卫星遥感叶面积指数在陆面模型中的同化及应用

发布时间：2018-04-05 03:32

  本文选题：叶面积指数　切入点：陆面模型　出处：《南京大学》2016年博士论文

【摘要】：作为表征地表植被特性的一个重要参数,叶面积指数(LAI),是影响地表辐射传输、物质能量平衡的重要植被生物物理学参数,同时也是连接植被光合作用、呼吸作用等微观生物地球化学过程的重要参数。由于陆地下表面强烈的不均匀性,正确描述地表植被的冠层结构,以及物理、生物地球化学过程,是正确描述地表辐射特征、以及陆-气相互作用的基础和前提。直接观测是目前认为可以获得LAI最精确的方法,但是短期内还无法实现LAI在区域乃至全球范围内的统一台站观测;随着生物地球化学模块的应用和发展,越来越多的陆面模式能模拟出LAI等植被参数的动态演变,但模拟结果严重依赖于陆面过程模式的模块和参数化过程,以及模拟的初始条件等因素。陆面数据同化,利用滤波/变分方法将模型和观测融合在一起,同时结合陆面过程的动力框架进行约束,达到对目的状态量的最优估计,并实现资料的时空尺度扩展。现存的很多卫星遥感数据都可以提供时空连续的区域/全球分布的LAI,并且经过了严格的质量评估和校验,这为数据同化改善陆面模型中生物地球化学模块的模拟能力提供了数据基础。本文拟利用美国大气研究中心(NCAR)开发的公用陆面模式(CLM4)和数据同化研究平台(DART),将卫星遥感观测的全球LAI数据(GLASS LAI)同化到CLM4的碳-氮循环模块(CLM4-CN)中,并在同化的基础上进一步分析LAI的改变在全球尺度和区域尺度上对地表能量、水分平衡、植被-大气相互作用以及气候变化产生的影响。鉴于模型的初始条件分布为集合同化提供了初始误差,其发散程度对同化的进行和同化结果都会产生很大的影响,本论文首先对模型进行了单个大气数据和集合大气数据分别驱动CLM4-CN的模型初始化过程,以获得相应的初始条件集合。结果表明,在热带低纬度的森林覆盖地区、及草地、农田下垫面,集合模式足够发散,而在北方常绿针叶林和温带落叶阔叶灌木下垫面,集合模式的离散度则相对偏小：另外,模式模拟LAI的发散程度在植被的生长季明显优于非生长季。LAI值波动范围越大的季节或地区,初始条件集合的发散程度越高,越有利于同化过程的进行。为了找到有效的同化方案,本文设计了三组对比实验,分别为：①没有进行同化的控制实验(CTL实验)、②在同化的过程中不进行碳-氮约束的NO-CN实验,和③在同化过程中进行碳-氮约束的C-N实验。结果显示,未进行同化时,CLM4-CN模拟的LAI系统性高估了全球的LAI分布,且在低纬度地区尤其明显,最大偏差甚至超过5 m2/m2。在同化过程中没有进行碳-氮约束时,同化效果并不明显;而在启动了碳-氮循环模块之后,同化的LAI值与观测的偏差明显降低,且在低纬度的偏差能够控制在±1 m2/m2。由此可见植被动态物理约束过程(C-N循环)的加入对同化过程具有很好的约束和修正作用。作为数据同化研究平台,DART可以同时提供各种集合同化算法为研究所用,因此本文分别对集合调整卡尔曼滤波(EAKF)、集合卡尔曼滤波(EnKF)、卡尔曼滤波(KF)和粒子滤波(PF)这四种同化算法进行了对比分析,结果表明,集合同化(EAKF、EnKF)的结果优于单个变量数目的同化算法(KF);另外,由于模型模拟的LAI系统偏高于观测,PF法在迭代的过程中逐步减小了LAI观测的权重,也就是降低了观测值对同化过程中后验概率的计算贡献。由于EAKF在每一步都对增益矩阵的更新进行了调整,使其在不低估分析误差协方差的前提下对观测场产生尽量小的扰动,所以,确定在用EAKF方法同化的过程中,同时考虑C-N模块的约束,才是最优的同化方案。在挑选出最优同化方案方法的基础上,本文对模型输出LAI的能力进行了评估,结果得到：同化不仅能够很好地模拟出LAI的空间分布,也能够改进不同纬度区域平均LAI的年变化(23°S以南除外)。另外,同化后LAI改进最明显的典型植被覆盖区罗列如下：同化后LAI显著减小的区域分别为非洲中部、亚马逊东部、欧亚大陆南部、中国东北、和欧洲西部,主要覆盖下垫面为常绿/落叶阔叶林、混合森林;LAI显著增加的区域分别为欧亚中部、北美西部、和澳大利亚西部地区,主要覆盖下垫面是开放式灌木丛和草地。在挑选出最优同化方案的基础上,本文首先利用单向耦合的方法分析了2002年北半球夏季LAI对地表状态量、陆-气相互交换通量的影响。结果得到,在亚马逊中部、北美北部地区、非洲北部及中部、以及欧亚大陆大部分地区,模型均低估了地表2m温度;而在北美中部及南部,欧亚大陆东部及东南部地区,以及澳大利亚大部分地区,模型则严重高估地表2m温度。就改进效果而言,在北美西部地区、亚马逊大部分区域、非洲中部地区、欧亚大陆的中东部及南部,改进后的LAI使得这种低估的形式有所缓解;而对于非洲南部,亚马逊中部地区、以及欧亚大陆的中西部地区,改进后的LAI缓解了高估地表温度的情况。总体而言,同化后的LAI对低估地表气温的改进效果明显好于对高估地表气温的改进效果,这大概与叶面积指数的全球效应有关。具体而言,LAI在全球范围内的减小,会导致地表2m气温增高,且在欧亚大陆西部地区达到最高(1.6℃),其次是亚马逊东部和非洲中部,偏差分别为1.1℃和1.0℃;区域LAI的升高会造成当地2m气温的降低,其中影响最大的地区为美国西部,偏差达-0.5℃。除了亚马逊东部地区,LAI减小的区域,表层土壤湿度增大,反之亦成立;这可能是由于亚马逊东部地区的土壤湿度受地表径流等因素的影响更大。值得说明的是,在低纬等LAI改变最大的区域,并不是地表状态量和陆-气通量改变最明显的区域,这主要与该区域的植被覆盖类型和气候平均态有关。利用LAI同化结果,本文还分别对比分析了在陆-气耦合、海-气-陆-冰耦合情况下LAI改变对地表状态量、物质能量平衡、以及气象条件的影响,以分析植被变化对天气和气候过程的反馈效应。结果表明,在耦合了大气和陆面模式的情况下,同化后的LAI不仅改善了模型对低纬度2m气温的模拟能力,还能改善对高纬度2m气温的模拟能力,模拟偏差从-5℃-4℃减小到-3℃-2℃。耦合了大气的陆面模型同时能够改善在非洲中部、东南亚群岛、亚马逊北部、阿拉伯半岛和澳大利亚西部地区对降水的模拟能力。另外,加入了海-冰耦合的模型,在热带地区LAI改变的地表温度、降水等特征量变化的影响有所削减,却加强了中高纬度地区LAI改变对区域气候的影响。由于本文积分时间较短,对海气耦合以及海气陆冰耦合结论的分析还有待深入;另外,对LAI导致的气候变化的机理解释还有待进一步验证。
[Abstract]:As an important parameter for characterizing the surface vegetation characteristics, leaf area index (LAI), is the effect of surface radiative transfer, important vegetation biophysical parameters material and energy balance, but also the connection of vegetation photosynthesis, respiration and other important parameters of micro biogeochemical processes. Due to the nonuniformity of land surface strong, correct description of the canopy the structure, surface vegetation and physical and biogeochemical processes, is the correct description of surface radiation characteristics, and land atmosphere interaction is the basis and premise. Direct observation is that LAI method can get the most accurate, but the short term can not achieve LAI in regional and global unification Observatory; with the application and the development of biogeochemical module, land surface model and more can simulate the dynamic evolution of LAI vegetation parameters, but the simulation results rely heavily on Module and parameterization in the land surface model, and the simulation of the initial conditions and other factors. Land data assimilation, using filter / variational method to model and observation together, combined with the dynamic framework of the land surface process constraints, achieve optimal estimation of objective state quantity, and