Abstract

A lake sediment sample is usually a mix of deposits derived from multiple processes operating over the time span of its accumulation, and the grain-size distribution (GSD) of a sediment sample is therefore a mix of one or more subpopulations produced by these processes. The log-normal distribution function has been used to describe the GSD of natural sediments for over a century, but its symmetric property makes it incapable of describing the distribution skewness that is commonly seen in the GSD of clastic sediments. This study presents a new nonparametric method called the skew normal distribution deconvolution (SNDD) that parses a sample GSD into one or more component subpopulations, each of which can be fully quantified by a parameter set () of a skew normal distribution (SND). This method was implemented with the “adaptive mesh refinement” algorithm so that the optimal partition can be achieved deterministically to a user-desired degree of match between original GSD data and SNDD result. A comparison between SNDD and normal distribution partition indicates that SNDD deconvolved a sample GSD to five discrete subpopulations but the normal distribution would partition the same GSD to 12 subpopulations. This implies that SNDD may parse GSD data much more effectively to subpopulations indicative of natural deposits. Furthermore, each SND will have a set of rich statistical parameters (different mean and median, characteristic skewness and kurtosis) that can be used to characterize sediments and depositional environments. We acquired clastic GSD data of 530 samples from Lake Bosumtwi using an LS-200 Coulter particle-size analyzer and used this new method to deconvolve these data to their component subpopulations. The results reveal seven disjointed subpopulation groups in the mean grain-size distribution (MGSD) field, and each of their volume frequencies can be fitted to a normal distribution. They are one dominant subpopulation with the median grain size of 11.8 μm, and six minor subpopulations with median grain sizes of 0.86 μm, 7.61 μm, 24.4 μm, 60.1 μm, 129.8 μm, and 439.6 μm. They provide strong indications that seven discrete depositional processes and/or environments might be responsible for producing these clastic grains over 94,500 years of deposition in Lake Bosumtwi. We also observed that an SNDD may re-partition grains of the same size into different components. On average, SNDD of our 530 samples partitioned 91.5% of clay grains to components whose mean grain sizes (MGS) are in the silt grain-size range, and 40.2% of sand grains to components whose MGS's are in the silt grain-size range; this implied that the majority of clay grains and 40% of sand grains were genetically related to silt components.