5:15 PM - 7:15 PM
[HQR05-P09] A Novel Graph-Based Approach for Correlation Modelling in Sediment Samples
Keywords:correlation model, age model, depth management, application
In order to precisely decode the time-series information preserved in sediments, it is essential to obtain a continuous material. However, technical limitations often make it difficult to recover all target horizons in a single drilling operation, and as a result, many actual sediment samples are composed of multiple sections. Due to the gaps between these sections, the samples often do not provide sufficient continuity for analysis.
In recent years, therefore, virtually continuous sediment samples have also been reconstructed by splicing together multiple samples from different boreholes, such as in the Lake Suigetsu project and the IODP. Sediment samples obtained from different holes each have independent drilling depth scales. In order to splice them together and treat them as a single continuous sample, it is necessary to construct a precise “correlation model” with a unified depth scale. Various approaches and definitions have been proposed for the construction and management of correlation models. Here, we discuss the modelling method basically following the definitions of Nakagawa et al. (2012, QSR) using the varved sediments of Lake Suigetsu in Fukui Prefecture, Japan as an example.
Several drilling campaigns have been carried out at Lake Suigetsu since 1993. After 2006, coring was carried out from multiple holes, and correlation models were constructed in which all sections could be defined on a unified depth scale by precisely correlating each section. The depths and ages of sediment samples have been semi-dynamically calculate based on the correlation model including pre-calculated depths, using the “Level Finder” correlation model management application (http://polsystems.rits-palaeo.com/). However, the previous method imposed certain constraints on construction and representation of the model, for example, requiring that some depths to be pre-calculated. Therefore, to achieve efficient construction and management of correlation models, we propose a modelling approach based on graph theory that is suitable for dynamic construction. We also introduce a newly designed “Level Compiler” application for constructing and managing correlation models (https://github.com/keitaroyamada/Level-Compiler), which we have designed and built as a proof-of-concept platform for implementing and testing this approach.
A correlation model has a complex network structure, as it is constructed by connections between individual laminae in each core section. To simplify this structure, we have successfully modelled it as a directed graph by considering each lamina as a node and vertical/horizontal connections between lamina as edges. By adjusting the length of the edges, this approach can also incorporate erosional and depositional events and is therefore potentially applicable to a wide variety of sedimentary environments. In addition, because this correlation model is essentially a directed graph, existing exploration algorithms can be applied. In this study, we applied a depth-first search algorithm to the correlation model, resulting in dynamic and efficient correlation modelling and depth/age calculations.
In recent years, with the development of technology, a variety of analyses have been performed, leading to the use of parallel sections for sampling. It is more important than ever to construct precise correlation models and manage their depth/ages. The modelling approach proposed in this study and its implementation, Level Compiler, provide one of the approaches to achieve precise depth management and its visualisation of such sediment samples. In particular, the intuitive “Level Compiler” may contribute to more efficient management the depth and age of various samples, including Quaternary sediments.
In recent years, therefore, virtually continuous sediment samples have also been reconstructed by splicing together multiple samples from different boreholes, such as in the Lake Suigetsu project and the IODP. Sediment samples obtained from different holes each have independent drilling depth scales. In order to splice them together and treat them as a single continuous sample, it is necessary to construct a precise “correlation model” with a unified depth scale. Various approaches and definitions have been proposed for the construction and management of correlation models. Here, we discuss the modelling method basically following the definitions of Nakagawa et al. (2012, QSR) using the varved sediments of Lake Suigetsu in Fukui Prefecture, Japan as an example.
Several drilling campaigns have been carried out at Lake Suigetsu since 1993. After 2006, coring was carried out from multiple holes, and correlation models were constructed in which all sections could be defined on a unified depth scale by precisely correlating each section. The depths and ages of sediment samples have been semi-dynamically calculate based on the correlation model including pre-calculated depths, using the “Level Finder” correlation model management application (http://polsystems.rits-palaeo.com/). However, the previous method imposed certain constraints on construction and representation of the model, for example, requiring that some depths to be pre-calculated. Therefore, to achieve efficient construction and management of correlation models, we propose a modelling approach based on graph theory that is suitable for dynamic construction. We also introduce a newly designed “Level Compiler” application for constructing and managing correlation models (https://github.com/keitaroyamada/Level-Compiler), which we have designed and built as a proof-of-concept platform for implementing and testing this approach.
A correlation model has a complex network structure, as it is constructed by connections between individual laminae in each core section. To simplify this structure, we have successfully modelled it as a directed graph by considering each lamina as a node and vertical/horizontal connections between lamina as edges. By adjusting the length of the edges, this approach can also incorporate erosional and depositional events and is therefore potentially applicable to a wide variety of sedimentary environments. In addition, because this correlation model is essentially a directed graph, existing exploration algorithms can be applied. In this study, we applied a depth-first search algorithm to the correlation model, resulting in dynamic and efficient correlation modelling and depth/age calculations.
In recent years, with the development of technology, a variety of analyses have been performed, leading to the use of parallel sections for sampling. It is more important than ever to construct precise correlation models and manage their depth/ages. The modelling approach proposed in this study and its implementation, Level Compiler, provide one of the approaches to achieve precise depth management and its visualisation of such sediment samples. In particular, the intuitive “Level Compiler” may contribute to more efficient management the depth and age of various samples, including Quaternary sediments.