07-11, 11:00–11:10 (Europe/Amsterdam), While Loop (4.2)
In this talk, we present QuantileMatching.jl, a package that provides exhaustive high-performance functions to perform quantile matching of an actual distribution and a target distribution with Julia. The functionalities will be illustrated in this talk with stationary examples of climate models and the impacts of bias correction.
Quantile mapping is a statistical technique used in hydrology and climatology to adjust the distribution of observed data to match a desired distribution, typically that of a climate model or historical record. This method involves mapping the quantiles of the observed data to the corresponding quantiles of the reference distribution, ensuring a more accurate representation of extreme events and improving the reliability of climate projections. By aligning the distributions, quantile mapping helps address biases in climate model outputs and enhances the utility of climate information for impact assessments and decision-making.
There are several types of quantile mapping methods, each with its specific approach to adjust the distribution of observed data to a reference distribution. Generally speaking, methods are classified within parametric or non-parametric categories.
In a stationary context, quantile mapping assumes a constant relationship between observed and reference distributions over time.
I am currently enrolled as a graduate student in applied mathematics at Ecole Polytechnique de Montréal in Canada. My research focuses on fitting extreme statistical models to climate models in order to better predict the occurrences of natural hazards like extreme rainfalls or heat waves. Ultimately, my research will benefit governments and engineers so that new infrastructure and urban planning can withstand extreme events. On another level, I also work as a senior software engineer at Coveo, a genAI platform that creates relevant experiences for customers online.