集合卡尔曼滤波耦合辐射传输模型的亚热带典型森林MODIS LAI时间序列同化

Subtropical forest plays an important role in the global carbon cycle, and LAI is a key parameter in the study of carbon cycle in forest ecosystem. Obtaining LAI and its time series change by using remotely sensed technology has been the current research hotspot and the important direction in the future. However, many existing LAI products based on satellite remote sensing data have the shortcomings of high noise, low accuracy and large fluctuation in time series, which restrict the application of LAI products. In this research, we take three typical subtropical forest including Moso bamboo forest in Anji County, Lei bamboo and evergreen-deciduous broadleaf forests in Lin'an City of Zhejiang Province as the study object. The assimilation system is constructe by aggregating ensemble Kalman filter assimilation comprehensive ground measurements of leaf and canopy reflectance, physical, chemical and biological parameters, MODIS reflectance and LAI products data, coupled with PROSAIL model, a connecting bridge between LAI dynamic model and MODIS reflectance observed. On the basis of model and system optimization, MODIS LAI products of three typical forest observation stations will be assimilated, and LAI time series of high precision will be acquired. At last, the spatial and temporal distributions of LAI in the three species of forest were obtained by the assimilation system, which participate as parameters of the carbon cycle model for the simulation of the spatial and temporal patterns of carbon cycle. This results is expected to provide theoretical and practical reference to the function of subtropical forest carbon sequestration in our country and its response to global change.

亚热带森林在全球碳循环中具有重要作用，LAI是森林生态系统碳循环研究的关键参数，通过遥感技术获取LAI及其时间序列变化，一直是当前研究热点及未来重要方向，而现有诸多基于卫星遥感数据的LAI产品噪声大、精度较低且在时间序列上波动较大，从而制约了LAI产品的应用。本研究将以浙江省安吉县毛竹林、临安市雷竹林以及常绿落叶阔叶林等三种亚热带典型森林为对象，通过集合卡尔曼同化滤波综合地面测量叶片及冠层反射率、物理生物化参数、MODIS反射率和LAI产品等数据，耦合并以PROSAIL模型为连接LAI动态模型和MODIS 观测反射率之间桥梁，构建同化系统；在模型、系统优化的基础上，对三种典型森林观测站点MODIS LAI产品进行同化，获得高精度LAI时间序列；最后，应用同化系统获得三种森林LAI时空分布并参与碳循环时空格局模拟。研究结果将为评价我国亚热带森林碳汇功能及其对全球变化的响应提供理论和实践参考。

亚热带森林在全球碳循环中具有重要作用，LAI是森林生态系统碳循环模拟的关键参数，然而现有卫星遥感数据LAI产品噪声大、精度较低且在时间序列上波动较大，从而制了LAI产品的应用。本研究以浙江省安吉县毛竹林、临安市雷竹以及常绿叶落阔叶林等三种亚热带典型森林为对象，通过集合卡尔曼等同化滤波综合地面测量叶片及冠层反射率、物理生物化参数、MODIS反射率和LAI产品等数据，耦合并以PROSAIL模型为连接LAI动态模型和MODIS 观测反射率之间桥梁，构建LAI同化模型系统；在模型、系统优化的基础上，对三种典型森林MODIS LAI产品进行同化；最后，LAI同化结果驱动BEPS等碳循环模型，模拟亚热带碳循环时空分布格局。研究表明，（1）LAI、Cab和Cw对冠层反射率模拟具有较大的影响，在此基础上得到了一套PROSAIL模型最优模型参数，实现了毛竹林、雷竹林和常绿落叶阔叶混交林冠层反射率高精度模拟，精度分别为0.86、0.90、0.93；（2）PROSAIL耦合同化滤波所建立MODIS LAI同化系统，实现了亚热带典型森林LAI高精度时空同化，同化LAI与实测LAI之间的相关关系极为显著，相关系数R在0.9以上，其中粒子滤波LAI同化精度略高于集合卡尔曼滤波。同化得到的LAI时空分布能反映亚热带典型森林季节性生长规律；（3）建立的层状贝叶斯耦合PROSAIL辐射传输模型的LAI多尺度同化方法，实现了竹林LAI高精度多尺度时空同化，如毛竹林在1000 m、500m、250m三个尺度上同化LAI与实测LAI之间的决定系数R2分别为0.91、0.90和0.86。（4）同化LAI耦合随机森林模型，较好的估算了浙江省竹林AGB，比未同化LAI竹林AGB估算精度提高了57.8%，为基于低分辨率遥感数据的森林生物量估算提供了一种新的方法。（5）LAI驱动BEPS、InTEC等模型，实现了浙江省竹林等生态系统碳循环较高精度模拟，如基于粒子滤波的LAI同化产品模拟的竹林生态系统碳通量与未同化LAI结果相比，GPP和NEE模拟精度R2分别提高了27.5% 和45.2%，RMSE分别降低了29.9%和23.7%，另外，LAI同化结果在碳循环多尺度模拟等方面也进行了很好的探索。研究结果为评价我国亚热带森林碳汇功能及其对全球变化的响应提供了重要的理论和数据支撑。