Spatiotemporal storm surge modeling via computationally efficient bayesian inference with diagrammatic perturbation technique

Radar nowcasting methodologies have evolved from traditional optical flow and extrapolation techniques to advanced deep learning algorithms. However, accurately modeling growth and decay processes remains a significant challenge. This study explores spatio-temporal statistical models inspired by physics-based stochastic partial differential equations (SPDEs). Specifically, the solution to the advection-diffusion PDE is framed as a vector autoregressive process with coloured noise, characterized by non-uniform spectral properties.We investigate the stochastic component using Gaussian Processes (GPs) and Gauss Markov Random Fields (GMRFs), evaluating covariance structures such as exponential, squared exponential, and dynamically weighted covariance and precision matrices. Nowcasts employing state-dependent GPs and GMRFs are assessed over lead times ranging from 15 minutes to 2 hours. The approach is tested on simulated data and UK precipitation events from the Met Office Nimrod system, focusing on a 200 km × 200 km region. Training data spans January 2014 to December 2020, with observational dimensions on the order of 10^4. To enable computationally efficient Bayesian inference, we utilize sparse matrix methods and Laplace approximations.

New York City, New York (NYC), is extremely vulnerable to coastal flooding; thus, verification and improvements in storm surge models are needed in order to protect both life and property. This paper highlights the Stony Brook Storm Surge (SBSS) modeling system. It utilizes surface winds and sea level pressures from the fifth-generation Pennsylvania State University (PSU)-National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) or the Weather Research and Forecasting (WRF) model to drive the Advanced Circulation Model for Coastal Ocean Hydrodynamics (ADCIRC). For this study, the MM5 is utilized at 12-km grid spacing and ADCIRC is run on an unstructured grid down to ~10-m resolution in areas around Long Island and NYC. This paper describes the SBSS and its performance across the NYC region during the 11–12 December 1992 nor'easter and Tropical Storm Floyd on 16–17 September 1999. During the 1992 event, east-northeasterly surface winds of 15–25 m s−1 (30–50 kts) persisted for nearly 24 h, while hurricane-force winds (35–40 m s−1) occurred for a few hours just south of western Long Island. This created a 1.0–1.5-m storm surge around NYC and western Long Island Sound over three tidal cycles. ADCIRC successfully simulated the peak water levels to within ~10%, and it realistically simulated some of the flooding across lower Manhattan. The surface winds for Tropical Storm Floyd were only 5–10 m s−1 weaker than the 1992 event, but no coastal flooding occurred during Floyd, because the storm approached during a low tide. Additional Floyd simulations were completed by shifting the storm's landfall to the spring high tide the previous week, and by doubling the wind speed to mimic a category-1 hurricane. A combination of the spring high tide and a category-1 hurricane scenario during Floyd would have resulted in moderate flooding at several locations around NYC.

Efficient Bayesian inference using adversarial machine learning and low-complexity surrogate models

A sub-space greedy search method for efficient Bayesian Network inference

Efficient Bayesian inference with latent Hamiltonian neural networks in No-U-Turn Sampling

