We develop a new statistical model and an algorithm for estimating any Thermal Comfort Index (TCI) by combining two practical considerations: 1) the climate model estimates are imperfect (noisy) due to model inaccuracies. 2) the climate variables exhibit complex non-linear dependence structure and follow non-Gaussian distributions, which we capture via Statistical Copula modelling. We first show that our model can be interpreted from a statistical perspective as a hierarchical statistical model and is challenging to perform inference of the TCI values due to its mathematical intractability. To overcome this challenging problem, we develop a novel inferential procedure, based on an Importance sampling algorithm to estimate the statistical properties of the TCI.We show that our algorithm provides much lower Mean Squared Error (MSE) of estimating the TCI values than a conventional approach which does not take the uncertainties into account. We then demonstrate the suitability of our algorithm for real-world examples using WRF climate model simulations and the Heat index thermal comfort model. The proposed algorithm can help policymakers make well-informed decisions regarding health and climate-informed urban design.