KMS XINJIANG INSTITUTE OF ECOLOGY AND GEOGRAPHY，CAS
| 全球陆地生态系统总初级生产力时空变化格局 | |
| Alternative Title | Spation-temporal patterns of gross primary production of global terrestrial ecosystems |
| 董嘉琪 | |
| Subtype | 博士 |
| Thesis Advisor | 李龙辉 |
| 2020-06-30 | |
| Degree Grantor | 中国科学院大学 |
| Place of Conferral | 北京 |
| Degree Discipline | 理学博士 |
| Keyword | 总初级生产力 温度-绿度模型 增强型植被指数 陆地表面温度 植被功能类型 |
| Abstract | 地球表层系统与其周围的地球圈层存在物质和能量的交换，是一个巨大的光合合成和消耗分解的系统。陆地生态系统碳循环和水循环与大气圈、地圈和生物圈之间相互作用，在地球气候变化和碳循环中起着重要的作用，也为人类提供了物质和能量基础。植被是陆地生态系统中的重要组成部分，绿色植物通过光合作用，将大气中CO2转变成有机碳，为自然界动物和生物等提供生存所需的物质条件。总初级生产力（Gross primary productivity, GPP）又称为总第一性生产力，它是指在单位时间内生物，主要是绿色植物通过光合作用的途径所固定的有机碳量。在生态系统水平上，总初级生产力是全球最大的碳通量，是人工和水域等生态系统功能的驱动力。基于此，准确地把握陆地生物圈总初级生产力在时间和空间尺度上的变化特征，对理解全球气候变化与生态系统变化之间的相互作用及碳循环在全球陆地生态系统演变中的作用研究具有重要的科学意义。 过去几十年，通过地面、大气和空间观测，在理解和量化陆地总初级生产力的时空格局方面取得了重大的进展。然而，现阶段基于直接观测手段还无法获得GPP，已有的多种GPP数据产品都是基于不同的模型估算得出，由于缺乏真实的观测量作为参考，不同数据产品之间存在较大的差异性。本论文目的是通过基于遥感的方法，即“温度-绿度”模型（Temperature-Greeness，TG），应用全局敏感性分析（Morris）和蒙特卡洛-马尔科夫链（MCMC）方法对TG模型的参数进行敏感性检测和参数优化，实现对全球陆地生态系统GPP的估算。在此基础上，分析TG模型模拟的全球陆地生态系统近20年（2000-2018）平均GPP的空间分布、变化趋势以及年际变化（IAV），并与其他三种全球GPP数据产品相比较，得到如下研究结论：（1）从Morris参数敏感性筛选方法中，筛选出TG模型在不同植被类型中的敏感参数。Xn（最低温度）和Xo（最适温度）是落叶阔叶林、落叶针叶林、常绿阔叶林、常绿针叶林、混交林和草地植被类型中的主要影响参数；Xo（最适温度）则为郁闭灌丛、作物、湿地、木质稀疏草原、稀疏草原和开阔灌丛的影响参数；然而Xm（最高温度）未被识别为任何植被功能类型中的主要影响参数。应用MCMC方法对TG模型敏感参数进行优化结果表明，对Xn敏感的植被类型的最优参数值与其初始值（0 ℃）有显著的差异，其范围从常绿针叶林的-10℃到湿地环境中5 ℃。参数Xo的优化值在不同植被类型中也表现出差异性，特别是在常绿阔叶林中Xo的最优值为15 ℃，是TG模型默认值（30 ℃）的一半。在其余植被类型中，Xo的最优值范围在27.5 ℃ - 32.5 ℃区间内；将MCMC参数优化后应用于TG模型，与全球155个通量站点处理得到的GPP数据进行验证。结果表明，参数优化提高了TG模型在超过一半的植被类型的GPP估算性能，尤其显著改善了常绿阔叶林和常绿针叶林中的GPP估算精度，也提高了稀树草原植被类型的GPP估算性能。 （2）应用优化前后的TG模型估算了全球近20年（2000-2018）陆地生态系统GPP，生成了两套全球GPP数据产品（TGdef GPP和TGopt GPP）。分析了近20年GPP时空变化格局。研究结果表明，TGdef和TGop估算的全球陆地生态系统多年平均GPP在空间分布中呈现一定的差异，主要表现在热带赤道亚马逊森林附近，TGopt显著高于TGdef的GPP值，相差约为500 gC year-1；但估算的近20年GPP变化趋势较为一致。在北美加拿大东北区域、亚马逊森林南部区域，印度的北方和中国东南部区域GPP呈现显著增长的趋势，在巴西东北部区域GPP呈现显著的下降趋势。 （3）近20年全球陆地生态系统GPP呈显著增加的趋势。GPP的增加主要是增强型植被指数（Enhanced Vegetation Index，EVI）的增加而引起的，而陆地表面温度（Land Surface Temperature，LST）对GPP的变化趋势影响有限。有超过9种植被类型的EVI与GPP偏相关系数高于0.5，其中在郁闭灌丛植被中相关性最好为0.8，在草原和木质稀树草原相关性也较好。LST表现与GPP的相关性不高，12种植被类型偏相关系数均低于0.5。 （4）基于四种基于遥感模型（TGopt、VPM、MODIS和PML）估算的全球GPP的时空格局的分析发现，在植被功能类型和柯本气候分区数据的辅助下，分析了四种GPP数据产品之间的差异。结果表明，四种数据产品的年总GPP估算范围较大，在126至164 Pg C year-1之间。GPP年均值的纬度梯度遵循生物群落的总体分布，但是在北纬热带区域和非洲的高纬度热带地区，四种数据产品之间年均值具有高达500 g C year-1的差异，MODIS GPP产品给出的年均值相对较低。其中，TGopt与PML呈现出较强的线性相关性。在常绿阔叶林和落叶阔叶林植被类型中，四种数据产品均呈现出较高的年均GPP，但是在开放灌木丛植被类型中，年均GPP最低。从气候分区在赤道带气候分区中VPM和MODIS整体比TGopt和PML GPP低，与其在赤道带和干旱带的表现较为一致。从全球陆地生态系统年GPP变化趋势来看，MODIS GPP近20年变化趋势不明显。从植被功能类型来看，四种GPP数据产品在郁蔽灌丛和开阔灌木丛中呈现出较好的一致性，在稀树草原植被类型中产品之间数值差异性较大。从柯本气候分区来看，温暖带气候和亚寒带年GPP呈现显著增加趋势（4.2-4.4 gC year−1），而在赤道带地区，GPP变化趋势呈现降低现象（-0.4 gC year−1），且四种数据产品之间差异较大。相比变化趋势，四种数据GPP年际变化（IAV）之间呈现较好的一致性，尤其是TGopt与VPM之间（R2=0.64）。总体上，四种数据产品中除了MODIS GPP数据产品，其余三种数据产品较好地体现出了干旱区GPP IAV较大，而湿润区GPP IAV较低的特点。柯本气候分区类型中，干旱带GPP IAV也显著高于亚寒带。关键词：总初级生产力，温度-绿度模型，增强型植被指数，陆地表面温度，植被功能类型，柯本气候分区 |
| Other Abstract | There is an exchange of matter and energy between the Earth's surface system andits surrounding Earth's circle. It is a huge system of photosynthesis and consumption.The interaction between the carbon cycle and water cycle of the terrestrial ecosystemand the atmosphere, geosphere and biosphere plays an important role in the Earth'sclimate change and carbon cycle, and also provides a material and energy basis forhuman activities. Vegetation is an important part of the terrestrial ecosystem. Greenplants convert CO2 in the atmosphere into organic carbon through photosynthesis, andprovide material conditions required for the survival of animals and organisms in ourenvironment. Gross primary productivity (GPP) is also called total primary productivity.It refers to the amount of organic carbon fixed by organisms, mainly green plants,through photosynthesis in a unit time. At the ecosystem level, gross primaryproductivity is the world's largest carbon flux and is the driving force