Identifying the burden of disease and its inequality between geographical regions is an important issue to study health priorities. Estimating burden of diseases using statistical models is inevitable especially in the context of rare data availability. To this purpose, the spatio-temporal model can provide a statistically sound approach for explaining the response variable observed over a region and various times. However, there are some methodological challenges in analysis of these complex data. Our primary objective is to provide some remedies to overcome these challenges. Data from nationally representative surveys and systematic reviews have been gathered across contiguous areal units over a period of more than 20 years (1990 - 2013). Generally, observations of areal units are spatially and temporally correlated in such a way that observations closer in space and time tend to be more correlated than observations farther away. It is critical to determine the correlation structure in space-time process which has been observed over a set of irregular regions. Moreover, these data sets are subject to high percentage of missing, including misaligned areal units, areas with small sample size, and may have nonlinear trends over space and time. Furthermore, the Gaussian assumption might be overly restrictive to represent the data. In this setting, the traditional statistical techniques are not appropriate and more flexible and comprehensive methodology is required. Particularly, we focus on approaches that allow extending spatio-temporal models proposed previously in the literature.Since statistical models include both continuous and categorical outcomes, we assume a latent variable framework for describing the underlying structure in mixed outcomes and use a conditionally autoregressive (CAR) prior for the random effects. In addition, we will employ misalignment modeling to combine incompatible areal units between data sources and/or over the years to obtain a unified clear picture of population health status over this period. In order to take parameter uncertainties into account, we pursue a Bayesian sampling-based inference. Hence, a hierarchical Bayes approach is constructed to model the data. The hierarchical structure enables us to "borrow information" from neighboring areal units to improve estimates for areas with missing values and small number of observations. For their general applicability and ease of implementation, the MCMC methods are the most adapted tool to perform Bayesian inference. This study aims to combine different available data sources and produce precise and reliable evidences for Iranian burden of diseases and risk factors and their disparities among geographical regions over time. Providing appropriate statistical methods and models for analyzing the data is undoubtedly crucial to circumvent the problems and obtain satisfactory estimates of model parameters and reach accurate assessment.

In this study, a storm surge model of the Salish Sea was developed to simulate storm surge in Puget Sound using the unstructured-grid, finite-volume community ocean model (FVCOM). The model was based on Pacific Northwest National Laboratory’s existing Puget Sound hydrodynamic model that has been validated extensively in previous studies. To identify representative historical storm events, a systematic approach was developed based on the non-tidal residual (NTR) method. A total of 34 storm surge events between 1980 and 2016 were identified and simulated. To accurately simulate storm surge inside Puget Sound, the model was driven by reanalyzed Climate Forecast System Reanalysis meteorological forcing and observed water levels at the entrance of Strait of Juan de Fuca and southern entrance of Johnstone Strait. Model results for total water level, storm surge, and currents were compared with field measurements. A series of error statistics parameters were also calculated to quantify the model’s skills in simulating storm surge in the Salish Sea. Error statistics of model performance metrics demonstrated that the Salish Sea storm surge model was able to simulate storm surge inside Puget Sound with high accuracy. Storm surge propagation into Puget Sound is a nonlinear process that cannot be simply determined by a static approach for the entire domain. The maximum NTR distribution inside Puget Sound suggests that storm surge tends to be amplified in several sub-basins and inlets of the Puget Sound, such as Bellingham Bay, Hood Canal, south Puget Sound, and the multi-inlet basin behind Agate and Rich Passages. Therefore, future study is necessary to further quantify the storm surge level in these high-risk areas through field measurements and refined model simulations.

Since its introduction in 2001, MrBayes has grown in popularity as a software package for Bayesian phylogenetic inference using Markov chain Monte Carlo (MCMC) methods. With this note, we announce the release of version 3.2, a major upgrade to the latest official release presented in 2003. The new version provides convergence diagnostics and allows multiple analyses to be run in parallel with convergence progress monitored on the fly. The introduction of new proposals and automatic optimization of tuning parameters has improved convergence for many problems. The new version also sports significantly faster likelihood calculations through streaming single-instruction-multiple-data extensions (SSE) and support of the BEAGLE library, allowing likelihood calculations to be delegated to graphics processing units (GPUs) on compatible hardware. Speedup factors range from around 2 with SSE code to more than 50 with BEAGLE for codon problems. Checkpointing across all models allows long runs to be completed even when an analysis is prematurely terminated. New models include relaxed clocks, dating, model averaging across time-reversible substitution models, and support for hard, negative, and partial (backbone) tree constraints. Inference of species trees from gene trees is supported by full incorporation of the Bayesian estimation of species trees (BEST) algorithms. Marginal model likelihoods for Bayes factor tests can be estimated accurately across the entire model space using the stepping stone method. The new version provides more output options than previously, including samples of ancestral states, site rates, site dN/dS rations, branch rates, and node dates. A wide range of statistics on tree parameters can also be output for visualization in FigTree and compatible software.

