A5063727990- Soil Geostatistics and Mapping
- Atmospheric and Environmental Gas Dynamics
- Geochemistry and Geologic Mapping
- Cryospheric studies and observations
- Spatial and Panel Data Analysis
- Atmospheric Ozone and Climate
- Data Management and Algorithms
- Winter Sports Injuries and Performance
- Gaussian Processes and Bayesian Inference
- Geophysics and Gravity Measurements
- Statistical Methods and Bayesian Inference
- Statistical Methods and Inference
- Methane Hydrates and Related Phenomena
- Bayesian Methods and Mixture Models
- Remote Sensing and LiDAR Applications
- Climate variability and models
- Point processes and geometric inequalities
- Neural Networks and Applications
- Landslides and related hazards
- Atmospheric chemistry and aerosols
- Meteorological Phenomena and Simulations
- Air Quality Monitoring and Forecasting
- Hydrocarbon exploration and reservoir analysis
- Land Use and Ecosystem Services
- Fault Detection and Control Systems
University of Wollongong
2016-2025
Australian Bureau of Statistics
2015-2021
University of Bristol
2013-2019
Brigham Young University
2018
Charles River Laboratories (Netherlands)
2016
At Bristol
2008-2015
University of California, Davis
2015
University of Edinburgh
2012-2013
British Heart Foundation
2012
Queen's Medical Centre
2012
The Gaussian process is an indispensable tool for spatial data analysts. onset of the "big data" era, however, has lead to traditional being computationally infeasible modern data. As such, various alternatives full that are more amenable handling big have been proposed. These methods often exploit low-rank structures and/or multi-core and multi-threaded computing environments facilitate computation. This study provides, first, introductory overview several analyzing large Second, this...
Abstract. Accurate accounting of emissions and removals CO2 is critical for the planning verification emission reduction targets in support Paris Agreement. Here, we present a pilot dataset country-specific net carbon exchange (NCE; fossil plus terrestrial ecosystem fluxes) stock changes aimed at informing countries' budgets. These estimates are based on “top-down” NCE outputs from v10 Orbiting Carbon Observatory (OCO-2) modeling intercomparison project (MIP), wherein an ensemble inverse...
We present spatiotemporal mass balance trends for the Antarctic Ice Sheet from a statistical inversion of satellite altimetry, gravimetry, and elastic‐corrected GPS data period 2003–2013. Our method simultaneously determines annual in ice dynamics, surface anomalies, time‐invariant solution glacio‐isostatic adjustment while remaining largely independent forward models. establish that over 2003–2013, Antarctica has been losing at rate −84 ± 22 Gt yr −1 , with sustained negative mean trend...
Abstract. We present a hierarchical Bayesian method for atmospheric trace gas inversions. This is used to estimate emissions of gases as well "hyper-parameters" that characterize the probability density functions (PDFs) priori and model-measurement covariances. By exploring space "uncertainties in uncertainties", we show results more complete estimation their uncertainties than traditional inversions, which rely heavily on expert judgment. an analysis shows effect including hyper-parameters,...
Modern conflicts are characterized by an ever increasing use of information and sensing technology, resulting in vast amounts high resolution data. Modelling prediction conflict, however, remain challenging tasks due to the heterogeneous dynamic nature data typically available. Here we propose spatiotemporal modelling tools for identification complex underlying processes such as diffusion, relocation, escalation, volatility. Using ideas from statistics, signal processing, ecology, provide a...
Deep neural network models have become ubiquitous in recent years and been applied to nearly all areas of science, engineering, industry. These are particularly useful for data that strong dependencies space (e.g., images) time sequences). Indeed, deep also extensively used by the statistical community model spatial spatiotemporal through, example, use multilevel Bayesian hierarchical Gaussian processes. In this review, we first present an overview traditional machine learning perspectives...
Neural Bayes estimators are neural networks that approximate estimators. They fast, likelihood-free, and amenable to rapid bootstrap-based uncertainty quantification. In this paper, we aim increase the awareness of statisticians relatively new inferential tool, facilitate its adoption by providing user-friendly open-source software. We also give attention ubiquitous problem estimating parameters from replicated data, which address in network setting using permutation-invariant networks....
Statistical models for spatial processes play a central role in analyses of data. Yet, it is the simple, interpretable, and well understood that are routinely employed even though, as revealed through prior posterior predictive checks, these can poorly characterise heterogeneity underlying process interest. Here, we propose new, flexible class spatial-process models, which refer to Bayesian neural networks (SBNNs). An SBNN leverages representational capacity network; tailored setting by...
Abstract In this work we assess the most recent estimates of glacial isostatic adjustment (GIA) for Antarctica, including those from both forward and inverse methods. The assessment is based on a comparison estimated uplift rates with set elastic‐corrected GPS vertical velocities. These have been observed an extensive network computed using data over period 2009–2014. We find systematic underestimations in methods specific regions Antarctica characterized by low mantle viscosities thin...
