Ensemble-based parameter estimation for improving ocean biogeochemistry
in an Earth system model
Abstract
Improved ocean biogeochemistry (BGC) parameters in Earth System Models
can enhance the representation of the global carbon cycle. We aim to
demonstrate the potential of parameter estimation (PE) using an ensemble
data assimilation method to optimise five key BGC parameters within the
Norwegian Earth System Model (NorESM). The optimal BGC parameter values
are estimated with an iterative ensemble smoother technique, applied
a-posteriori to the error of monthly climatological estimates of
nitrate, phosphate and oxygen produced by a coupled reanalysis that
assimilates monthly ocean physical observed climatology. Reducing the
ocean physics biases while keeping the default parameters (DP) initially
reduces BGC state bias in the intermediate depth but deteriorates near
the surface, suggesting that the DP are tuned to compensate for physical
biases. Globally uniform and spatially varying estimated parameters from
the first iteration effectively mitigate the deterioration and reduce
BGC errors compared to DP, also for variables not used in the PE (such
as C0$_2$ fluxes and primary production). While spatial PE performs
superior in specific regions, global PE performs best overall. A second
iteration can further improve the performance of global PE for
near-surface BGC variables. Finally, we assess the performance of the
global estimated parameters in a 30-year coupled reanalysis,
assimilating time-varying temperature and salinity observations. It
reduces error by 20\%, 18\%,
7\%, and 27\% for phosphate, nitrate,
oxygen, and dissolved inorganic carbon, respectively, compared to the
default version of NorESM. The proposed PE approach is a promising
innovative tool to calibrate ESM in the future.