欢迎访问林草资源研究

林草资源研究 ›› 2024›› Issue (5): 48-55.doi: 10.13466/j.cnki.lczyyj.2024.05.006

• 科研研究 • 上一篇    下一篇

哑变量模型和混合模型的比较——以构建不同区域杉木林和落叶松林生物量模型为例

邹文涛1(), 曾伟生2(), 杨学云2, 温雪香2   

  1. 1.中国林业科学研究院林业科技信息研究所,北京 100091
    2.国家林业和草原局林草调查规划院,北京 100714
  • 收稿日期:2024-08-26 修回日期:2024-09-26 出版日期:2024-10-28 发布日期:2025-04-18
  • 通讯作者: 曾伟生,教授级高级工程师,博士,主要从事森林资源调查监测与林业数学建模等工作。Email:zengweisheng0982@126.com
  • 作者简介:邹文涛,副研究员,博士,主要从事林业遥感和森林生态系统固碳能力评估工作。Email:zouwentao1982@126.com
  • 基金资助:
    国家重点研发计划课题“典型人工林立地质量评价与生产力提升技术”(2022YFD2200501)

Comparison of Dummy Variable Model and Mixed Model: A Case Study on Constructing Biomass Models for Cunninghamia lanceolata and Larix spp.Forests in Different Regions

ZOU Wentao1(), ZENG Weisheng2(), YANG Xueyun2, WEN Xuexiang2   

  1. 1. Research Institute of Forestry Policy and Information,Chinese Academy of Forestry,Beijing 100091,China
    2. Academy of Inventory and Planning,National Forest and Grassland Administration,Beijing 100714,China
  • Received:2024-08-26 Revised:2024-09-26 Online:2024-10-28 Published:2025-04-18

摘要:

以构建杉木林和落叶松林生物量模型为例,利用第九次全国森林资源清查的3 152个杉木林和2 495个落叶松林固定样地数据,将反映不同区域的分类变量作为哑变量或随机变量,对比分析哑变量模型和混合模型两种建模方法。结果显示:两种建模方法得出的不同区域杉木林和落叶松林生物量模型的确定系数(R2)均达到0.9以上,平均预估误差(EMP)均小于1.5%,总体相对误差(ETR)均等于或接近于0,平均系统误差(EAS)基本在±5%以内,平均百分标准误差(EMPS)大多数小于15%;华东、中南、西南3个区域的杉木林生物量模型之间存在一定程度的差异,但中南和西南地区之间的差异不显著;东北、华北、西部3个区域的落叶松林生物量模型之间均存在显著差异。结果表明:哑变量模型和混合模型两种建模方法均可用于比较不同区域或不同类型林分生物量模型的差异并分析其差异显著性,且结果基本一致,混合模型比哑变量模型更为适用,其结果也更为稳定;所建生物量模型为国家及区域尺度的杉木林和落叶松林生物量估计提供了科学依据。

关键词: 哑变量, 随机变量, 生物量模型, 加权回归, 落叶松, 杉木

Abstract:

This study develops and compares dummy variable models and mixed models for biomass modeling of Cunninghamia lanceolata and Larix spp.forests.Using data from 3 152 Cunninghamia lanceolata and 2 495 Larix spp.permanet sample plots collected during the 9th national forest inventory.Indicative variables representing three distinct regions were incorporated as either dummy or random variables in the models.The results demonstrated that the determination coefficients(R2)of the biomass models from two approaches for Cunninghamia lanceolata and Larix spp.forests in different regions exceeded 0.9.The models achieved mean prediction errors(EMP)under 1.5%,the total relative errors(ETR)near zero,the average systematic errors(EAS)within ±5%;and the mean percent standard error(EMPS) almost under 15%.While biomass models for Cunninghamia lanceolata forests differed among East,Central-South and Southwest China,difference between South and Southwest was not significant.There were significant differences among the biomass models of Larix spp.forests in Northeast,North and West China.The study confirms that both dummy and mixed variable models can effectively compare and analyze regional and typological differences in stand-level biomass.However,the mixed model proved more robust and applicable.The developed biomass models offer a scientific foundation for estimating biomass of Cunninghamia lanceolata and Larix spp.forests on national and regional scales.

Key words: dummy variable, random variable, biomass model, weighted regression, Cunninghamia lanceolata, Larix spp.

中图分类号: