【作者】 徐小军；

【导师】 周国模；

【作者基本信息】 北京林业大学 ， 森林经理学， 2013， 博士

【摘要】 森林生态系统碳水通量观测和模型模拟是森林碳水循环监测研究的主要方向之一,该研究有利于准确地评估森林生态系统碳水通量和编制森林固碳经营管理措施以应对气候变化,丰富森林可持续经营内涵。极端天气事件对森林造成的干扰已经成为森林碳水通量模型模拟中的重要组成部分,给精准监测碳水通量带来很大不确定性。与其他森林类型相比,竹林具有显著的蒸腾和固碳潜力,其生物学特征及受干扰程度具有明显的差异。由于缺乏针对竹林生态系统的长期观测站点,冠层尺度竹林生态系统碳水循环特征有待研究。结合安吉县毛竹林生态系统通量塔地面观测、卫星遥感影像(Landsat TM和MODIS)和再分析数据(Modern-Era Retrospective Analysis for Research and Applications, MERRA)等数据集,该文定量分析了毛竹(Phyllostachys edulis)林生态系统碳通量变化特征及其与环境和生物因子的关联性：构建了适合毛竹林生态系统的碳水通量监测模型,实现站点尺度向区域尺度上推；着重揭示了干旱和冰雪灾害等自然干扰对毛竹林生态系统碳水通量的影响。该研究为评价毛竹林生态系统对区域碳水平衡的贡献提供参考依据、为区域尺度碳水通量精准监测及空间分析提供方法、为制定应对气候变化的毛竹林固碳经营管理措施提供参考资料。该文得出的结果和结论主要包括以下几个方面：(1)根据湍流稳态测试、方差相似性关系和能量平衡闭合分析对通量塔观测数据质量评价,结果表明：站点观测的湍流通量数据满足常通量层假设,符合涡度相关技术的基本要求,数据质量相对较可靠；垂直风速和温度归一化的标准差是大气稳定度的普适函数；适宜的夜间通量摩擦风速u*界限值为0.24～0.33m s-1；能量闭合程度为71%,夜间能量平衡闭合程度较差,降水和异质性地形是引起本站点能量不闭合的主要原因。(2)对2011年毛竹林生态系统碳通量观测数据进行校正、插补和组分分解,结果表明：对于生态系统呼吸(Re)通量组分,基于夜间数据和基于白天数据的2种插补方法差异性较大,表明通量塔Re数据存在较大的不确定性；2011年毛竹林年生态系统净交换量(NEE)、总初级生产力(GPP)和Re分别为417.5gCm-2y-1、1899.7gCm-2y-1和-1482.2gCm-2y-1.NEE在10：00～11：00达到最高峰,午后的NEE略低于上午。毛竹林固碳能力较强,四季均表现为碳汇。(3)采用通量塔观测数据对基于遥感的Penman-Monteith蒸散模型(RS-PM)和光能利用率模型(EC-LUE)进行参数化,建立了毛竹林生态系统蒸发散(ET)和GPP监测模型,并以MODIS和MERRA数据为驱动变量,估算安吉县毛竹林ET和GPP,结果表明：与通量塔观测结果相比,最优模型估算的8天ET和GPP平均值相对均方根误差(RMSEr)分别为22.35%和17.96%；考虑GPP饱和效应有助于提高EC-LUE模型精度,RMSEr从32.79%降低到17.96%。(4)碳通量各组分与环境和生物因子相关性分析结果表明：GPP与温度(T)相关性最高(P<0.01),依次为叶面积指数(LAI)>归一化植被指数(NDVI)>水蒸气压差(VPD)>光合有效辐射(PAR)>土壤含水量(SVWC); PAR、LAI和VPD是NEE的主要影响因子(P<0.01)；Re与T相关性最高,其次为NDVI和LAI(P<0.01)。建议将LAI和NDVI引入区域尺度毛竹林生态系统Re估算模型中。(5)将整个干旱过程分成开始期、干旱期和恢复期3个阶段,分析了干旱对毛竹林生态系统NEE、GPP和Re的影响,结果表明：干旱显著地改变环境因子与GPP、NEE和Re的相关性；在干早期不同阶段,起关键作用的环境因子发生变化；综合VPD和SVWC能较好地表征干旱对碳通量的影响；动态线性回归模型模拟结果表明干旱降低了GPP和NEE,两者降低量分别为12.7%和44.8%,但对Re影响较少,可见减少了固碳能力和增加了碳排放量。将VPD和SVWC引入EC-LUE模型中,能提高干旱期GPP估算精度。(6)利用2004-2011年安吉县毛竹林ET和GPP分布图,分析了2008年冰雪灾害对毛竹林生态系统ET和GPP的影响以及地形与冰雪灾害影响程度的关系,结果表明：冰雪灾害降低了毛竹林生态系统年GPP和ET,平均GPP减少量为0.17gC m-2d-1;处于小年期的毛竹林GPP和ET受冰雪灾害的影响程度稍高于大年毛竹林；GPP受冰雪灾害影响程度与地形的关联度高于ET,海拔和坡度与GPP降低量呈正相关关系。更多还原