for ecosystemfunctions such as artificial and water areas. Based on this, it is possible to accuratelygrasp the changes in the temporal and spatial scale of the total primary productivity ofthe terrestrial circle, which is important for us to understand the interaction betweenglobal climate change and ecosystem changes and the change of the carbon cycle in theglobal terrestrial ecosystem.Over the past few decades, significant progress has been made in quantifying andunderstanding the temporal and spatial pattern of total terrestrial primary productivitythrough surface, atmospheric, and space observations. However, at this stage, there isno GPP data obtained based on direct observation. Many data sets are estimated basedon different types of models. Due to the lack of real observations, the data set has a lotof uncertainty. The purpose of this paper is to optimize the parameters of the TG modelby using the remote sensing method, that is, the "Temperature-Greeness" model (TG),and the Monte Carlo-Markov chain (MCMC) method to estimate global GPP interrestrial ecosystem; The spatial distribution, trends and interannual variability of theglobal terrestrial ecosystem simulated by the TG model for nearly 20 years (2000-2018)are analyzed. And compared with the other three global GPP data sets, the followingresearch conclusions are obtained:(1) From the Morris parameter sensitivity screening method, the sensitiveparameters of the TG model in 12 plant function types were screened. In TG model, the parameters Xn (minimum temperature) and Xo (optimum temperature) were sensitiveparameters in deciduous broadleaved forest, deciduous needleleaved forest, evergreenbroadleaved forest, evergreen needleleaved forest, mixed forest and grasslandvegetation types; Xo (optimum temperature) is sensitive to shrubland, crops, wetlands,woody savanna, savanna and open shrubland; however, Xm (maximum temperature)was not identified as the main influence parameter in any plant function type. Afterselecting sensitive parameters, the MCMC method was used to optimize the sensitiveparameters of the TG model. The parameter optimization results show that the optimalparameter values Xn of the plant function types are significantly different from theinitial value Xn (0 ℃), ranging from -10 ℃ in evergreen needleleaved forest to 5 ℃ inwetland. The optimized value of the parameter Xo also shows differences amongdifferent plant function types, especially in the evergreen broadleaved forest, theoptimal value of Xo is 15 ℃, which is half of the default value of the TG model (30 ℃).Among the remaining plant function types, the optimal value of Xo is in the range of27.5 ℃-32.5 ℃; the MCMC parameters are optimized and applied to the TG model,and verified with GPP by 155 flux sites worldwide. The results show that parameteroptimization improves the performance of the TG model in more than half of the plantfunction types, especially significantly improves the estimation performance of themodel in evergreen broadleaved forest, evergreen needleleaved forest and savanna. Theanalysis shows that the vegetation index EVI in the TG model is more sensitive toecosystems with limited water resources.(2) The TG model with parameter optimization is used to estimate the globalterrestrial ecosystem GPP for nearly 20 years (2000-2018), and the global GPP data set(TGopt GPP) based on the optimized TG model and the initial TG model are obtainedglobal GPP data set (TGdef GPP). Analyze the effects of GPP's spatiotemporal changesin the past 20 years, as well as the yearly changes in the temperature indicator LST andthe greenness index EVI. The research results show that the global terrestrial ecosystemestimated based on the initial and optimized TG models has a large difference in spatialdistribution over many years, especially near the tropical equatorial Amazon forest.TGopt is significantly higher than the GPP value of TGdef. The difference is about 500gC year-1; The