3:30 PM - 4:00 PM
[MGI33-06] Sedimentology as a science of reconstruction
★Invited Papers
Keywords:inverse analysis, turbidite, tsunami deposit
Sedimentology is a geological science that aims to understand the mechanisms of sedimentation and erosion that form landforms, and to reconstruct the past environment of the Earth's surface from information recorded in strata. Strata are accumulations of sedimentary rocks formed by buried landforms that once existed on the Earth's surface. Landforms change their shape in response to variation in sediment supply and water discharge caused by orogenic activity and climatic changes, and these topographic changes are preserved in the strata. Sedimentology solves the inverse problem of the environmental changes that resulted in the landform changes from the strata. In the field of sedimentology, however, quantitative inverse analyses have rarely been performed.
This situation is changing drastically with the recent development of morphodynamics. Morphodynamics is a research field that combines fluid dynamics and sediment transport models to understand the dynamics of landform development due to sedimentation and erosion. Morphodynamics has been developed in the fields of river and coastal engineering, and has been mainly applied to river and coastal management and disaster prevention. Recently, however, it has been recognized that morphodynamics is an effective research method for understanding longer-term, large-scale geomorphic development, and has been widely accepted as a research method in the field of sedimentology. In this talk, we will discuss the recent advances in the field of geomorphology and sedimentology, including the use of morphodynamics-based forward models and inverse analysis. We will show the recent trend of research using inverse analysis in the field of sedimentology, and introduce the efforts of our research group.
In sedimentology, inverse analysis is particularly popular in (1) reconstruction of uplift rates of mountainous regions from longitudinal profiles of rivers and (2) estimation of hydraulic conditions from tsunami deposits. In addition to the above, our research group is working on (3) estimation of hydraulic conditions of turbidity currents from turbidites. (1) The inverse analysis of mountain uplift rate can reconstruct the uplift rate of the continental crust over a wide area and a long period of time (more than several hundred thousand years). (2) The inverse analysis of tsunami sediments is an important research topic for social application of sedimentology because it can estimate the magnitude of tsunamis in the past and the disaster risk of the region. (3) The inverse analysis of turbidite is an attempt to estimate the conditions of turbidity currents that occur in deep sea environments, which are difficult to observe directly. Turbidity currents that deposit turbidites build huge topographic features called submarine fans on the deep seafloor, and the deposits often become large-scale oil and gas reservoirs, making them an economically important subject of research. The ability to read the hydraulic conditions of turbidity currents is expected to shed light on the shape and properties of deep-sea reservoir rocks from limited geological information.
In order to solve these research problems, we are working on the development of inverse models using artificial neural networks. A serious problem in inverse analysis in sedimentology is the high computational loads of forward models. Common inverse modeling methods such as the Markov Chain Monte Carlo (MCMC) method often require tens of thousands of iterations of the forward model calculation, which cannot be parallelized and therefore cannot be used for inverse analysis of sedimentological subjects. Therefore, we attempted to generate an inverse model by performing several thousand iterations of the forward model with randomly generated initial conditions, and training the neural network with the resulting combination of sedimentary data and initial conditions as teacher data. The advantage of this method is that the generation of the teacher data can be completely parallelized, so that the computational load of the forward model becomes less of a problem. This method is particularly useful for the inverse analysis of tsunamis and turbidity currents in (2) and (3), and it has been confirmed that it can reproduce the hydraulic conditions of modern tsunamis and recover the hydraulic conditions of experimental turbidity currents from turbidites in experiments. In the future, it is expected that the analysis of past tsunami sediments and submarine fan deposits will be carried out.
This situation is changing drastically with the recent development of morphodynamics. Morphodynamics is a research field that combines fluid dynamics and sediment transport models to understand the dynamics of landform development due to sedimentation and erosion. Morphodynamics has been developed in the fields of river and coastal engineering, and has been mainly applied to river and coastal management and disaster prevention. Recently, however, it has been recognized that morphodynamics is an effective research method for understanding longer-term, large-scale geomorphic development, and has been widely accepted as a research method in the field of sedimentology. In this talk, we will discuss the recent advances in the field of geomorphology and sedimentology, including the use of morphodynamics-based forward models and inverse analysis. We will show the recent trend of research using inverse analysis in the field of sedimentology, and introduce the efforts of our research group.
In sedimentology, inverse analysis is particularly popular in (1) reconstruction of uplift rates of mountainous regions from longitudinal profiles of rivers and (2) estimation of hydraulic conditions from tsunami deposits. In addition to the above, our research group is working on (3) estimation of hydraulic conditions of turbidity currents from turbidites. (1) The inverse analysis of mountain uplift rate can reconstruct the uplift rate of the continental crust over a wide area and a long period of time (more than several hundred thousand years). (2) The inverse analysis of tsunami sediments is an important research topic for social application of sedimentology because it can estimate the magnitude of tsunamis in the past and the disaster risk of the region. (3) The inverse analysis of turbidite is an attempt to estimate the conditions of turbidity currents that occur in deep sea environments, which are difficult to observe directly. Turbidity currents that deposit turbidites build huge topographic features called submarine fans on the deep seafloor, and the deposits often become large-scale oil and gas reservoirs, making them an economically important subject of research. The ability to read the hydraulic conditions of turbidity currents is expected to shed light on the shape and properties of deep-sea reservoir rocks from limited geological information.
In order to solve these research problems, we are working on the development of inverse models using artificial neural networks. A serious problem in inverse analysis in sedimentology is the high computational loads of forward models. Common inverse modeling methods such as the Markov Chain Monte Carlo (MCMC) method often require tens of thousands of iterations of the forward model calculation, which cannot be parallelized and therefore cannot be used for inverse analysis of sedimentological subjects. Therefore, we attempted to generate an inverse model by performing several thousand iterations of the forward model with randomly generated initial conditions, and training the neural network with the resulting combination of sedimentary data and initial conditions as teacher data. The advantage of this method is that the generation of the teacher data can be completely parallelized, so that the computational load of the forward model becomes less of a problem. This method is particularly useful for the inverse analysis of tsunamis and turbidity currents in (2) and (3), and it has been confirmed that it can reproduce the hydraulic conditions of modern tsunamis and recover the hydraulic conditions of experimental turbidity currents from turbidites in experiments. In the future, it is expected that the analysis of past tsunami sediments and submarine fan deposits will be carried out.