Search

It is a general problem in geoscience to estimate the time-series of velocity and temperature fields for a fluid based on limited observations, such as the flow velocity at the fluid surface and/or a temperature snapshot after flow. In this study, an adjoint-based data assimilation method (also known as 4-D variational data assimilation) was used to reconstruct the thermal convection in a highly viscous fluid (e.g. Earth's mantle) to investigate which observations constrain the thermal convection and how accurately the convection can be reconstructed for different wavelengths. The data assimilated to the adjoint-based model were generated synthetically from forward models with convecting cells of different length-scales. Based on the surface velocity and temperature snapshot, our simulations successfully reconstructed thermal convection over 50 Myr in the case that the wavelength of the convective cells is sufficiently large. We obtained two main results from this parametric study. (1) When we only considered instantaneous thermal structure fitting in the cost function, the convection reconstruction tended to fail. However, there are some cases where the laminar thermal convection can be reconstructed by assimilating only the velocity along the fluid surface. (2) There is a limit to the reconstruction of thermal convection in the case that the convecting cells are small (∼1000 km for a 50 Myr reconstruction). We propose that (1) is related to the balance of forces due to the thermal buoyancy and viscous stress around the thermal anomalies and (2) is related to how information is preserved (i.e. how the previous thermal structure is maintained in the observable state throughout the convection process). The results enable the use of geological records to estimate time-series of velocity and temperature in Earth's deep interior, even though the records may only contain information from shallow parts of Earth. [ABSTRACT FROM AUTHOR]

The detection of earthquakes is a fundamental prerequisite for seismology and contributes to various research areas, such as forecasting earthquakes and understanding the crust/mantle structure. Recent advances in machine learning technologies have enabled the automatic detection of earthquakes from waveform data. In particular, various state-of-the-art deep-learning methods have been applied to this endeavour. In this study, we proposed and tested a novel phase detection method using deep learning, which is based on a standard convolutional neural network in a new framework. The novelty of the proposed method is its separate explicit learning strategy for global and local representations of waveforms, which enhances its robustness and flexibility. Prior to modelling the proposed method, we identified local representations of the waveform by the multiple clustering of waveforms, in which the data points were optimally partitioned. Based on this result, we considered a global representation and two local representations of the waveform. Subsequently, different phase detection models were trained for each global and local representation. For a new waveform, the overall phase probability was evaluated as a product of the phase probabilities of each model. This additional information on local representations makes the proposed method robust to noise, which is demonstrated by its application to the test data. Furthermore, an application to seismic swarm data demonstrated the robust performance of the proposed method compared with those of other deep learning methods. Finally, in an application to low-frequency earthquakes, we demonstrated the flexibility of the proposed method, which is readily adaptable for the detection of low-frequency earthquakes by retraining only a local model. [ABSTRACT FROM AUTHOR]

Since deep low‐frequency tremors are considered to be associated with large earthquakes that occur adjacently on the same subducting plate interface, it is important to investigate tremors that occurred before the establishment of modern seismograph networks such as the High Sensitivity Seismograph Network (Hi‐net). We propose a deep‐learning solution to detect evidence of tremors in scanned images of paper seismogram records from over 50 years ago. In this study, we fine‐tuned a convolutional neural network (CNN) based on the Residual Network, which was pre‐trained using images of synthetic waveforms from our previous study, using a data set comprised of images generated from real seismic data recorded digitally by Hi‐net to facilitate a supervised analysis. The fine‐tuned CNN was able to predict the presence or absence of tremors in the Hi‐net images with an accuracy of 98.64%. Gradient‐weighted Class Activation Mapping heatmaps created to visualize model predictions indicated that the CNN's ability to detect tremors is not degraded by the presence of teleseisms. Once validated using the Hi‐net images, the CNN was applied to paper seismograms recorded from 1966 to 1977 at the Kumano observatory in southwest Japan, operated by Earthquake Research Institute, The University of Tokyo. The CNN showed potential for detecting tremors in scanned images of paper seismogram records from the past, facilitating downstream tasks such as the creation of new tremor catalogs. However, further training using an augmented data set to control for variables such as inconsistent plotting pen thickness is required to develop a universally applicable model. Plain Language Summary: In 2002, deep low‐frequency tremors, which periodically occur with much smaller amplitudes and longer durations than common earthquakes, were discovered owing to the establishment of dense seismic arrays in Japan. These tremors are considered to be associated with megathrust earthquakes because they occur in deeper regions of plate boundaries compared with common earthquakes. Investigating tremors that occurred before the establishment of dense seismic arrays is important since megathrust earthquakes recur on intervals of 100–200 years, and digital records from modern seismic arrays cover only a fraction of that time. In this study, we developed a convolutional neural network based on the Residual Network (ResNet) that extracts tremors from images of waveforms recorded more than 50 years ago, when seismometers drew waveforms directly on drum‐rolled paper using a pen. The proposed ResNet, which was trained on seismogram images generated from synthetic waveforms and real data recorded digitally by modern seismic arrays, was applied to scanned images of paper seismograms recorded from 1966 to 1977 at an observatory in southwest Japan. A list of tremors was successfully obtained; however, further training using data that accounts for variables such as inconsistent plotting pen thickness is required to develop a universally applicable model. Key Points: Convolutional neural network model for detection of deep low‐frequency tremors from seismogram images is proposedThe model trained with seismogram images converted from real seismic data successfully detects tremorsThe detection performances of the trained model for the paper records at the Kumano observatory in southwest Japan are discussed [ABSTRACT FROM AUTHOR]

