【作者】 金云翔；

【导师】 徐斌；

【作者基本信息】 中国农业科学院 ， 农业遥感， 2012， 博士

【摘要】 草原生态系统是我国最大的陆地生态系统，草原生产力是维护草原生态系统的物质基础，是反映草原状况最直接的指标，对草原生态系统功能的强弱起着决定性的作用，草原植被生物量动态研究一直是陆地生态学的重点和热点问题。及时准确地了解草原产草量的时空分布状况，掌握草原年际间变化动态规律，对于草原可持续利用和管理具有重要意义。目前，无论从大尺度乃至全球尺度上，还是小到某区域或草地类型的估算草原生物量及碳密度，地面样方的建模和验证数据较少，模型的可靠性和估算精度难以保证，而对地下生物量的研究也相对缺乏，长期以来是生态学研究中的一个薄弱环节，尤其是不同利用方式和草原退化、沙化程度下，地下生物量与地上生物量的关系及动态变化和规律缺少深入系统的研究。本文以内蒙古正蓝旗为研究区，基于“3S”技术手段，以MODIS和TM为遥感数据源，通过地面样方数据与对应的气象要素、数字高程要素等空间数据库，构建草原生物量的遥感估算模型，阐明了2005-2010年草原生物量的时空动态分布，探讨了影响草原生物量时空分布的主要控制因素，揭示了草原地下/地上生物量的大小及其分配机制，并进一步估算了不同沙化程度草原生物量及碳贮量。主要的研究结论与进展如下：（1）利用地面样方生物量数据结合同期MODIS、气象要素、数字高程等数据，建立传统经验统计法和B-P神经网络法的生物量遥感估算模型，通过精度验证和对比分析，优选出B-P神经网络法模型为生物量遥感估算模型，模型精度达到79%。利用B-P神经网络模型估算了正蓝旗2005-2010年的草原地上生物量，多年平均干草量为798620吨，平均生物量密度为797.36kg/hm~2，空间上呈中部高两边低的分布格局，降水量是影响草原生物量时空分布的关键因素之一。（2）基于2010年正蓝旗TM遥感数据，参考草原沙化分级指标体系，采用混合像元分解法将正蓝旗草地分为未沙化草地、轻度沙化草地、中度沙化草地和重度沙化草地，分类精度达到80%。结果表明，正蓝旗草原沙化主要集中在北部，沙化面积约为4483.95km~2，占总草原面积的45.88%。（3）根据2010年野外调查的不同沙化程度下样方生物量资料得出，未沙化草地、轻度沙化草地、中度沙化草地、重度沙化草地的平均地上生物量密度分别为214.72g/m~2、135.41g/m~2、91.42g/m~2、24.01g/m~2，平均地下生物量密度分别为2601.35g/m~2、2318.45g/m~2、413.25g/m~2、117.25g/m~2，地下/地上生物量比值在4.5-17.2间。随着沙化程度的加剧，土壤质地粗化，土壤肥力下降，影响了植物体的生长发育，从而导致草原生物量大幅降低。（4）在2010年正蓝旗草原沙化分类结果和草原地上生物量遥感估算结果的基础上，结合地下/地上生物量比值信息，估算了正蓝旗2010年不同沙化程度下的草原生物量及碳贮量，草原生物量约为755万吨，总碳贮量为3.4TgC，碳密度在0.99MgC/hm~2-11.12MgC/hm~2范围内。其中未沙化草地平均碳密度为4.39MgC/hm~2，碳贮量为2.31TgC；轻度沙化草地碳密度为5.61MgC/hm~2，碳贮量为0.55TgC；中度沙化草地碳密度为1.58MgC/hm~2，碳贮量为0.33TgC；重度沙化草地碳密度为1.53MgC/hm~2，碳贮量为0.22TgC。由此看出，草原沙化对草原碳固定量具有明显的负作用，草原沙化的加剧导致草原固碳能力的下降。（5）为了完成上述研究目标和内容，选用正蓝旗周围51个气象台站资料，以2010年年均温和年降水为例，对比分析了克里格插值法和ANUSPLIN插值法的优缺点，最终选择ANUSPLIN插值法对正蓝旗2005-2010年的各气象要素（年均温、年降水量、生长季均温、生长季降水量）进行插值计算。正蓝旗年均温在1.3℃-3.9℃间，年降水量在200mm-450mm间，年际间气温和降水量均存在较大的波动性。生长季均温在13.5℃-18.5℃间，年际间变化幅度较小；生长季降水量在130mm-380mm间，多年的平均生长季降水量占年总降水量的80%以上。本文在以下两个方面有所创新：（1）综合考虑影响草原生物量的主要因素，利用B-P神经网络法构建草原生物量遥感估算模型，深入分析了正蓝旗2005-2010年草原生物量的时空动态变化。同时，在构建草原生物量遥感估算模型过程中，地面样方数据选用多年连续观测资料建模，降低了年际间生物量波动的影响，有效地提高了估算模型的稳定性。（2）参照草原沙化分级指标体系，将正蓝旗草原区划分为未沙化草地、轻度沙化草地、中度沙化草地和重度沙化草地，探讨了不同沙化程度下草原地下/地上生物量的大小及其分配机制，估算了正蓝旗不同沙化程度下的草原碳贮量，揭示了草原沙化对草地生态系统碳循环的影响作用，提高了草原生物量及其碳贮量的估算精度。更多还原