Accurate estimation of heading direction from optic flow is a crucial aspect of human spatial perception. Previous psychophysical studies have shown that humans are typically biased in their heading estimates, but the reported results are inconsistent. While some studies found that humans generally underestimate heading direction (center bias), others observed the opposite, an overestimation of heading direction (peripheral bias). We conducted three psychophysical experiments showing that these conflicting findings may not reflect inherent differences in heading perception but can be attributed to the different sizes of the response range that participants were allowed to utilize when reporting their estimates. Notably, we show that participants' heading estimates monotonically scale with the size of the response range, leading to underestimation for small and overestimation for large response ranges. Additionally, neither the speed profile of the optic flow pattern nor the response method (mouse vs. keyboard) significantly affected participants' estimates. Furthermore, we introduce a Bayesian heading estimation model that can quantitatively account for participants' heading reports. The model assumes efficient sensory encoding of heading direction according to a prior inferred from human heading discrimination data. In addition, the model assumes a response mapping that linearly scales the perceptual estimate with a scaling factor that monotonically depends on the size of the response range. This simple perception-action model accurately predicts participants' estimates both in terms of mean and variance across all experimental conditions. Our findings underscore that human heading perception follows efficient Bayesian inference; differences in participants reported estimates can be parsimoniously explained as differences in mapping percept to probe response.

The floods associated with the effects of an incoming tropical cyclone have an immense effect in the Philippines, especially with respect to agriculture, industry, livelihood, and public safety. Knowledge of how such storm surge flooding can affect the community is therefore of great importance. In this study, the mechanisms behind Typhoon Haiyan’s anomalous storm surge flooding in 2013, which resulted in more than 6300 casualties and 2.86 billion USD worth of damage in the Philippines, were investigated. The Japan Meteorological Agency (JMA) storm surge model and the FLO-2D flood model were used to simulate Typhoon Haiyan’s storm surge height and the extent of inundation, respectively. The storm surge input data were obtained from JMA typhoon data, and the digital terrain models used were gathered from the airborne interferometric synthetic aperture radar data. The model’s accuracy was also validated using field validation data of the extent of the observed storm surge in affected coastal areas. Topographical analysis of the inundated regions showed the effects of coastal shape, elevation, and position relative to the typhoon’s approach angle on storm surge flow depth and velocity. Storm surge maximum velocity appears to increase as the fluid flows to an increasingly elevated area. Observing fluid velocity in a coastal area with uniform storm surge discharge from all directions also showed that flow velocity tends to increase at the center. Greater flood depths were experienced in areas with lower coastal elevation and not directly located at the coast, compared to higher elevation coastal areas. Greater extents of storm surge flooding are expected in coastal areas that have a concave shape, as fluid is more likely to be dispersed when hitting a convex coast. Extents are likewise observed to be greater in coastal regions that are located perpendicular to the direction of the typhoon. The research also validated the option of using a combination of typhoon and flood models to simulate the inundation flooding caused by extreme weather events.

In the storm surge model, the wind drag coefficient Cd is a critical parameter that has a great influence on the forecast of the storm surge level. In the present study, the effect of various wind drag coefficient parameterizations on the storm surge level is investigated in the Bohai Sea, Yellow Sea and East China Sea for Typhoons 7203 and 7303. Firstly, the impacts of initial values of a and b in the linear expression Cd = (a + b × U10) × 10−3 on the pure storm surge model are evaluated based on the data assimilation method. Results indicate that when a and b (i.e., the wind drag coefficients given by Smith, Wu, Geernaert et al. and Mel et al.) are non-zeros, the performance of the model has little difference, and the result from Wu is slightly better. However, they are superior to the performance of the model adopting zero initial values. Then, we discuss the influences of diverse ways of calculating wind drag coefficients, which are inverted by data assimilation method (including both linear and constant Cd) and given in the form of linear formulas, on simulating pure storm surge level. They show that the data assimilation-based coefficients greatly exceed those of the ordinary coefficient formulas. Moreover, the wind drag coefficient in the linear form is a little better than that in constant form when the data assimilation method is used. Finally, the assessment of the impact of astronomical tides on the storm surge level is conducted, and the simulation demonstrates that the storm surge model, which has the combination of four constituents (M2, S2, K1, O1) and wind drag coefficient inverted by the data assimilation method with the linear Cd, exhibits the best performance.