【Abstract】 Forest ecosystem carbon and water fluxes modeling is one of the main directions of the forest carbon cycle monitoring research. This research is helpful to accurately assess the contribution of forest ecosystem to global carbon and water balance and to make forest carbon planning for mediating climate change. Forest disturbances by extreme weather events cause large uncertainties in carbon and water flux estimates and become an important part of forest carbon and water fluxes modeling. Compared with other forest types, bamboo forest has significant transpiration and carbon sequestration potential and its biological characteristics are special. It is intensively managed by farmers. Due to the lack of long-term observation site in bamboo forest ecosystem, bamboo forest ecosystem carbon and water cycle in the canopy scale was rarely reported. Therefore, combined the flux observation tower of Moso bamboo forest ecosystem in Anji county, Landsat Thematic Mapper (TM), MODerate resolution Imaging Spectroradiometer (MODIS) data, and Modern Era Retrospective-Analysis for Research and Applications (MERRA), this study quantitatively analyzed variations of Moso bamboo(Phyllostachys edulis) forest ecosystem carbon and water fluxes and their correlations with environmental and biological factors; Models for carbon and water fluxes were built and used to scale site-based fluxes up to regional-based fluxes; Effects of snow damages and drought on Moso bamboo forest ecosystem carbon and water fluxes were also analyzed. Results from this study were a reference for evaluating the contribution of Moso bamboo forest ecosystem to the regional carbon and water balance, provided methods for accurately estimating carbon and water fluxes in regional scale, and presented reference data for making Moso bamboo forest carbon planning for mediating climate change. Results and conclusions are mainly included in the following aspects:(1) Quality of turbulent flux measurements of Moso bamboo forest flux tower was evaluated using steady state test, Monin-Obukhov similarity function, and energy balance closure. Results showed that:COt flux measurements are met with the constant flux layer hypothesis and basic requirements of eddy covariance technique; Standard deviates of vertical wind speed and temperature normalized by a scaling velocity or temperature are universal functions of atmospheric stability; Minimum values of friction velocity for each seasonal nocturnal flux are between0.24and0.33m s-1; Energy balance closure is relatively low with value of71%due to precipitation and heterogeneous terrain, especially for nocturnal period.(2) Correction, gap-filling and flux partitioning were implemented to carbon flux measurements of Moso bamboo forest in2011. Results showed that, for the ecosystem respiration (Re), there is significant difference between night-based method and daytime-based method, implying huge uncertainty in Re; Net ecosystem exchange (NEE), gross primary productivity (GPP) and Re in2011were417.5g C m-2y-1,1899.7g C m-2y-1, and-1482.2g C m-2y-1. NEE reaches its peak at10:00-11:00, and it is slightly lower in the afternoon than in the morning. Moso bamboo forest has strong ability in carbon sequestration and is a carbon sink in each season.(3) The eddy covariance (EC) data collected from the Anji flux tower were used in this study to calibrate and validate the remote sensing-driven Penman-Monteith (RS-PM) and Eddy Covariance Light Use Efficiency (EC-LUE) models, and the models driven by the MODIS and MERRA datasets were used to estimate the evapotranspiration (ET) and GPP of Moso bamboo forest in Anji county. Results showed that the relative root mean square error of ET and GPP estimates were22.35%and17.96%compared with measurements, respectively; The saturation effect of GPP was taken into account in the EC-LUE model, which played an important role in improving the model performance with relative root mean square error reduced from32.79%to17.96%.(4) Correlation analysis between carbon flux components and environmental and biological factors showed that the correlation between GPP and temperature (T) was the highest (P<0.01), followed by Leaf Area Index (LAI)> Normalized Difference Vegetation Index (NDVI)> Vapor Pressure Deficit (VPD)> Photosynthetically Active Radiation (PAR)> Soil Volumetric Water Content (SVWC); PAR, LAI, and VPD were main control factors for NEE (P<0.01); Correlation between Re and T was the highest, followed by NDVI, LAI (P<0.01). LAI and NDVI can serve as inputs to model for estimating regional Re.(5) The whole drought period was divided into beginning phase, drought phase and recovery phase. The influences of drought on NEE, GPP and Re were analyzed. Results showed that the correlations between environmental factors and GPP, NEE and Re were significantly changed due to drought effect; Key control factors for the carbon flux were different in each phase of drought; Combined VPD and soil moisture content can be used to well describe occurrence and influence of drought; Simulation results from the dynamic linear regression model showed that the decreases in GPP and NEE caused by drought were12.7%and44.8%, respectively. Sensitivity of Re to drought was lower than those of GPP and NEE. Re did not decrease during drought period. The accuracy of GPP estimates during drought period was improved when VPD and SVWC were inputted into the EC-LUE model.(6) The impacts of ice storm on GPP and ET of Moso bamboo forests and their relationships with terrain factors were analyzed based on ET, GPP and digital elevation data from2004to2011. Results showed that the early2008ice storm caused a slight decrease in annual ET and GPP, and this ice storm decreased annual mean GPP by0.17g C m-2d-1; Ice storm damage effects on ET and GPP for the off-year were slightly greater than those for the on-year; Relationships between decreases in GPP and terrain factors were more significant than those between decreases in ET and terrain factors. The correlations between decreases in GPP and altitude and slope were positive.更多还原