Statistics and Probability seminar
Speaker: Dipali Vasudev Mestry, Institute of Chemical Technology, Mumbai
Host: Radhendushka Srivastava
Title: Identifying Ecological Drivers of Decline in the Abundance of Chitala Populations Through Bayesian Analysis of an Intraguild Predation System
Time, day and date: 4:00:00 PM - 5:00:00 PM, Thursday, November 20
Venue: Ramanujan Hall
Abstract: Understanding how populations grow, interact, and respond to environmental pressures is central to ecological science. Mathematical models provide a principled framework for representing these processes, while Bayesian statistical methods offer powerful tools for analysing noisy ecological data, quantifying uncertainty, and formally evaluating competing ecological hypotheses. This study integrates Bayesian modelling, Monte Carlo computation, and sensitivity analysis to investigate ecological dynamics through a data-driven lens. In this study, we analysed the rapid decline of Notopterus chitala, an endangered freshwater fish species widely distributed across South and Southeast Asia. Motivated by field evidence and ecological theory, the study investigates the trophic mechanisms contributing to this decline within a three-species intraguild predation (IGP) system involving Chitala (IG predator), Mugil (IG prey), and shrimp (resource). Two variants of coupled ordinary differential equation (ODE) models were studied: one assuming that Chitala feeds solely on Mugil and shrimp, and another incorporating an alternative food source. Parameter estimation was conducted within a fully Bayesian framework. Posterior distributions of the model parameters wereobtained using the Gibbs sampler, enabling rigorous quantification of parameter uncertainty. To address structural uncertainty arising from competing model formulations, Reversible-Jump Markov Chain Monte Carlo (RJMCMC) was employed for Bayesian model selection and comparison. The results identify two dominant ecological drivers behind the decline of Chitala: (i) a substantial reduction in nutritional gain from Mugil, which forms a major component of its diet, and (ii) diminished predation efficiency on shrimp during periods when Mugil is scarce. These mechanisms limit the species’ ability to compensate for fluctuations in prey availability, thereby accelerating its decline. The findings underscore the ecological importance of restoring Mugil populations as a central component of conservation strategies. More broadly, the study demonstrates the value of Bayesian inference for extracting mechanistic insights from multi-species ecological systems and for informing evidence-based management decisions.

Supervised Learning Project seminar
Speaker: Nirav Bhattad, IIT Bombay
Host: Prachi Mahajan
Title: Rado’s Theorem in Complex Analysis
Time, day and date: 4:00:00 PM – 5:00:00 PM, Thursday, November 20
Venue: Room 114