EUVIMEDOdkrijte EUVIMED
Crossref

Integrating Regularized Coefficient Optimization and Explainable Machine Learning to Improve Ecosystem Service Valuation

Peihong Song, Wenyin Wu, Ting Luo, Jing Luo, Jianan Yu, Jingli Lu, Zhongyi Sun, Jie Zhang, Rui Yu, Peng Wang, Yamin Jiang, Chuan Jin and 5 more

Ecosystem Health and Sustainability · 2026-01-01 · 12

DOI: 10.34133/ehs.0547 ↗

At a glance

Journal article in Ecosystem Health and Sustainability, published online on 2026-05-29. Written by Peihong Song and 16 co-authors. 53 references. Funded by: Key Research and Development Project of Hainan Province, Sanya Science and Technology Special Fund, Open Project Program of Huangshan Observation and Research Station for Land-Water Resources.

PovzetekBibliografski podatkiCiteRelated studies

Abstract (original language)

Ecosystem service value (ESV) assessment is essential for supporting ecological management and sustainable regional development. To improve ESV estimation for Hainan Island (HN), this study first quantified the ESV of Sanya using the biophysical-economic method and then integrated these estimates with existing ESV samples from multiple regions of HN to calibrate region-specific unit value coefficients using regularized nonnegative least squares. These coefficients, together with standard equivalence values and ecosystem area, were used to estimate major ecosystem services, including climate regulation, soil retention, water conservation, flood mitigation, and carbon sequestration and oxygen release. The extreme gradient boosting–Shapley additive explanations (SHAP) framework was further employed to examine the relative importance and nonlinear associations of natural and socioeconomic factors influencing ESV. The results showed that the total ESV of HN increased from 253.6 billion yuan in 2010 to 507.9 billion yuan in 2020 and that the fitted coefficient improved the regional applicability and consistency of the valuation results. Normalized difference vegetation index, elevation, gross domestic product density, population density, and distance to coast were among the most influential factors associated with ESV. Stratified analyses for the full-island, coastal, and inland regions further revealed regional heterogeneity in factor importance and SHAP dependence relationships. These findings highlight the value of coefficient calibration and spatially stratified interpretable machine learning for improving ESV assessment in tropical island systems.

Funding

  • Key Research and Development Project of Hainan Province
  • Sanya Science and Technology Special Fund
  • Open Project Program of Huangshan Observation and Research Station for Land-Water Resources
  • Key Laboratory of Biodiversity Conservation and Characteristic Resource Utilization in Southwest Anhui

Cite

APA

Song, P., Wu, W., Luo, T., Luo, J., Yu, J., Lu, J., Sun, Z., Zhang, J., Yu, R., Wang, P., Jiang, Y., Jin, C., Fang, M., Huang, H., Li, K., Yang, H., & Liu, W. (2026). Integrating Regularized Coefficient Optimization and Explainable Machine Learning to Improve Ecosystem Service Valuation. Ecosystem Health and Sustainability, 12. https://doi.org/10.34133/ehs.0547

Vancouver

Song P, Wu W, Luo T, Luo J, Yu J, Lu J, et al. Integrating Regularized Coefficient Optimization and Explainable Machine Learning to Improve Ecosystem Service Valuation. Ecosystem Health and Sustainability. 2026;12. doi:10.34133/ehs.0547

BibTeX
@article{Song2026,
  author = {Song, Peihong and Wu, Wenyin and Luo, Ting and Luo, Jing and Yu, Jianan and Lu, Jingli and Sun, Zhongyi and Zhang, Jie and Yu, Rui and Wang, Peng and Jiang, Yamin and Jin, Chuan and Fang, Mengyang and Huang, Haizhi and Li, Ke and Yang, Huai and Liu, Wenjie},
  title = {Integrating Regularized Coefficient Optimization and Explainable Machine Learning to Improve Ecosystem Service Valuation},
  journal = {Ecosystem Health and Sustainability},
  year = {2026},
  volume = {12},
  doi = {10.34133/ehs.0547}
}
Start for free →