鄱阳湖湿地干湿交替过程土壤有机碳三维时空变异性研究

Wet-dry cycling environment frequently over the typical land use/coverage landscape of Poyang Lake wetland, but little is known of its impact on soil carbon variations and soil carbon source/sink, process that are potentially important in global carbon cycle. To investigate this issue, understanding the scaling effect on the spatial variation of soil organic carbon density (SOCD) is necessary for determination of reasonable soil sampling intervals. In this study, Ancillary variable, such as soil apparent electrical conductivity (ECa), which is measured by an EM38 sensor, will assume to provide indirect information about the spatial distribution of SOCD. By using the method of discrete wavelet transform ( DWT ) on the spatial sampling data of SOC density along the a ECa transect. We derive wavelet approximate information and wavelet detailed information over six different scales. Then, optimizing sampling for SOCD. Ordinary kriging is based on the assumption that the variable of interest is intrinsically stationary. If there is trend, however, the assumption is untenable, and a more elaborate model of the variation is needed to take into account the trend. Extending geostatistics interpolation techniques to the three-dimensional or space-time domain is not simply adding another dimension, as there are some fundamental differences between the space and time domain , and the variation of SOCD at the vertical direction is much larger then the horizontal plane. A linear mixed model (LMM) is specified for the data, and the fixed and random effects by ordinary least squares (OLS) regression or REML. An anisotropic model, Bilonick model, is used to fit the variogram of the residuals. At last, the empirical best linear unbiased predictor or ordinary kriging is used for interpolation, and adding the fixed effects back for the predictions. Then, the three-dimensional geostatistics model with external drift is used for modeling the temporal-space variation on the transect or the three-dimensional variation of SOCD. For assessment the uncertainty of the prediction of soil carbon stocks by modeling the variation of SOCD at different scales, by different geostatistcis methods, by different soil sampling design method and different ancillary variables. At last, the total soil carbon source/sink is calculated by constraining the confidence limits of the variograms and the prediction confidence limits. The work can provide experience for soil carbon stocks prediction, and is important for identifying the effects of LULC change and climate on the SOC change, as well as wetland protection.

针对鄱阳湖湿地典型土地利用/覆盖景观格局干湿交替过程土壤碳源汇效应尚不明确。本研究首先利用离散小波方法分析断面样条采样电导率的尺度效应，为SOC采样提供决策依据。针对土壤变量的非稳态和三维建模尺度的不统一问题，将回归克立格和REML等非稳态地统计方法扩展到三维空间，引进Bilonick各向异性半方差函数模型，增加近地传感器测定的因子作为辅助变量，研制一套成熟的（外部趋势）三维地统计学方法。然后将其应用到典型断面或样区SOC时空或三维空间建模研究中；构建不同尺度、建模方法、采样设计方法、辅助变量等多不确定性影响因子下的SOC三维空间模型，提高碳库估算精度；以动态或静态比较的方式表征干湿交替过程土壤有机碳的三维时空变异规律，分析其与相关环境因子的耦合规律，揭示其碳源汇效应。相关研究可为土壤碳库的估算提供理论依据和方法支持，对于碳储量响应气候和土地利用/覆盖变化研究和湿度保护也具有重要意义。

鄱阳湖湿地是世界上重要的湿地资源，季节性的涨水使得鄱阳湖湿地形成了典型的“高草-低草-泥滩”景观格局，而这种干湿交替过程的土壤碳源汇效应尚不明确。本研究利用原位可见光-近红外高光谱遥感技术，利用线性PLSR和非线性LS-SVM等数据挖掘方法，采用额外参数正交化法（External parameter orthognolization，EPO）、光谱直接转换法（Direct Standardization，DS）、光谱分段直接转换法（Piecewise Direct Standardization，PDS）三种方法对吸收率转换后的原位光谱中的环境影响因素进行去除，开展基于水分去除的原位可见光-近红外光谱土壤有机碳预测研究；同时利用地面和Landsat8卫星遥感信息，建立反演模型，预测鄱阳湖湿地土壤有机碳，为后续数字土壤制图研究提供数据源。针对三维数字土壤制图研究中区域化变量的非稳态问题和土壤三维地统计插值尺度的不统一问题，引入REML方法和各向异性比关键参数，去除土壤有机碳在水平和垂直方向上的趋势，同时拟合残差的三维各向同性和各向异性半方差函数模型，再采用E-BLUP方法进行预测，利用预测方差表示模型的不确定性，形成一套成熟的三维地统计学建模方法。以典型断面为例，利用RK和REML方法构建多年份土壤有机碳时空变异模型，通过静态比较的方法，揭示鄱阳湖湿地典型景观上土壤有机碳的时空变异规律；最后，基于多源信息数据，利用地理加权回归克里格方法对环鄱阳湖区农田土壤有机碳开展空间变异及其影响因子研究。研究构建土壤有机碳空间变异模型，揭示了鄱阳湖湿地典型景观样区和断面干湿交替过程土壤有机碳时空变异规律。本研究开展三维地统计学方法及其在SOC时空变异性方面的研究可丰富和拓展当前计量土壤学（Pedometrics）、数字土壤制图（Digital Soil Mapping, DSM）研究领域的理论与方法，具有广阔的应用前景。