trends of the two data sets are consistent in the past 20 years. In NortheastCanada, North America, South of the Amazon Forest, North India and Southeast China,GPP shows a significant growth trend, and in Northeast Brazil, GPP shows a significantdownward trend; TGopt changes in the interannual changes of Australia and southwestern South Africa in the southern hemisphere. The magnitude is large, showinga downward trend in the Amazon forest in South America. In southeastern China, itmay be due to the initial success of policies such as returning farmland to forests andafforestation. The GPP has changed significantly.(3) The global terrestrial ecosystem GPP has been increasing in the past 20 years.The increase of GPP is mainly caused by the increase of Enhanced Vegetation Index(EVI), and the land surface temperature (LST) has a limited impact on the changingtrend of GPP. There are more than 9 types of EVI and GPP partial correlationcoefficients higher than 0.5. Among them, the best correlation is 0.8 in canopyshrubalnd vegetation, and the correlation is better in grassland and woody savanna. Thecorrelation between LST performance and GPP is not high, and the partial correlationcoefficients of 12 types of plant cover are all lower than 0.5.(4) Spatio-temporal pattern analysis of four types of global GPP estimated basedon remote sensing ecological models (TGopt, VPM, MODIS, and PML), with theassistance of vegetation function types and Köppen climate zone data, analyze thedifferences between data sets. The results show that the annual total GPP estimates ofthe four data sets have a wide range, ranging from 126 to 164 PgC year-1. The GPPannual mean latitude gradient follows the overall distribution of biological communities,but in the tropical north latitudes and the high latitudes in Africa, there is a differenceof 500 gC year-1 between the four data sets. The year given by MODIS GPP productsthe mean is relatively low. Among them, TGopt and PML show a strong linearcorrelation. Among the evergreen broadleaved fores and deciduous broadleaved forestvegetation types, the four datasets all showed higher average annual GPP levels, butamong the open shrubland vegetation types, the average annual GPP was the lowest.From the climate zone in the dry zone climate zone, the overall VPM and MODISunderestimate the GPP compared to TGopt and PML. This result is consistent with theperformance in the tropical zone and dry zone. From the perspective of the globalterrestrial ecosystem year GPP trend, the MODIS GPP trend in the past 20 years is notobvious. From the perspective of plant function types, the four GPP data sets showedgood consistency in the canopy shrubs and open shrubland, and the values GPP of theproducts in the savanna were significantly different. From the perspective of theKöppen climate zone, the temperate climate and cold temperate zone GPP show anincreasing trend (4.2-4.4 gC year−1), while in the tropics the GPP change trend shows adecreasing phenomenon (-0.4 gC year−1). There are large differences among these data sets. Compared with the interannual variability, the four types of data GPP IAV showgood consistency, especially between TGopt and VPM (R2 = 0.64). In general, inaddition to the MODIS GPP data sets, the remaining three data sets better reflect thecharacteristics of larger GPP IAV in the arid zone and lower GPP IAV in the wet zone.In the Koppen climate zoning type, the dry zone GPP IAV is also significantly higherthan the wet zone. |
| Subject Area | 地图学与地理信息系统 |
| Language | 中文 |
| Document Type | 学位论文 |
| Identifier | http://ir.xjlas.org/handle/365004/15415 |
| Collection | 中国科学院新疆生态与地理研究所 研究系统 |
| Affiliation | 中国科学院新疆生态与地理研究所 |
| First Author Affilication | 中国科学院新疆生态与地理研究所 |
| Recommended Citation GB/T 7714 | 董嘉琪. 全球陆地生态系统总初级生产力时空变化格局[D]. 北京. 中国科学院大学,2020. |
| Files in This Item: | There are no files associated with this item. | |||||
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment