摘要
森林蓄积量的研究对了解森林生态系统、林分生产力、森林生物量具有重要意义,探究影响 4 种树种(组)林分蓄积量变化的因子,为后期建立天然林生长模型构建提供理论支撑.以福建省最新一次的森林资源连续清查中的天然马尾松、阔叶林树种(组)、针阔混交树种(组)、针叶混交树种(组)的蓄积量为研究对象,气象、地貌等环境因子为自变量,利用决策树回归、随机森林回归、adaboost回归、梯度提升树回归(GBDT)、CatBoost回归、ExtraTrees回归、XGBoost回归、LightGBM回归方法分析环境因子对 4种天然林树种(组)蓄积量的影响情况开展探讨.结果表明:梯度提升树回归(GB-DT)能较好地拟合各环境因子与 4 种树种(组)蓄积量的关系,4 种树种(组)蓄积量R2 均为 0.999,MSE、RMSE、MAE、MAPE均在 0.1 范围内;林分年龄与蓄积量的密切关系,重要性达 0.50 以上;气象和地貌因子对 4 种树种(组)蓄积量的重要性存在差异,建议在具体建模过程中应进行剥离分析.
Abstract
The study of forest volume is of great significance for understanding forest ecosystems,stand productivity,and forest biomass.It explores the factors that affect the changes in forest volume of four tree species(groups)and provides theoretical support for establishing natural forest growth models in the future.The study focuses on the stock volume of natural Pinus massoniana,broad-leaved forest species(groups),mixed coniferous and broad-leaved tree species(groups),and mixed coniferous tree species(groups)in the latest continuous inventory of forest resources in Fujian Province.Environmental factors such as meteorology and geomorphology are used as in-dependent variables,and decision tree regression,random forest regression,adaboost regression,gradient lifting tree regression(GBDT),CatBoost regression,ExtraTrees regression,XGBoost regression are used The LightGBM regression method is used to analyze the impact of environmental factors on the stock volume of four natural forest tree species(groups).The results showed that Gradient Ascending Tree Regression(GBDT)could better fit the relationship between environmental factors and the stock volume of four tree species(groups).The stock volume R2 of all four tree species(groups)was 0.999,and MSE,RMSE,MAE,and MAPE were all within the range of 0.1;The close relationship between stand age and stock volume,with an importance of over 0.50;There are differences in the importance of meteo-rological and geomorphological factors on the accumulation of four tree species(groups),and it is recommended to conduct stripping anal-ysis during the specific modeling process.
基金项目
福建省自然科学基金(2021J011141)
南平市资源化学产业科技创新联合项目(N2021Z010)
福建省哲学社会科学规划项目(FJ2022X019)
大学生创新训练项目(S202210397048)
NETL
NSTL
万方数据