Team: Dr Wanchuang Zhu
Rank aggregation can obtain an aggregated ranking list based on multiple ranking lists. Its performance closely relies on equal reliabilities of all the rankers. The problem of discerning reliability of rankers based only on the rank data is of great interest to many practitioners, but has received less attention from researchers. This project aims to develop new methods which can distinguish the reliabilities of rankers. This project expects to simultaneously obtain a consensus ranking list and qualities of rankers.
Paper has been accepted by the Journal of the American Statistical Association.
Team: Dr Roman Marchant, Dr Vincent Chin, Dr Kendra Travaille
Partner: The Minderoo Foundation
There are numerous stock assessment methods specifically designed to deal with data-limited stocks with catch-only data. These models are applied in complex environments with large amounts of missing data in order to better understand the state of fish stocks and ultimately, guide management decisions. New and improved models are frequently proposed, including versions of CMSY, SSCOM, COMSIR and SRA+ among others. This project aims to provide a comprehensive and thorough comparison of catch-only stock assessment methods using three different simulation frameworks (FLR, DLMTool, and Rpath). These simulations generate a wide variety of ground truth data and removes bias in the evaluations arising from the selection of a single simulation technique. By using diverse simulation frameworks, we aim to better quantify and characterise the accuracy of each model as a function of stock characteristics, such as fish life histories, data availability and effort dynamics.
Team: Professor Mark Jessell, Dr Mark Lindsay, Dr Richard Scalzo
Inference of subsurface geological structure is a key part of decision-making in a range of contexts, including mineral exploration, environmental impacts from mining operations, natural hazard assessment, and engineering. Direct probes of geology from borehole logs are sparse and expensive, and must be fused with sensor data and geological prior knowledge to yield an uncertain model of the subsurface. While both geophysical imaging and geological modeling are mature areas, uncertainty-quantified inference of geological parameters or histories from geophysics is still uncommon, and poses interesting methodological challenges.
This research stream focuses on parametric representations of geology, including “implicit” and “kinematic” models that embody simplified descriptions of geological processes, and how to embed them within Bayesian statistical models for uncertainty-quantified inference. Like many inverse problems, the posterior for the resulting model often has complex geometry, with strong correlations, curvature, and multiple modes. Research focuses on how to quickly and efficiently sample these posteriors, and particularly on approximate methods, such as emulators and surrogates, to accelerate sampling.
DARE Team: Associate Professor Willem Vervoort, Dr Richard Scalzo, Dr Ignacio Fuentes
DARE Partners: Alan Turing Institute, WaterSense
Remote sensing time series holds great promise for data-driven management of water resources at catchment scale. This research stream develops new uncertainty-enabled water accounting methodologies, fusing data from the microwave-band Sentinel satellite with in-situ measurements. Techniques used include probabilistic modeling with physical components, Gaussian process mixtures, and deep learning.
An initial paper using Sentinel data to estimate the time-varying capacity of lakes and on-farm reservoirs has been published in ScienceDirect.