基于BEPS模型的塞罕坝植被净初级生产力时空变化分析研究

Analysis of Spatial and Temporal Variation of Vegetation Net Primary Productivity in Saihanba Based on BEPS Model

叶面积指数(leaf area index,LAI)是北部生态系统生产力模拟模型(boreal ecosystem productivity simulator,BEPS)的关键驱动数据,获取高精度LAI对区域森林生态系统碳循环十分重要,然而当前大多研究采用的MODIS LAI产品缺乏可信度。为此,基于LAI动态模型、PROSAIL辐射传输模型和层状贝叶斯网络(Hierarchical Bayesian Network,HBN)构建数据同化系统,获得空间分辨率为20 m的LAI数据,驱动BEPS模型,模拟塞罕坝机械林场2011—2021年的植被净初级生产力(Net Primary Productivity,NPP),并对NPP时空变化特征及影响因子进行分析。结果表明,基于贝叶斯同化方法获得的高分辨率LAI数据极大提高了MODIS LAI产品的精度;基于同化后的LAI数据驱动BEPS模型获取模拟森林NPP,与样地实测数据计算NPP间相关性较高(R^(2)=0.77);2011—2021年塞罕坝机械林场植被NPP平均值为307.4 g/(m^(2)·a),森林NPP呈现平稳增长趋势;不同植被类型模拟NPP存在较大差异,针叶林、落叶林及混交林模拟NPP分别为484.9、402.4、287.9 g/(m^(2)·a);植被NPP与温度因子相关性较高,偏相关系数为0.2~0.8,而植被NPP与降水量的相关性总体而言相对较低,其偏相关系数为-0.3~0.4,在该地区降水量对植被NPP的影响较低,温度为该地区NPP变化的主导因子。研究结果可获取高空间分辨率的LAI数据,为森林生态系统碳循环的精准时空模拟提供依据。

Leaf area index(LAI) is the key driving data of BEPS model,and it is important to obtain high accuracy LAI for regional forest ecosystem carbon cycle,however,the MODIS LAI products used in most current studies lack credibility.To this end,this study constructed a data assimilation system based on LAI dynamic model,PROSAIL radiative transfer model and Hierarchical Bayesian Network(HBN) to obtain LAI data with a spatial resolution of 20 m to drive the BEPS model and simulate the vegetation net primary productivity(NPP) of Saihanba Mechanical Forest during 2011—2021,and the spatial and temporal variation of NPP and the influencing factors of NPP were analyzed.The results showed that the high-resolution LAI data obtained based on Bayesian assimilation method greatly improved the accuracy of MODIS LAI products;the correlation between the simulated forest NPP obtained from BEPS model driven by assimilated LAI data and the NPP calculated from the sample plots was high(R^(2)=0.77);the mean value of vegetation NPP in Saihanba Mechanical Forest during 2011—2021 was 307.4 g/(m^(2)·a),and the NPP of forest showed a steady growth trend;the simulated NPP of different vegetation types were different,and the simulated NPP of coniferous,deciduous and mixed forests were 484.9,402.4,287.9 g/(m^(2)·a);the correlation between vegetation NPP and temperature factor was high,and the bias correlation coefficient was 0.2-0.8,while the correlation between vegetation NPP and precipitation was relatively low in general,with bias correlation coefficients of-0.3-0.4.The influence of precipitation on vegetation NPP was low in this region,and temperature was the dominant factor of NPP variation in this region.In this study,high spatial resolution LAI data were obtained to provide a basis for accurate spatial and temporal simulation of the carbon cycle in forest ecosystems.