realize the spatial and temporal scales data expansion. Many existing satellite remote sensing data can provide continuous spatio-temporal regional / global distribution of LAI, and after a strict quality assessment and verification, which provide the data basis for data assimilation to improve the simulation ability of biogeochemical module land surface model. This paper intends to use the National Center for Atmospheric Research (NCAR) the development of public land surface model (CLM4) and data assimilation research platform (DART), the satellite observations of global LAI (GLASS LAI) data assimilation to carbon nitrogen cycle CLM4 module (CLM4-CN), and in the same On the basis of further analysis of the change of LAI in the global and regional scale on surface energy, water balance, influence vegetation atmosphere interaction and climate change. In view of the distribution of the initial conditions of the model provides the initial error for collection and assimilation, assimilation results on the assimilation of the divergence degree will have a great effect of model initialization process firstly carried out on the model of single atmospheric data and atmospheric data set respectively to drive CLM4-CN, to obtain the corresponding set of initial conditions. The results show that the coverage area in the low latitude tropical forest and grassland, farmland surface, collection mode enough divergence, while in the north temperate evergreen needle Ye Linhe deciduous shrub surface, discrete degree set model is relatively small. In addition, the degree of divergence in the LAI model to simulate the growth of vegetation in non Ji Mingxianyou Long season.LAI value fluctuation range more seasons or regions, the higher the level of divergence of the initial conditions set, more conducive to the assimilation process. In order to find the effective assimilation scheme, this paper designs three groups of experiments, respectively: no control experiment of assimilation (CTL experiment), NO-CN experiment II for carbon and nitrogen assimilation in the constraints in the process of C-N experiment of carbon nitrogen and the constraints in the assimilation process. The results showed that, without assimilation, CLM4-CN simulation system of LAI overestimates LAI global distribution, and is especially evident in the low latitude region, the maximum deviation even more than 5 m2/m2. carbon - nitrogen constraint in the process of assimilation, assimilation effect is not obvious; and after the start of the cycle of carbon and nitrogen assimilation module, the LAI value and the observed deviation is significantly reduced, and the deviation in the low latitude can be controlled within 1 m2/m2. in vegetation The dynamic process of physical constraints (C-N cycle) with good correction effect on the constraint and assimilation process. As the data assimilation research platform, DART can also provide a variety of ensemble algorithm used for research, so this paper set adjustment (EAKF), a collection of Calman filter Calman filter (EnKF), Calman filter (KF) and particle filter (PF) of the four assimilation algorithms are compared, the results show that the set of contracts (EAKF, EnKF) assimilation algorithm results is better than that of single variable number (KF); in addition, because the model simulation of LAI system is higher than observation, PF in the iterative process of gradually reduced the weight of LAI observations, is reduced in the process of assimilation of observations to calculate the posterior probability contribution. Because EAKF has been adjusted in every step of the gain matrix is updated, so that it does not underestimate the premise of the analysis of the error covariance The observation field as far as possible small disturbance, so determined by EAKF method in the process of assimilation, taking into account the C-N module constraints, is optimal. In the selection of assimilation scheme based on the method of optimal assimilation scheme, the ability of the model output of LAI was evaluated, the results can not only obtain: assimilation good simulation