ADEARTH

Advances in Earth Science

   

Weighted Kernel Centroid Optimization Method and Its Application in Optimizing Dynamic Vegetation Parameters of a Land Surface Model: A Case Study of Grassland Areas in the Yellow River Basin*

Liu Yonghe1， Liu Long2， Huang Pinghua1， Yan Yin1， Liu Siyuan2   

  1. (1. School of Resources and Environment, Henan Polytechnic University, Jiaozuo Henan 454000, China; 2. School of Design, Henan Polytechnic University, Jiaozuo 454000 Henan, China)

  • About author:Liu Yonghe, mainly engaged in land surface processes and hydrology research. E-mail: yonghe_hpu@163.com
  • Supported by:
    Project supported by the Natural Science Foundation of Henan Province (General Program) (Grant No. 252300421467); Henan Provincial Key Science and Technology Program (Grant No. 252102320004).
PDF ( ) Cited

Liu Yonghe, Liu Long, Huang Pinghua, Yan Yin, Liu Siyuan. Weighted Kernel Centroid Optimization Method and Its Application in Optimizing Dynamic Vegetation Parameters of a Land Surface Model: A Case Study of Grassland Areas in the Yellow River Basin*[J]. Advances in Earth Science, DOI: 10.11867/j.issn.1001-8166.2026.068.

Abstract：When traditional methods are used to implement parameter optimization for land surface process models, they face challenges such as slow convergence and a tendency to fall into local optima, and they also suffer from poor stability for models with non-deterministic outputs. This paper proposes a new global optimization algorithm—the Weighted Kernel Centroid (WKC) method. By calculating the centroid of superior solutions through inverse fitness weighting and combining it with an adaptive space contraction strategy, it achieves exploration on cost functions that are unimodal or ridge-type and contain fluctuating noise. To verify its effectiveness, this study conducted tests on a variety of standard mathematical functions, and, based on 49 sample points representing grassland types in the Yellow River Basin, carried out joint optimization of four dynamic vegetation parameters in the NMH-CS model, namely VCMX25 (carboxylation rate at 25 degrees C), MP (stomatal conductance slope), DILEFC (leaf death coefficient due to temperature stress), and QE25 (quantum efficiency at 25 degrees C). A systematic comparison with the genetic algorithm (GA), steepest descent method (GD), and shuffled complex evolution algorithm (SCE-UA) is conducted. The results show that: ① WKC has a certain global optimization capability, but is more suitable for non-multimodal problems; ② WKC has the advantages of fast convergence and strong stability; ③ parameter optimization brings the correlation coefficient of leaf area index and gross primary productivity simulations to around 0.8, with RMSE and relative bias well controlled; ④ the optimized parameters exhibit significant spatial heterogeneity, where VCMX25 is significantly correlated with altitude (r=0.89), reflecting photosynthetic adaptation characteristics under high-altitude lowtemperature environments; MP is weakly positively correlated with precipitation, reflecting the water conservation property of reduced stomatal conductance in arid regions. This study provides a concise, efficient, and stable parameter optimization method, and through parameter optimization for a land surface model reveals the spatial differentiation patterns of key grassland vegetation parameters in the Yellow River Basin and their environmental adaptation mechanisms.
No related articles found!
Viewed
Full text


Abstract

Address: No. 8, Tianshui Middle Road, Lanzhou, Gansu
QQ：114723517
Tel：0931-8762293　E-mail：adearth@lzb.ac.cn
陇ICP备2021001824号-5　 甘公网安备62010202000687