Uncovering the distribution of magnitudes and arrival times of aftershocks is a key to comprehending the characteristics of earthquake sequences, which enables us to predict seismic activities and conduct hazard assessments. However, identifying the number of aftershocks immediately after the main shock is practically difficult due to contaminations of arriving seismic waves. To overcome this difficulty, we construct a likelihood based on the detected data, incorporating a detection function to which Gaussian process regression (GPR) is applied. The GPR is capable of estimating not only the parameters of the distribution of aftershocks together with the detection function, but also credible intervals for both the parameters and the detection function. The property that the distributions of both the Gaussian process and aftershocks are exponential functions leads to an efficient Bayesian computational algorithm to estimate hyperparameters. After its validation through numerical tests, the proposed method is retrospectively applied to the catalogue data related to the 2004 Chuetsu earthquake for the early forecasting of the aftershocks. The results show that the proposed method stably and simultaneously estimates distribution parameters and credible intervals, even within t ≤ 3 hr after the main shock. [ABSTRACT FROM AUTHOR]

We propose a method to predict grain growth based on data assimilation by using a four-dimensional variational method (4DVar). When implemented on a multi-phase-field model, the proposed method allows us to calculate the predicted grain structures and uncertainties in them that depend on the quality and quantity of the observational data. We confirm through numerical tests involving synthetic data that the proposed method correctly reproduces the true phase-field assumed in advance. Furthermore, it successfully quantifies uncertainties in the predicted grain structures, where such uncertainty quantifications provide valuable information to optimize the experimental design. [ABSTRACT FROM AUTHOR]

Cellular structures are hydrodynamically interconnected, such that force generation in one location can move distal structures. One example of this phenomenon is cytoplasmic streaming, whereby active forces at the cell cortex induce streaming of the entire cytoplasm. However, it is not known how the spatial distribution and magnitude of these forces move distant objects within the cell. To address this issue, we developed a computational method that used cytoplasm hydrodynamics to infer the spatial distribution of shear stress at the cell cortex induced by active force generators from experimentally obtained flow field of cytoplasmic streaming. By applying this method, we determined the shear-stress distribution that quantitatively reproduces in vivo flow fields in Caenorhabditis elegans embryos and mouse oocytes during meiosis II. Shear stress in mouse oocytes were predicted to localize to a narrower cortical region than that with a high cortical flow velocity and corresponded with the localization of the cortical actin cap. The predicted patterns of pressure gradient in both species were consistent with species-specific cytoplasmic streaming functions. The shear-stress distribution inferred by our method can contribute to the characterization of active force generation driving biological streaming. [ABSTRACT FROM AUTHOR]

Tide gauge observations along the coastline of Japan have recorded the land sinking due to the continuous subduction of the oceanic plates, indicating that stress and strain energies have been accumulating at the plate boundary, which would eventually cause large oceanic trench earthquakes like the 2011 Great East Japan Earthquake. The proposed method extracts such long-term activities of the Earth's crust together with rapid displacements related to earthquakes, even before the establishment of the global positioning system, from monthly mean data of the sea levels. A state space model decomposes the tide gauge time series into trend, seasonal, autoregressive and observation noise components, each of which are estimated using the particle filter algorithm. The spatial and temporal distributions of the extracted trend component clearly indicate high-risk regions, near which giant earthquakes have occurred or are predicted to occur. A multivariate analysis of the observatories located at the northeast coast of Japan successfully determines the past crustal displacement in the case of the 1978 Off-Miyagi Earthquake. The proposed method has the potential application for monitoring crustal activities related to the accumulation of earthquake energy. [ABSTRACT FROM PUBLISHER]

In Japan, ~3 million individuals are estimated to be infected with hepatitis B virus (HBV) or hepatitis C virus (HCV). The rates of hepatitis virus infection amongst dentists is higher than that amongst other healthcare workers due to increased exposure to both saliva and blood. However, an efficient method for the testing of hepatitis virus infections amongst dentists remains to be established. The aim of the present study was to examine the rate of hepatitis virus infection amongst dental healthcare workers (DHWs) by introducing a health checkup that included screening for HBV and HCV infections. A total of 1,834 members of the Dental National Health Insurance Society in the Oita Prefecture, consisting of dentists and other employees, were tested for hepatitis B surface antigen (HBsAg), antibodies to HBsAg (anti-HBs) and antibodies to HCV (anti-HCV) during routine medical checkups. Anonymized data, including the age, sex, occupation (dentist or employee), and presence of a hepatitis virus marker, was collected and analyzed. The positive rates of HBsAg, anti-HBs and anti-HCV in the study sample were 0.6, 44.1 and 0.5%, respectively; the positive rates were higher amongst dentists than the employees. Furthermore, the positive rates of HBsAg and anti-HCV increased with age and were higher in subjects aged 50-79 (1.7-2.2%). The positive rate of presence of anti-HBs was significantly higher in the dentists compared with employees (56.4 vs. 39.6%; respectively; P<0.0001). The three factors associated with anti-HB positivity were HBsAg negativity, occupation (dentist) and age (20-29 years) with adjusted odds ratios of 8.29, 2.27 and 1.59, respectively (P<0.05). These results suggest that introducing a hepatitis virus examination during routine health checkups of DHWs may prove useful in identifying infected individuals. [ABSTRACT FROM AUTHOR]

