Journal cover Journal topic
Geoscientific Model Development An interactive open-access journal of the European Geosciences Union
doi:10.5194/gmd-2017-48
© Author(s) 2017. This work is distributed
under the Creative Commons Attribution 3.0 License.
Development and technical paper
05 Apr 2017
Review status
This discussion paper is under review for the journal Geoscientific Model Development (GMD).
Parameter Calibration in Global Land Carbon Models Using Surrogate-based Optimization
Haoyu Xu1, Tao Zhang1,2, Yiqi Luo2,3, Wei Xue1,2, and Xin Huang1,2 1Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China
2Department of Earth System Science, Ministry of Education Key Laboratory for Earth System Modelling, Tsinghua University, Beijing 100084, China
3Department of Microbiology and Plant Biology, University of Oklahoma, Norman, Oklahoma, USA
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.

Citation: Xu, H., Zhang, T., Luo, Y., Xue, W., and Huang, X.: Parameter Calibration in Global Land Carbon Models Using Surrogate-based Optimization, Geosci. Model Dev. Discuss., doi:10.5194/gmd-2017-48, in review, 2017.
Haoyu Xu et al.
Haoyu Xu et al.
Haoyu Xu et al.

Viewed

Total article views: 246 (including HTML, PDF, and XML)

HTML PDF XML Total Supplement BibTeX EndNote
183 53 10 246 76 1 12

Views and downloads (calculated since 05 Apr 2017)

Cumulative views and downloads (calculated since 05 Apr 2017)

Viewed (geographical distribution)

Total article views: 246 (including HTML, PDF, and XML)

Thereof 241 with geography defined and 5 with unknown origin.

Country # Views %
  • 1

Saved

Discussed

Latest update: 24 Apr 2017
Publications Copernicus
Download
Short summary
To reduce the parameter uncertainty in Soil Organic Carbon(SOC) models with reasonable computation cost, we apply Bayesian MCMC, global optimization and surrogate-based optimization (SBO) to three representative SOC models. The experiment results indicate that both global optimization and SBO can find the best parameters (lowest predication RMSE) and SBO we designed only requires no more than 100 samples. Thus, SBO is the most accurate and efficient method for SOC parameter calibration.
To reduce the parameter uncertainty in Soil Organic Carbon(SOC) models with reasonable...
Share