Hierarchy of Initial Conditions Reconstruction Strategies in a Perfect-Model Target Framework

State-of-the-art decadal prediction systems are initialized from reconstruction simulations - often achieved by Newtonian relaxation called nudging. While physical state variables of the climate system are nudged directly, biogeochemical state variables follow these physical variables indirectly instead of being directly nudged into the model themselves. Here we ask what the potential benefit of having perfect biogeochemical observations to reconstruct initial conditions to would be.

We define 50 years of a control simulation as our target representing observations. We nudge variables from this target onto arbitrary initial conditions 150 years later mimicking an assimilation simulation generating initial conditions for hindcast experiments of prediction systems. We investigate the tracking performance when nudging an increasing set of variables with a focus on biogeochemical variables explaining variations in atm. CO₂.

We find that indirect nudging of biogeochemistry reproduces the target variations with an anomaly correlation coefficient (ACC) above 0.6. While reproducing the large scale variations, nudging introduces small systematic biases in the physical state variables, on which biogeochemical cycles react very sensitively. Global surface oceanic pCO₂ initial conditions are best indirectly reconstructed for boreal winter months with ACC of 0.95 and root mean square error (RMSE) of 0.5 ppm. Direct reconstruction only lightly improves initial conditions in ACC by +0.1 only in boreal summer and RMSE by -0.2 ppm. Indirect reconstruction of global terrestrial biogeochemical initial conditions for terrestrial CO₂ flux track the target by ACC of 0.7 and RMSE of 0.1 PgC/month. Direct reconstruction brings improvements in ACC by +0.2 and RMSE by -0.02 PgC/month only from January to June. Global atm. CO₂ is indirectly tracked by ACC of 0.75 and RMSE of 0.6 ppm. Active reconstruction improves ACC by 0.15 and RMSE by -0.3 ppm.

Our research shows that indirect reconstruction of biogeochemistry yields satisfying results for much less effort and risk, especially for the ocean, whereas terrestrial biogeochemistry holds more potential for direct reconstruction. These results strengthen the current practice of indirect reconstruction in marine biogeochemical prediction systems.