Grain-size distribution from different surface sediment samples of Lake Towuti, Indonesia@en

A digital elevation model (DEM) of Lake Towuti and its surrounding was calculated using ArcGIS (Esri, Inc., Redlands, CA, USA). The model is based on open source satellite data for Sulawesi provided by the United States Geological Survey (Aster Global DEM based on the Shuttle Radar Topography Mission carried out by the National Aeronautics and Space Administration at 1 arc-second 30 m spatial resolution). Spatial interpolation of the analytical surface sediment data was carried out with the software Surfer 9 (Golden Software Inc., Golden, CO, USA) using the kriging method. Statistical analyses employed on the surface sediment data sets comprise end-member (EM) unmixing, principal component analysis (PCA) and a redundancy analysis (RDA). EM analyses were carried out on normalized and standardized grain-size (EMGS), chemical (EMChem) and mineralogical (EMMin) data sets. Assuming a sedimentary mixture from different sources the mixing model in all cases can be written as: X = AS + E (1) where X represents the n-by-m matrix of n samples (one per row) and m variables (relative abundance of individual data). Matrix A (n-by-l) denotes the mixing proportion of l end-members for the n samples, S represents the m properties of the l EMs and E is the error matrix of residuals. The uncertainties of the EM analyses are controlled by the errors of the data sets used. The EM algorithm developed by Heslop and Dillon (2007) adopting the approach of Weltje (1997) was applied. The decision criterion of how many EMs are included in the three models is based partly on the coefficients of determination derived from the PCA. Nevertheless, the number of the respective EMs should also be reasonable in the geological context of the data set (Weltje, 1997; Weltje and Prins, 2007). Residuals of the EM models include analytical errors and non-identified additional sources of variability. All other multivariate statistical analyses were carried out with the Excel-based software Addinsoft XLSTAT (STATCON GmbH, Witzenhausen, Germany) The PCA was conducted with the sand content and the concentrations of selected elements determined by ICP-MS and XRF analyses (Fe, Mg, Al, Si, K, Ca, Cr and Ni). In the RDA, the results derived from the PCA are expanded by the concentrations of major minerals, the MS and TOC values and the C/N ratio. […]本研究针对托武蒂湖（Lake Towuti）及其周边区域的数字高程模型（DEM，digital elevation model），依托美国地质调查局提供的苏拉威西岛开源卫星数据构建，该数据为基于美国国家航空航天局航天飞机雷达地形测绘任务的ASTER全球DEM，空间分辨率为1弧秒（30米），并通过ArcGIS软件（Esri公司，美国加利福尼亚州雷德兰兹市）完成计算。针对表层沉积物分析数据的空间插值，采用克里金插值法，借助Surfer 9软件（Golden Software公司，美国科罗拉多州戈尔登市）完成。针对表层沉积物数据集的统计分析涵盖端元（EM，end-member）解混、主成分分析（PCA，principal component analysis）及冗余分析（RDA，redundancy analysis）。端元分析针对经过归一化与标准化处理的粒度（EMGS）、化学（EMChem）及矿物学（EMMin）数据集开展。假设沉积物由多种来源混合而成，所有场景下的混合模型均可表示为：X = AS + E （1）其中X代表n个样本（每行对应1个样本）与m个变量（单一组分的相对丰度）构成的n×m矩阵；矩阵A（n×l）表示n个样本中l个端元的混合占比；S代表l个端元的m项属性；E为残差误差矩阵。端元分析的不确定性由所用数据集的误差控制。本研究采用Heslop与Dillon（2007）提出的、借鉴Weltje（1997）方法的端元解混算法。三种模型中端元数量的判定标准，部分基于主成分分析得到的决定系数，但端元数量也需符合数据集的地质背景合理性（Weltje，1997；Weltje与Prins，2007）。端元模型的残差包含分析误差与未识别的额外变异性来源。其余所有多元统计分析均借助基于Excel的Addinsoft XLSTAT软件（STATCON GmbH公司，德国维岑豪森市）完成。主成分分析以砂粒含量及电感耦合等离子体质谱法（ICP-MS）与X射线荧光光谱法（XRF）测定的选定元素（Fe、Mg、Al、Si、K、Ca、Cr及Ni）浓度为基础开展。在冗余分析中，主成分分析的结果辅以主要矿物浓度、磁化率（MS）、总有机碳（TOC）值及C/N比值。[…]