Analysing Extreme Rainfall via a Geometric Framework

arXiv (Cornell University) · 2026-03-18

Motivated by the EVA 2025 Data Challenge, we address the problem of predicting extreme rainfall in the eastern United States using data from a large ensemble of climate model runs. The challenge focuses on three quantities of interest related to the spatial extent and/or temporal duration of extreme rainfall, each requiring extrapolation. To tackle these questions, we adopt the recently developed geometric framework for extreme-value analysis, offering substantial flexibility for capturing complex extremal dependence structures and enabling extrapolation across the entire multivariate tail. In this work, we focus on the spatial geometric framework for analysing the spatial extent and consider a sampling procedure that retains the temporal information in the data, thereby enabling estimation of the duration of extreme rainfall events. We also account for the non-stationary behaviour, arising from topographical and seasonal effects, that commonly characterises extreme weather events in both space and time. Using diagnostic metrics, we demonstrate that the proposed model is appropriate for inferring extreme events on this dataset and apply it to estimate target quantities of interest.

Analysing Extreme Rainfall via a Geometric Framework

ArXiv.org · 2026-03-18

Motivated by the EVA 2025 Data Challenge, we address the problem of predicting extreme rainfall in the eastern United States using data from a large ensemble of climate model runs. The challenge focuses on three quantities of interest related to the spatial extent and/or temporal duration of extreme rainfall, each requiring extrapolation. To tackle these questions, we adopt the recently developed geometric framework for extreme-value analysis, offering substantial flexibility for capturing complex extremal dependence structures and enabling extrapolation across the entire multivariate tail. In this work, we focus on the spatial geometric framework for analysing the spatial extent and consider a sampling procedure that retains the temporal information in the data, thereby enabling estimation of the duration of extreme rainfall events. We also account for the non-stationary behaviour, arising from topographical and seasonal effects, that commonly characterises extreme weather events in both space and time. Using diagnostic metrics, we demonstrate that the proposed model is appropriate for inferring extreme events on this dataset and apply it to estimate target quantities of interest.

Piecewise-linear modeling of multivariate geometric extremes

arXiv (Cornell University) · 2024-12-06

A recent development in extreme value modeling uses the geometry of the dataset to perform inference on the multivariate tail. A key quantity in this inference is the gauge function, whose values define this geometry. Methodology proposed to date for capturing the gauge function either lacks flexibility due to parametric specifications, or relies on complex neural network specifications in dimensions greater than three. We propose a semiparametric gauge function that is piecewise-linear, making it simple to interpret and provides a good approximation for the true underlying gauge function. This linearity also makes optimization tasks computationally inexpensive. The piecewise-linear gauge function can be used to define both a radial and an angular model, allowing for the joint fitting of extremal pseudo-polar coordinates, a key aspect of this geometric framework. We further expand the toolkit for geometric extremal modeling through the estimation of high radial quantiles at given angular values via kernel density estimation. We apply the new methodology to air pollution data, which exhibits a complex extremal dependence structure.

A geotechnical review of the transition from narrow incline undercut to w-undercut design at the Deep Mill Level Zone mine, PT Freeport Indonesia

2024-01-01 · 1 citations

The Deep Mill Level Zone (DMLZ) panel cave mine is located at PT Freeport Indonesia’s operations in Papua, Indonesia. Originally designed and constructed with a narrow incline undercut design with 15 m pillar spacing between parallel drill drives, the DMLZ mine experienced significant issues related to seismicity, damages and rockbursts in undercut and extraction levels during undercutting and production. One of the mitigations identified was to transition the undercut design to a wide-undercut (w-undercut). A w-undercut design doubles the spacing between drill drives to 30 m, which reduces extraction ratios and increases the size of the major apex. A transition from the narrow incline undercut design to a w-undercut layout was initiated through slot blasting towards the existing cave. A critical challenge associated with the transition included the initiating slots which were located near high-stress and rockburst-prone ground. Following the interconnection, a systematic investigation of the ground response in the first four drill drifts of the wundercut layout was undertaken. A substantial improvement in the rock mass behaviour has since been observed in the undercut level, including reduced rockburst frequency and severity as well as an overall decrease in excavation deformation and ground support rehabilitation requirements. This paper discusses the geotechnical review of the narrow incline undercut and w-undercut layout in terms of the ground response due to stress along the cave abutment, seismicity and rockbursts, as well preventative support maintenance.

Monitoring stress-induced brittle rock mass damage for preventative support maintenance

International Journal of Rock Mechanics and Mining Sciences · 2024-10-31 · 1 citations

Integrating Stress Fracturing and Bulking Monitoring for Deformation-Based Ground Support Design Calibration in a Deep Caving Operation

