A large fraction of uncertainty in future carbon–climate feedbacks originates not in the atmosphere or ocean, but in terrestrial ecosystem dynamics.

This is not primarily a question of missing carbon pools, but of structural uncertainty in how land processes are represented in Earth system models.

1. Nonlinear vegetation dynamics

Vegetation is not a passive tracer of climate. It exhibits:

• threshold behaviour
• disturbance-driven regime shifts
• competition across scales of space and time

These dynamics are only partially resolved in most ESMs.

2. Disturbance as a first-order control

Fire, pests, and land-use change act as:

• fast carbon release mechanisms
• structural resets of ecosystem state
• modifiers of successional trajectories

Yet they are often represented via empirical or semi-empirical schemes.

3. Soil and legacy effects

Soil carbon turnover is strongly influenced by:

• microbial processes
• hydrological state
• historical disturbance

These create long memory effects that are difficult to constrain observationally.

4. Why this dominates uncertainty

Even when atmospheric physics is well constrained, land surface feedbacks introduce:

• divergence in net land sink strength
• divergence in transient CO₂ trajectories
• uncertainty in timing of peak carbon uptake

In many projections, land processes are the leading source of spread in cumulative emissions-compatible pathways.

Conclusion

Reducing uncertainty in climate projections is increasingly a question of: improving structural representation of land–biosphere systems, not just refining atmospheric physics.