沉积学报2025，Vol.43Issue(2)：361-375,15.DOI:10.14027/j.issn.1000-0550.2023.117
大数据背景下粒度分布沉积信息挖掘方法进展
Progress on Mining Methods of Sedimentological Information from Grain-size Distribution under the Background of Big Data
摘要
Abstract
[Significance]The grain sizes of sediments contain information on multiple factors:transport path,depo-sitional process,and environment.Grain-size distribution(GSD)is defined in sedimentology and geology as the fre-quency of occurrence of different-diameter particles.GSD is a record of the original sedimentological information.It is one aspect of the basic data used to reveal modern and ancient depositional environments in rivers,lakes,oceans,deserts,loess,etc.The traditional GSD analytical methods adopted to describe the overall features of depositional processes and environments,either qualitatively or semi-quantitatively,may not overcome problems of quantification and multiple solutions.[Progress]This study summarizes the range of different classification standards of grain-size scale,and compares moment and graphical frequency-curve methods of describing GSDs with morphological descrip-tion standards.The applicability and usage of traditional methods of sedimentary environment analysis by GSD are re-viewed,and some unconventional approaches are developed using mathematical methodology to tackle the entire range of GSD.Unsupervised clustering algorithms calculate the similarity of GSDs using their frequency,cumulative frequency or statistical parameters,then depositional environments are sorted according to the classes of clustering.Multifractal analysis is used to extract fractal parameters that represent the complexity of GSD frequency data.The dif-ferent fractal structures reveal different depositional properties.When applied to multiple sedimentary processes in different sedimentary environments and dynamics and the GSD is superposed by multi-subpopulations,the corre-sponding frequency curve is found to be bimodal or multimodal.This implies that an inverse unmixing model of the sediments is ideally suited for obtaining genetically meaningful interpretations of these subpopulations.Two tech-niques are used to separate the grain-size component from GSD frequency data.To apply the statistical finite-mixture model,single-sample unmixing(SSU)uses a probability density function(normal,skew normal or Weibull distribu-tion)to unmix the GSD by curve-fitting techniques.Each grain-size component is distributed in a unimodal fashion such that its statistical parameters(mean,sorting,skewness,kurtosis and percentage)may be calculated.The end-member modeling algorithm(EMMA)decomposes grain-size end-members from a GSD dataset.These unimodal or multimodal grain-size end-members are linearly independent and fixed within a single GSD dataset.Many improved EMMAs are available in different open-source tools.To introduce examples of the application of these unconventional methods,in this study 27 GSDs from the central bar of the Kangshan River in the Poyang Lake drainage are pro-cessed by clustering,multifractal,SSU and EMMA.[Conclusions and Prospects]Problems of sedimentation analy-sis and the big-data properties of GSDs are solved.The trend of development of the depositional significance of GSDs is proposed based on analytical methods.With the advent of various modern grain-size analysis techniques and more sophisticated artificial intelligence procedures in earth sciences,new increasingly intelligent mining methods for GS-Ds are emerging for understanding the spatio-temporal grain-size patterns in sediments.Some excellent sedimentologi-cal related databases have been constructed.Accordingly,an open-access database will be established for GSDs to in-clude various kinds of data,intelligent methods and a literature of reported research.Under the background of big da-ta,GSD big-data technology will provide a new driver for mining depositional properties intensively,and integrate them into sedimentological big data.Four phases,initial,exploratory,early development and rapid development,de-scribe the history of GSD research.The future must hold a big-data phase for intelligent mining using sedimentologi-cal GSD information.关键词
大数据/粒度分布/沉积信息/智能挖掘Key words
big data/grain-size distribution/sedimentological information/intelligent mining分类
地质学引用本文复制引用
袁瑞..大数据背景下粒度分布沉积信息挖掘方法进展[J].沉积学报,2025,43(2):361-375,15.基金项目
国家自然科学基金项目(42202113)[National Natural Science Foundation of China,No.42202113] （42202113）
