Extreme climate events (ECEs) like fires, floods, and heatwaves significantly impact people, communities and the environment and climate change is modifying their occurrence. Generalised Extreme Value (GEV) distributions are used to quantify the magnitude and occurrence of ECEs and guide human system design (e.g. the return value of extreme wind gust sets construction codes at a given location). We train a Machine Learning (ML) model utilising a suite of GEV distributions to determine the sample size required to estimate return values to an arbitrary uncertainty. For the typical GEV parameters for heat and rainfall extremes, accurate estimates of the return values can require a large sample size. If regional climate model projections aim to characterise the magnitude and occurrence of ECEs, a large sample size places substantial computational requirements on the regional climate simulations.