We propose a Bayesian inference methodology to evaluate unobservable parameters involved in multi-phase-field models with the aim of accurately predicting the observed grain growth, such as in rocks, metals and alloys. This approach integrates models and a set of observational image data of grain structures. Since the set of image data is not a time series, directly applying conventional inference techniques that require time series as the input data is difficult. The key idea in our methodology to overcome this difficulty is to construct a time series with an appropriate statistic that characterizes static image data of grain structures. Our methodology implements the empirical Bayes method. It can estimate not only a probability density function of the parameters but also an initial phase-field, which is generally unobservable in real experiments. After validating the proposed method through numerical tests using synthetic data, we apply it to real experimental images of grain structures in a steel alloy. The proposed method properly estimates unobservable parameters together with their uncertainties, and successfully selects the initial phase-field that best explains the experimental data from among candidate initial phase-fields. [ABSTRACT FROM AUTHOR]

Motivation Biochemical reactions in cells are made of several types of biological circuits. In current systems biology, making differential equation (DE) models simulatable in silico has been an appealing, general approach to uncover a complex world of biochemical reaction dynamics. Despite of a need for simulation-aided studies, our research field has yet provided no clear answers: how to specify kinetic values in models that are difficult to measure from experimental/theoretical analyses on biochemical kinetics. Results: We present a novel non-parametric Bayesian approach to this problem. The key idea lies in the development of a Dirichlet process (DP) prior distribution, called Bayesian experts, which reflects substantive knowledge on reaction mechanisms inherent in given models and experimentally observable kinetic evidences to the subsequent parameter search. The DP prior identifies significant local regions of unknown parameter space before proceeding to the posterior analyses. This article reports that a Bayesian expert-inducing stochastic search can effectively explore unknown parameters of in silico transcription circuits such that solutions of DEs reproduce transcriptomic time course profiles. [ABSTRACT FROM AUTHOR]

Data assimilation (DA) is a fundamental computational technique that integrates numerical simulation models and observation data on the basis of Bayesian statistics. Originally developed for meteorology, especially weather forecasting, DA is now an accepted technique in various scientific fields. One key issue that remains controversial is the implementation of DA in massive simulation models under the constraints of limited computation time and resources. In this paper, we propose an adjoint-based DA method for massive autonomous models that produces optimum estimates and their uncertainties within reasonable computation time and resource constraints. The uncertainties are given as several diagonal elements of an inverse Hessian matrix, which is the covariance matrix of a normal distribution that approximates the target posterior probability density function in the neighborhood of the optimum. Conventional algorithms for deriving the inverse Hessian matrix require O(CN^{2}+N^{3}) computations and O(N^{2}) memory, where N is the number of degrees of freedom of a given autonomous system and C is the number of computations needed to simulate time series of suitable length. The proposed method using a second-order adjoint method allows us to directly evaluate the diagonal elements of the inverse Hessian matrix without computing all of its elements. This drastically reduces the number of computations to O(C) and the amount of memory to O(N) for each diagonal element. The proposed method is validated through numerical tests using a massive two-dimensional Kobayashi phase-field model. We confirm that the proposed method correctly reproduces the parameter and initial state assumed in advance, and successfully evaluates the uncertainty of the parameter. Such information regarding uncertainty is valuable, as it can be used to optimize the design of experiments.

Construction of quantitative models is a primary goal of quantitative biology, which aims to understand cellular and organismal phenomena in a quantitative manner. In this article, we introduce optimization procedures to search for parameters in a quantitative model that can reproduce experimental data. The aim of optimization is to minimize the sum of squared errors (SSE) in a prediction or to maximize likelihood. A (local) maximum of likelihood or (local) minimum of the SSE can efficiently be identified using gradient approaches. Addition of a stochastic process enables us to identify the global maximum/minimum without becoming trapped in local maxima/minima. Sampling approaches take advantage of increasing computational power to test numerous sets of parameters in order to determine the optimum set. By combining Bayesian inference with gradient or sampling approaches, we can estimate both the optimum parameters and the form of the likelihood function related to the parameters. Finally, we introduce four examples of research that utilize parameter optimization to obtain biological insights from quantified data: transcriptional regulation, bacterial chemotaxis, morphogenesis, and cell cycle regulation. With practical knowledge of parameter optimization, cell and developmental biologists can develop realistic models that reproduce their observations and thus, obtain mechanistic insights into phenomena of interest.