摘要
林地分等定级是一项基于多维度指标体系的全国性自然资源评估,各指标数据主要来自于“森林资源一张图”和“林草生态综合监测成果”数据集。在指标体系中,土壤腐殖质厚度为必选指标之一。然而,在数据集中常常存在缺失值,补充土壤腐殖质厚度数据需要大量的外业调查。为减少采集土壤腐殖质厚度数据的工作量,基于腐殖质厚度与已有林地及环境信息的关联性,探讨了是否能够通过预测模型可靠预测腐殖质厚度级别。研究使用“森林资源一张图”和“林草生态综合监测成果”中已有的林地及环境数据作为预测因子,包含植被覆盖类型、林地保护等级、龄组、郁闭度、生物量、自然度和枯枝落叶厚度等,构建了预测土壤腐殖质厚度级别的随机森林机器学习模型。研究结果表明:包含土壤因子,经过采样处理的随机森林机器学习模型能够可靠预测土壤腐殖质厚度级别,生产者精度均大于0.8,使用者精度均大于0.9。因此,在林地分等定级中,可通过已有数据预测和补充土壤腐殖质厚度的缺失数据,从而提高林地分等定级工作的效率。
Forest gradation and classification(FGC)is a national-wide natural resource evaluation based on indicator system.The data of indicators are mainly from the‘One Map of Forest Resource’and‘Ecological Monitoring of Forest and Grassland’.In FGC,humus thickness is a required indicator,of which filling the missing data by field surveys is labor-intensive.Here,predictive models for the humus thickness grade based on the already-existed data from‘One Map of Forest Resource’and‘Ecological Monitoring of Forest and Grassland’is built.It is found that the up-sampling random forest model with the predicators of forest cover,protection level,forest age,closeness,biomass,naturalness,and thickness of deciduous leaves layer,can robustly predict the humus thickness grade(User’s accuracy>0.8 and Producer’s accuracy>0.9).This model can be used to impute the missing values of humus thickness grade in the‘One Map of Forest Resource’.
作者
黄程
董子燕
陈拓
于海璁
刘颖
Huang Cheng;Dong Ziyan;Chen Tuo;Yu Haicong;Liu Ying(China Research Center for Special Economic Zones,Shenzhen University,Shenzhen,Guangdong 518061,China;Shenzhen Center for Evaluation and Development of Natural Resources and Real Estate,Shenzhen,Guangdong 518040,China)
出处
《绿色科技》
2023年第10期42-45,51,共5页
Journal of Green Science and Technology
关键词
林地分等定级
森林资源一张图
腐殖质厚度
预测模型
随机森林
forest gradation and classification
One Map of Forest Resource
humus thickness
predictive model
random forest