Development of operational multi-scale storm surge inundated model and application of 2013 typhoon Haiyan

Based on Finite Volume Coastal Ocean Model (FVCOM), this study constructed a numerical model covering the Bohai Sea, Yellow Sea, and East China Sea. National Centers for Environmental Prediction’s Climate Forecast System Reanalysis (NCEP-CFSR) data were used to drive the model to simulate a large storm surge generated by Typhoon Chan-hom. The model was validated by multiple in situ observations of water levels taken at tidal gauge stations. The effects of the nonlinear terms and topography on the modeling of storm surges were then studied. First, the tide–surge interaction during the storm surge process was analyzed. The results show that the tide–surge interaction can suppress the storm surge at the climax of the astronomical tide and benefit the growth of the storm surge at the ebb of the astronomical tide. The tidal constituents M2, S2 and K1 were added to analyze the influences of the amplitudes and periods of tides on the nonlinear reaction. The results indicate that the nonlinear effect will be enhanced by an increased tidal height; additionally, the nonlinear interaction of semidiurnal tides is more significant than that of diurnal tides. The semidiurnal fluctuation that appears near the peak-value time is also attributed to the tide–surge interaction. Moreover, the tide–surge interaction can be influenced by the water depth and position of the area of interest. According to the numerical results, topography has a certain impact on the storm surge: The peak value of the storm surge will decrease with increasing slope. The existence of the Ryukyu Islands reduces the area influenced by the storm surge on the southeast coast of China but expands the high-value storm surge area.

Summary The inferential model (IM) framework provides valid prior-free probabilistic inference by focusing on predicting unobserved auxiliary variables. But, efficient IM-based inference can be challenging when the auxiliary variable is of higher dimension than the parameter. Here we show that features of the auxiliary variable are often fully observed and, in such cases, a simultaneous dimension reduction and information aggregation can be achieved by conditioning. This proposed conditioning strategy leads to efficient IM inference and casts new light on Fisher's notions of sufficiency, conditioning and also Bayesian inference. A differential-equation-driven selection of a conditional association is developed, and validity of the conditional IM is proved under some conditions. For problems that do not admit a conditional IM of the standard form, we propose a more flexible class of conditional IMs based on localization. Examples of local conditional IMs in a bivariate normal model and a normal variance components model are also given.

Bayesian inference is an effective approach for solving statistical learning problems, especially with uncertainty and incompleteness. However, Bayesian inference is a computing-intensive task whose efficiency is physically limited by the bottlenecks of conventional computing platforms. In this paper, a spintronics-based stochastic computing (SC) approach is proposed for efficient Bayesian inference. The inherent stochastic switching behaviors of spintronic devices are exploited to build a stochastic bitstream generator (SBG) for SC with hybrid CMOS/magnetic tunnel junction (MTJ) circuits design. Aiming to improve the inference efficiency, an SBG sharing strategy is leveraged to reduce the required SBG array scale by integrating a switch network between SBG array and SC logic. A device-to-architecture level framework is proposed to evaluate the performance of spintronics-based Bayesian inference system (SPINBIS). Experimental results on data fusion applications have shown that SPINBIS could improve the energy efficiency about 12× than MTJ-based approach with 45% design area overhead and about 26× than FPGA-based approach.