detailed information on water cycling, volume increment, and wood formation processes (especially belowground) are less common, as are records of tree mortality or terrestrial and airborne LiDAR that could help scale local observations. In addition, we found that the temporal resolution and length of existing time-series vary substantially across the current flux-tower network. Weighing the strengths and limitations of this and many other ecological monitoring networks, we conclude that the present data basis is insufficient to support accelerating vegetation model development.

Looking forward, we anticipate that not only the amount of tree-level observations needs to be increased – especially in tropical and boreal systems – but that the consistency, scalability, and predictability of forest carbon cycle observations needs to be improved. We also propose that intensive long-term monitoring sites be strategically paired with manipulative experiments at comparable sites to better connect past, present, and expected future dynamics. For this, we propose a versatile experimental framework and call for a community-wide discussion on the “yield on cost” of various field observations. We also list a number of key questions on how to best build and maintain cross-scale data archives in support of tree-centered vegetation modelling.