【Abstract】 Grassland ecosystem is the largest terrestrial ecosystem of China. Grassland productivity is thematerial base for maintaining grassland ecosystem, then the most direct index to reflect grassland statusand the decisive role to function of grassland ecosystem. It has been the important and hot spot ofterrestrial ecology to research grassland biomass. It is the important significance for sustainable usingand management of grassland to accurately and rapidly investigate the temporal-spatial distribution ofgrassland production and master the dynamic law of grassland. At present, whether on the large scale orglobal scale and the regional or grassland type to estimate grassland biomass and carbon storage, thedata of modeling and verification is less. It is difficult to ensure the reliability and precision of model.The relative deficiency of belowground biomass has long been one of the weak links of ecological study.Especially, there is a lack of systemic and through study on the relationship between below-andabove-ground biomass and their dynamic change under different utilization types and grasslanddegradation and desertification.In this study, we selected Zhenglan Banner, Inner Mongolia as our study area. Based on ‘3S’technology method and special database, we built Remote Sensing estimation model for grasslandbiomass. We illustrated temporal-spatial distribution of grassland biomass during2005to2010, theninvestigated the main control factors of temporal-spatial distribution of grassland biomass, and revealedabove-and below-ground biomass and allocation, then further estimated grassland biomass and carbonstorage under the different desertification grade. The main results and progress are summarized asfollows:(1) The estimating model of aboveground biomass based on RS was established by traditionalexperience statistical method and B-P neural network method using ground quadrat biomass data.Through precision verification, the optimal model was B-P neural network model and model precisionwas79%. By B-P neural network model, grassland aboveground biomass was estimated at annual mean798,620ton, biomass density at797.36kg/hm~2. In region spatial distribution pattern was “middle higherand sides lower”. Precipitation was one of key factors for temporal distribution of grassland biomass.(2) Based on TM image at Zhenglan Banner in2010, referencing for the index system of grasslanddesertification grade, grassland was divided into nonsandy grassland, slight type grassland, moderatetype grassland, severe type grassland. Classification precision was80%. The results showed thatgrassland desertification mainly centralized on Northern region. Desertification area was4,483.95km~2,accounted for45.88%of total grassland area.(3) Mean aboveground biomass density of nonsandy grassland, slight type grassland, moderatetype grassland and severe type grassland was estimated at214.72g/m~2、135.41g/m~2、91.42g/m~2、24.01g/m~2, respectively. Mean belowground biomass density was estimated at2601.35g/m~2、2318.45g/m~2、413.25g/m~2、117.25g/m~2, respectively. The ratio of below-and above-ground biomass was between4.5and17.2. With aggravation of desertification degree, it made the soil texture coursing and soil fertility reduced. Then it inhibited vegetation growth and reduced grassland biomass.(4) On the basis of grassland desertification classification and the estimation of abovegroundbiomass by Remote Sensing, combined with the ratio of below-and above-ground biomass, grasslandbiomass and carbon storage were estimated at7.55million ton and3.4TgC, respectively. Carbondensity was between0.99MgC/hm~2and11.12MgC/hm~2. Among them, mean carbon density ofnonsandy grassland was at4.39MgC/hm~2, carbon storage was at2.31TgC. Mean carbon density ofslight type grassland was at5.61MgC/hm~2, carbon storage was at0.55TgC. Mean carbon density ofmoderate type grassland was at1.58MgC/hm~2, carbon storage was at0.33TgC. Mean carbon density ofsevere type grassland was at1.53MgC/hm~2, carbon storage was at0.22TgC. The above resultsindicated that grassland desertification has significantly negative effect of carbon fixation. Theaggravation of the desertification may lead to reduce the ability of storage carbon.(5) In order to complete all the above research objects and contents, taking annual meantemperature and precipitation in2010for example, selecting51meteorological station data aroundZhenglan Banner and comparing analysis of advantages and disadvantages of kriging interpolation andANUSPLIN interpolation method, we finally chose ANUSPLIN interpolation method to calculate eachmeteorological element(including annual mean temperature, annual mean precipitation, growing seasontemperature, growing season precipitation) of Zhenglan Banner during2005to2010. Annual meantemperature was between1.3℃and3.9℃, annual mean precipitation was between200mm and450mm,inter-annual temperature and precipitation had large fluctuation. Growing season temperature wasbetween13.5℃and18.5℃. The change range was smaller in different years. Growing seasonprecipitation was between130mm and380mm. It accounted for more80%of total precipitation.There are innovations in tow aspects as follows:(1) Through comprehensively main factors effecting on grassland biomass, the estimating model ofgrassland biomass by Remote Sensing was established using B-P neural network method. This papermade a profound analysis on space-time dynamic change of grassland biomass during2005to2010.Meanwhile, six years continuous observation data was selected to establish model. It reduced thefluctuation infuluences of biomass annually and effectively improved the stability of model.(2) Referencing for the index system of grassland desertification grade, grassland was divided intononsandy grassland, slight type grassland, moderate type grassland, severe type grassland. Weinvestigated above-and below-ground biomass and allocation on the different desertification grade, thenestimated grass carbon storage on the different desertification grade, and further revealed the effect oncarbon cycle in grassland ecosystem and improved estimation accuracy of grassland biomass and carbonstorage.更多还原