The linear behavior of the joint initial-boundary-value predictability of the climate system

The primary sources of predictability of the Earth climate system come from initial conditions and external radiative forcings, which is a joint initial-boundary-value problem. It is crucial to clearly understand the contributions of initial conditions and external forcings (i.e. initial-value and boundary-value) to the total predictability, for which the traditional predictability study examines the growth of system uncertainties is however difficult to quantitatively identify. This study uses the variations of prediction skills in the space of prediction lead times to systematically examines the linear behaviors of initial-value, boundary-value and joint initial-boundary-value predictability, which are measured by anomaly correlation coefficients between system states in a series of perturbed experiments on initial and boundary conditions. The lead time scale at which the prediction skill induced from external forcing exceeds the skill from the initial condition is a key measurement of system uncertainty, after which the boundary-value predictability efficiently extends the total predictability. These results provide a guideline to understand how the accurate setting of external forcing caused by human activities promotes the prediction skill when a comprehensive Earth climate system model is initialized to make climate predictions.

Automated summary

This study examines the linear behaviors of initial-value, boundary-value and joint initial-boundary-value predictability using the anomaly correlation coefficients between system states. It finds that boundary-value predictability efficiently extends the total predictability when the prediction skill induced from external forcing exceeds the skill from the initial condition.