of the spatial distribution of LAI, also can change in different latitude areas improved average LAI (23 degrees south of S except). In addition, LAI after assimilating the most obvious improvement of the typical vegetation coverage area are as follows: the eastern region decreased significantly after LAI assimilation were central Africa, south of the Amazon, the Eurasian Chinese northeast, and Western Europe, mainly covering the surface for the evergreen / deciduous broad-leaved forest, mixed forest; LAI increased significantly in the region were central Eurasia in western North America, Australia, and the western region, covering the main surface is open Put the bushes and grass. In the selected optimal assimilation scheme based on the one-way coupling method of 2002 in the northern hemisphere summer LAI on surface state, land surface fluxes. Results show that in the Amazon region of northern North America, central, northern and Central Africa, and most of the Eurasia area model underestimates the surface temperature of 2m; while in the north central and southern, Eastern and southeastern regions of Eurasia, and Australia in most areas, the model overestimated 2m temperature. The improved surface effect, in the areas of western North America, most of the Amazon region, Central Africa, the Middle East and south part of Eurasia part, the improved LAI makes this underestimation form eased; and for South Africa, the central region of Eurasia, Amazon, and the central and western regions, the improved LAI mitigation The surface temperature of the overestimation. In general, the improvement effect after assimilation of LAI to underestimate the surface temperature is significantly better than the improvement effect on the overestimation of surface temperature and this is probably related with leaf area index of the global effect. Specifically, the decrease of LAI in the global scope, will lead to surface 2m temperature increased, and in the area the Western Eurasian continent reached the highest (1.6 DEG C), followed by Amazon in eastern and Central Africa, deviations were 1.1 degrees and 1 degrees; the increase of LAI area will decrease the local temperature 2m, one of the most influential area for the western United States, deviation of -0.5 degrees. In addition to the Eastern Amazon, LAI reduced surface area the soil moisture increases, and vice versa; this may be due to the soil moisture of the Eastern Amazon is affected by surface runoff factors. It is worth, the biggest change in the low latitude region such as LAI, and not The surface state and land air fluxes change the most obvious area, this mainly in the region and the type of vegetation cover and climate mean state. Using LAI assimilation results, this paper analyzed the land atmosphere coupling, air sea land ice under the condition of coupling LAI change of surface state, material the energy balance, and the influence of meteorological conditions, to analyze the feedback effects of vegetation change on weather and climate processes. The results show that, in the coupling of the atmosphere and land surface model under the condition of assimilation after LAI can improve not only the ability to simulate the low latitude 2m temperature model, also can improve the simulation ability of high latitude temperature 2m the simulation error is reduced from -5 to -3 DEG -4 DEG -2 DEG C. Coupled with land surface model and can improve the atmosphere in Central Africa, Southeast Asian islands, north of the Amazon, the simulation of precipitation in Arabia Peninsula and Western Australia. The other , joined the sea ice coupling model, the surface temperature change of LAI in the tropics, changes in precipitation characteristic quantity has reduced, but increased in the high latitudes of LAI influence on regional climate. The integral time is short, and the analysis of the coupled air sea ice coupling and conclusion further; in addition, explain the mechanism of climate change caused by LAI remains to be further verified.

【学位授予单位】：南京大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：Q948
，

本文编号：1713026