FRK is an R software package for spatial/spatio-temporal modeling and prediction with large datasets. It facilitates optimal spatial (kriging) on the most commonly used manifolds (in Euclidean space surface of sphere), both spatio-temporal fields. differs from many packages by avoiding stationary isotropic covariance variogram models, instead constructing a random effects (SRE) model fine-resolution discretized domain. The discrete element known as basic areal unit (BAU), whose introduction...
Non-Gaussian spatial and spatio-temporal data are becoming increasingly prevalent, their analysis is needed in a variety of disciplines. FRK an R package for modeling prediction with very large sets that, to date, has only supported linear process models Gaussian models. In this paper, we describe major upgrade that allows non-Gaussian be analyzed generalized mixed model framework. These vastly more general fitted using the Laplace approximation via software TMB. The existing functionality...
Neural Bayes estimators are neural networks that approximate in a fast and likelihood-free manner. Although they appealing to use with spatial models, where estimation is often computational bottleneck, applications have, date, been restricted data collected over regular grid. These also currently dependent on prescribed set of locations, which means the network needs be re-trained for new sets; this renders them impractical many impedes their widespread adoption. In work, we employ graph...
For offshore structures like wind turbines, subsea infrastructure, pipelines, and cables, it is crucial to quantify the properties of seabed sediments at a proposed site. However, data collection costly, so analysis must be made from measurements that are spatially sparse. Adding this challenge, structure exhibits both nonstationarity anisotropy. To address these issues, we propose GeoWarp, hierarchical spatial statistical modeling framework for inferring 3-D geotechnical sediments. GeoWarp...
Advancements in artificial intelligence (AI) and deep learning have led to neural networks being used generate lightning-speed answers complex questions, paint like Monet, or write Proust. Leveraging their computational speed flexibility, are also facilitate fast, likelihood-free statistical inference. However, it is not straightforward use with data that for various reasons incomplete, which precludes many applications. A recently proposed approach remedy this issue inputs an appropriately...
For offshore structures like wind turbines, subsea infrastructure, pipelines, and cables, it is crucial to quantify the properties of seabed sediments at a proposed site. However, data collection costly, so analysis must be made from measurements that are spatially sparse. Adding this challenge, structure exhibits both nonstationarity anisotropy. To address these issues, we propose GeoWarp, hierarchical spatial statistical modeling framework for inferring 3-D geotechnical sediments. GeoWarp...
Spherical regression, where both covariate and response variables are defined on the sphere, is a required form of data analysis in several scientific disciplines, has been subject substantial methodological development recent years. Yet, it remains challenging problem due to complexities involved constructing valid expressive regression models between spherical domains, difficulties quantifying uncertainty estimated maps. To address these challenges, we propose casting as optimal transport...
Abstract Accurate downscaling with uncertainty quantification and its inclusion in fitting biodiversity models to data are essential for accurate, valid inferences predictions. Here, we provide a general framework spatial modelling of that involves environmental covariates. We derive ecological based on spatial‐statistical model accounts change‐of‐support. Through simulation study, demonstrate our statistical provides accurate quantification. With the Monte Carlo samples downscaled...
Measuring sea surface currents (SSC) directly is challenging. Instead, SSC are often inferred from indirect measurements like altimetry. However, altimetry-based methods only provide large-scale (>100 km) geostrophically-balanced velocity estimates of SSC. Here, we present a statistical inversion model to predict fine-scale using remotely sensed temperature (SST) data. Our approach employs Gaussian Process (GP) regression, where the GP informed by two-dimensional tracer transport...
Multivariate geostatistics is based on modelling all covariances between possible combinations of two or more variables at any sets locations in a continuously indexed domain. spatial covariance models need to be built with care, since matrix that derived from such model must nonnegative-definite. In this article, we develop conditional approach for spatial-model construction whose validity conditions are easy check. We start bivariate and go demonstrate the approach’s connection...
Spatial processes with nonstationary and anisotropic covariance structure are often used when modeling, analyzing, predicting complex environmental phenomena. Such may be expressed as ones that have stationary isotropic on a warped spatial domain. However, the warping function is generally difficult to fit not constrained injective, resulting in "space-folding." Here, we propose modeling an injective through composition of multiple elemental functions deep-learning framework. We consider two...
Antarctica is the world's largest fresh-water reservoir, with potential to raise sea levels by about 60 m. An ice sheet contributes sea-level rise (SLR) when its rate of discharge and/or surface melting exceeds accumulation through snowfall. Constraining contribution sheets present-day SLR vital both for coastal development and planning, climate projections. Information on various processes available from several remote sensing data sets, as well in situ such global positioning system data....