2024-06-23 · 1 citations

ABSTRACT: In a highly stressed environment, massive rock masses fail through stress fracturing, resulting in rock slabs or spalls that can detach from the excavation perimeter either progressively in a non-violent manner as spalling, or suddenly and violently in the form of strainbursting. Both phenomena ultimately lead to rock mass bulking near the excavation boundary that reduces support system capacity, creating safety risks for workers and costly interruptions in production. An innovative and recent approach to designing support in highly stressed brittle rock, known as deformation-based support design (DBSD), offers advantages in addressing the challenges associated with brittle failures around underground excavations. However, the limited availability of purpose-focused field data and systematic procedures for effectively utilizing monitoring techniques to determine DBSD's critical parameters often hinder its optimization, particularly in deep caving operations. This paper presents a systematic method and recommendations to improve design justification and optimization of the DBSD approach for the PT Freeport Indonesia (PTFI) Deep Mill Level Zone (DMLZ) panel cave mine. The study introduces a systematic process that integrates stress fracturing and bulking monitoring techniques, leveraging monitoring data to calibrate the critical parameters of the DBSD. A site-specific predictive function is proposed for estimating the depth of stress fracturing and the corresponding displacement based on the transient nature of the bulking factor across the extraction level footprint. Implementing this procedure into PTFI's DBSD tool effectively improves the forecasting of preventative support maintenance in areas with high demand, thereby promoting the integrity and reliability of the excavation. 1. INTRODUCTION Experiences gained from past and ongoing deep mining and caving operations suggest that brittle failure around underground excavations have been a significant challenge and contributes to hazards in deep underground operations. This phenomenon poses a safety risk for workers and can result in costly interruptions to production. Brittle failure occurs through stress fracturing. Fractures begin to form when induced stresses exceed the crack initiation strength of the rock (Martin, 1997; Kaiser et al., 2000). These fractures will propagate parallel to the maximum compressive stress and open perpendicular to the direction of minimum confinement. This opening mode distinguishes these fractures as extensional fractures, and differentiates them from shear fractures that develop under higher confinements and involve a shearing displacement mode. Thus, near the excavation boundary where confining stresses are low, the failure process is characterized by the formation of extensional fractures growing parallel to the excavation boundary. This, in turn, results in the creation of a set of rock slabs or spalls that can detach from the perimeter of the excavation, known as spalling. As the deviatoric stresses increase relative to the strength of the rock, it progresses deeper into the rock mass. However, as the spalling moves deeper into the rock mass away from the excavation, the higher confining stresses encountered start to suppress and limit the progression of extensional fracturing, resulting in the transition toward the formation of shear fracturing.

Can Custom Models Learn In-Context? An Exploration of Hybrid Architecture Performance on In-Context Learning Tasks

arXiv (Cornell University) · 2024-11-06

In-Context Learning (ICL) is a phenomenon where task learning occurs through a prompt sequence without the necessity of parameter updates. ICL in Multi-Headed Attention (MHA) with absolute positional embedding has been the focus of more study than other sequence model varieties. We examine implications of architectural differences between GPT-2 and LLaMa as well as LlaMa and Mamba. We extend work done by Garg et al. (2022) and Park et al. (2024) to GPT-2/LLaMa hybrid and LLaMa/Mamba hybrid models - examining the interplay between sequence transformation blocks and regressive performance in-context. We note that certain architectural changes cause degraded training efficiency/ICL accuracy by converging to suboptimal predictors or converging slower. We also find certain hybrids showing optimistic performance improvements, informing potential future ICL-focused architecture modifications. Additionally, we propose the "ICL regression score", a scalar metric describing a model's whole performance on a specific task. Compute limitations impose restrictions on our architecture-space, training duration, number of training runs, function class complexity, and benchmark complexity. To foster reproducible and extensible research, we provide a typed, modular, and extensible Python package on which we run all experiments.

Managing excavation closure risk in caving operations

2024-01-01

A framework has been developed to manage production horizon excavation closure risk in caving operations. The focus is on monitoring and response procedures, which are linked to transient geotechnical hazard and remnant reserve value. The study encompasses an analysis of various factors influencing high-closure event risk, including the consequence of asset loss, contributing factors to potential high-closure events, and resisting factors such as rock strength, ground control measures, and rehabilitation work. The framework is supported by empirical benchmarking data and lessons learned from historic cases within PT Freeport Indonesia’s operations and other caving operations. An overview of typical failure mechanisms, best practice monitoring, situation-dependent ground control strategies, and temporary and permanent closure measures (i.e., concrete filling) is provided. The strategies outlined herein offer underground operators a valuable resource for enhancing safety and operational reliability when developing strategies to manage potential high-closure events in critical production areas.

Visual Thinking Strategies (VTS) for Promoting Reflection in Engineering Education: Graduate Student Perceptions

2021 ASEE Virtual Annual Conference Content Access Proceedings · 2024-02-20 · 9 citations

Abstract Visual Thinking Strategies (VTS), an educational technique that uses art to foster visual literacy through facilitated group discussion, has been shown to promote the development of skills that transfer to other domains. In this paper, we report findings from our use of VTS in an experimental graduate course in environmental engineering that aims to foster students' capacities for reflection. Using data from writing samples with methods of thematic analysis, we explore students' perceptions of their own learning from the VTS portion of this semester-long course called Developing Reflective Engineers through Artful Methods. One significant theme identified was "Knowledge/Skills", in which students identified specific knowledge gained or skills developed through their VTS experience, including skills of group discussion, listening/ paraphrasing, observation, imagination/creativity, and critical thinking. Another key theme identified was "Appreciating Others' Perspectives", in which students expressed appreciation of the differences in perspective that VTS discussions tend naturally to draw out. This finding highlights the potential of VTS as a tool for promoting and supporting diversity in engineering. Based on these data and a brief, associated survey, we learned that students found VTS to be highly effective at helping them become more reflective and was one of the most effective methods we have attempted for the development of reflective thinking in graduate engineering.

Work in Progress: Post-Pandemic Opportunities to Re-Engineer Engineering Education: A Pragmatic-Futurist Framework

2021 ASEE Virtual Annual Conference Content Access Proceedings · 2024-02-20 · 1 citations

I am a lifelong transformer.My personal, educational journey