Congratulations to Litty Mathew, a third year PhD student at the School of Computer Science, Trinity College Dublin whose research was recognised with the Best Paper Award at the 7th International Conference on Statistics: Theory and Applications (ICSTA 2025), held in Paris in August. ICSTA is a prestigious international forum that showcases leading advances in statistical theory and applications.
The award-winning paper, “Assessing Ecosystem Dynamics Under Disturbance: A Hidden Markov Model Framework for Species Detection Data”, introduces an innovative method to analyse species behaviour during sudden, short-term environmental changes. Drawing on processed passive acoustic monitoring data, the framework has wide applicability across terrestrial and aquatic ecosystems, providing fresh insights for conservation ecology.
She is funded by the E3 Kinsella Digitising Biodiversity project, further highlighting the School’s role in advancing impactful, cross-disciplinary research.
Litty describes the research as follows: My PhD is part of the E3 'Digitising Biodiversity' project, which aims to develop a biodiversity monitoring system by developing and integrating acoustic, visual, and mm-wave all-weather radar sensors to get reliable data at fine spatial and temporal scales and derive ecological insights. As part of this project, we develop statistical modelling approaches to analyse processed passive acoustic monitoring time-series data, which have statistical challenges including spatio-temporal dependence and non-stationarity. We present a general Gaussian hidden Markov model framework for analysing such datasets from terrestrial or aquatic ecosystems to identify changes in species' behavioural dynamics under sudden and short-term disturbances. To illustrate our framework, we analyse high-resolution bird-vocalisation time series from the Okinawa Environmental Observation Network (OKEON) in Okinawa, Japan, including two consecutive typhoon disturbance events. Increasing frequency and magnitude of sudden and short-term disturbances underscore the relevance of our framework for detecting early warning signals through shifts in community-level vocalisation dynamics.