耦合多源数据的森林碳水通量模型驱动参数同化机制

Huge uncertainties and coarse resolution of MODIS leaf area index (LAI)/normalized difference vegetation index (NDVI) and MERRA products have became the bottleneck to achieve stability and accuracy of forest carbon and water flux model.The best way to solve this problem is to couple the advantage of multi-source data and assimilate these driving variables, and then make these driving variables contain high quality and spatio-temporal resolution. This study focuses on coupling and assimilation mechanism for driving variables using the conjugate gradient method, ensemble Kalman filter, variational method based on data sets from sample plot survey, satellite remote sensing, aerial photography obtained by unmanned aerial vehicles, eddy flux tower and wireless sensing technology.The goals of this study were to develop:(1) photosynthetically active radiation (PAR) inversion and assimilation method based on wireless sensing data and the radiative transfer model;(2) the method for LAI estimation based on aerial photography and Beer-Lambertlaw; (3) LAI and NDVI inversion and assimilation method based on the combination of observed LAI, MODIS reflectance, Beer-Lambertlaw and PROSAIL coupled model;(4) a software for coupling and assimilation of driving variables.Finally, the effect of driving variables after assimilation on estimtes of forest carbon and water flux model will be evaluated.

MODIS LAI/NDVI和MERRA产品存在高不确定性和低空间分辨率等问题，成为大尺度森林碳水通量估算稳定性和准确性的瓶颈。如何耦合和同化多源数据，获取高质量、时空连续的驱动参数，是降低大尺度估算误差的有效途径。本项目将以样地调查、卫星遥感、无人机航拍、通量塔和无线传感等数据为基础，精度、时间和空间分辨率三要素融合为目的，采用共轭梯度法、集合卡尔曼滤波、变分法等优化算法，研究耦合多源数据的关键驱动参数同化机制，研发（1）同化无线传感数据和辐射传输模型的PAR反演方法；（2）耦合航片和Beer-Lambert定律的高精LAI转换方法；（3）基于实测LAI、MODIS反射率和Beer-Lambert定律+PROSAIL耦合模型的LAI/NDVI同化方法；（4）关键驱动参数耦合和同化应用软件。评价参数同化技术对提高竹林碳水通量反演精度的贡献，为降低大尺度碳水通量估算误差提供技术支撑。

由于现有的生态过程模型驱动参数，如叶面积指数（LAI）和光合有效辐射（PAR），存在空间分辨率低和不确定性大等问题，成为大尺度森林碳通量估算稳定性和准确性的瓶颈。如何获取高精度、高时空分辨率的驱动参数，成为当前森林碳通量反演研究的重要科学问题。本研究以开发多源数据的耦合和同化技术为研究目标，开展了基于双集合卡尔曼滤波的LAI同化研究、基于辐射传输模型的PAR遥感反演研究，评价了多源数据的耦合和同化技术在森林碳通量估算精度提升上的贡献。研究得到以下重要结果：采用双集合卡尔曼滤波同化技术显著地提高了LAI产品的精度，相比MODIS LAI产品，误差降低了60%以上，解决了MODIS LAI产品精度低的问题，实现了大尺度LAI产品高精度反演；基于辐射传输模型反演的PAR与实测值决定系数达到0.76，精度比MERRA PAR产品提高了32%，空间分辨率从0.5°×0.67°提高到1000 m，云量是制约PAR遥感反演精度的主要因子；以同化后的LAI和PAR产品驱动光能利用率模型和BEPS生态过程模型显著提高了森林碳通量估算精度，同化后的PAR比MERRA PAR产品得到的碳通量精度提高了13%，同化后的LAI比MODIS LAI产品得到的碳通量误差降低了30%以上。本研究开发的多源数据的耦合和同化技术显著地提高了生态过程模型驱动参数的质量，为精准监测大尺度森林碳通量提供强有力的技术支撑，在精准评价森林对减缓气候变化的贡献上具有重要科学意义。