文献类型：期刊文献

中文题名：基于最大熵模型的吉林省主要天然林潜在分布适宜性

英文题名：The Suitability of Potential Geographic Distribution of Natural Forest Types in Jilin Province Based on Maximum Entropy Models

作者：刘丹[1] 李玉堂[1] 洪玲霞[1] 国红[1] 谢阳生[1] 张卓立[1] 雷相东[1] 唐守正[1]

第一作者：刘丹

通信作者：Lei, Xiangdong

机构：[1]中国林业科学研究院资源信息研究所国家林业与草原局森林经营与生长模拟重点实验室

年份：2018

卷号：54

期号：7

起止页码：1-15

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;EI(收录号：20184406012456);Scopus;北大核心:【北大核心2017】；CSCD:【CSCD2017_2018】；

基金：林业公益性行业科研专项"我国主要林区林地立地质量和生产力评价研究"(201504303)

语种：中文

中文关键词：森林资源连续清查数据;天然林;混交林;物种分布模型;分布适宜性制图

外文关键词：national forest inventory data;natural forest;mixed forests;maximum entropy models;distribution suitability mapping

分类号：S757

摘要：【目的】研究基于物种分布模型预测天然林类型适宜分布区的方法,探讨影响天然林类型适宜性分布的主导环境因子及其阈值,制作主要天然林类型的潜在分布适宜性等级图,为东北林区天然林修复中的树种选择和结构调整提供依据。【方法】基于吉林省第八次森林资源连续清查固定样地数据,采用最大熵物种分布模型,划分11个主要天然林类型(含7类混交林),确定其现实分布点,选取影响林分生长的19个气候因子、33个土壤因子和3个地形因子共55个环境变量。对各天然林类型经变量筛选后进行分布建模,通过受试者工作特征曲线下的面积(AUC)评价模型精度。采用刀切法对模型进行检验,计算各环境变量对天然林类型分布的影响程度,筛选影响天然林类型分布的主导环境因子;将分布适宜性划分为5个等级,并制作不同天然林类型的分布适宜性等级图。【结果】基于最大熵模型的吉林省11个主要天然林类型适宜性分布的训练集和检验集的AUC在0.687 2~0.946 9之间,检验集的精度1个达到"极准确"、7个达到"很准确"、2个达到"较准确"、1个达到"一般"的水平,模型具有很好的泛化能力。各环境变量对天然林类型分布的影响程度由大到小依次为最热月的最高温度、海拔、最热季度的平均温度和年平均温度,最热月的最高温度为其中10个类型的主导环境因子,海拔为其中8个类型的主导环境因子,最热季度的平均温度为7个类型的主导变量因子,年平均温度为其中6个类型的主导环境因子,气温和海拔对预测的天然林类型分布有重要影响。基于建立的模型,形成吉林省11个主要天然林类型的潜在分布适宜性等级图。【结论】最大熵模型能够很好预测吉林省主要天然林类型的分布适宜区,筛选的影响每个天然林类型分布的主导环境因子及阈值合理,形成的吉林天然林类型潜在分布适宜性等级图可为东北林区天然林修复中的树种选择和结构调整提供科学依据。
【Objective】This study aims to investigate the method of distribution suitability of natural forests based on species distribution models,to examine the dominant environmental factors affecting the distribution of forest types and to generate the distribution suitability maps of major forest types. The results will provide the reference for natural forest restoration and structural adjustment in the region.【Method】Maximum entropy models were used for predicting potential distribution suitability for 11 natural forest types（including 7 mixed-species forests） in Jilin Province,northeast China.Data with tree species presence were obtained from permanent sample plots from the 8 th Chinese national forest inventory（NFI） with natural origins collected in Jilin Province. Totally,19 bioclimatic,33 soil and 3 terrain environmental variables were included. Model accuracy was evaluated by AUC（area under the receiver operating characteristics curves）values,and the Jackknife test showed the importance of different variables which determined dominant variables affecting the distribution of forest types. Distribution suitability maps of 11 forest types were generated with five levels.【Result】Themaximum entropy models were successful at discriminating between suitable and unsuitable habitat at the local scale for all11 forest types,the AUC values were from 0. 687 2 to 0. 946 9 for calibration and test data. Among the values,one forest type showed ＂excellent＂,7 forest types showed ＂very good＂,2 forest types showed ＂good ＂and 1 forest type showed＂fair＂. Therefore,the prediction results of the potential distribution of the 11 forest types in Jilin Province by the maximum entropy model were reliable. The environmental factors affecting the distribution of forest types were ranked as max temperature of the warmest month（for 10 forest types）,elevation（for 8 forest types）,mean temperature of the warmest quarter（for 7 forest types）,and annual mean temperature（for 6 forest types） in terms of their importance to the distribution of the specific forest types. Therefore,temperature and elevation were the most important factors to the distribution of the specific forest types in the region. Potential distribution suitability mapping for 11 forest types were completed. 【Conclusion】The maximum entropy model could reliably simulate the potential distribution area of forest types in Jilin Province. The dominant environment variables selected by the model were reasonable. The distribution suitability maps for 11 forest types could be as a reference for natural forest restoration and quality improvement in northeast China.