Parameter Calibration in Global Land Carbon Models UsingSurrogate-based Optimization

Haoyu Xu,Tao Zhang,Yiqi Luo,Wei Xue,Xin Huang

DOI: https://doi.org/10.5194/gmd-2017-48

2017-01-01

Geoscientific Model Development Discussions

Abstract:Abstract. Soil organic carbon (SOC) has a significant effect on the carbon emission and climate change. However, current SOC prediction accuracy of most models is very low. Most evaluation studies indicate that the prediction error mainly comes from parameter uncertainties, which can be obviously improved by parameter calibration. Data assimilation technique has been successfully employed for parameter calibration of SOC models. However, data assimilation algorithms such as Bayesian Markov Chain Monte Carlo (MCMC) generally require a large amount of computation cost and are not appropriate for complex global land models. This study proposes a new parameter calibration method based on surrogate optimization techniques for improving the prediction of SOC. Experiments on three types of land carbon cycle models, including Community Land Model with Carnegie-Ames-Stanford Approach biogeochemistry sub-model (CLM-CASA’) and two microbial models show that surrogate-based optimization method is more effective and efficient than MCMC on both accuracy and cost. The root mean squared errors (RMSE) between predictions of different SOC models calibrated by surrogate-base optimization and observations can be reduced up to 12% compared to the results by using Bayesian MCMC. Meanwhile, the corresponding computation cost required is only one thousandth of that with Bayesian MCMC.