The coupled natural convection-radiation heat transfer across cavities is an extremely important research issue commonly encountered in many innovative building envelope applications. Extensive both experimental and numerical studies were conducted to analyse the natural convection flow coupled with radiation heat transfer across vertical and inclined rectangular cavities, proposing accurate empirical or regression correlations according to the particularities of the studied cavity problem. Nevertheless, there is a lack of correlations accurate and adapted enough to model coupled natural convection-radiation across cavities in the vicinity of low emissivity surfaces. The present work aims to cover this subject, through an exhaustive numerical analysis validated and evaluated enough with the use of experimental data. A numerical model is firstly updated to simulate infrared radiation and convective heat transfer across a sophisticated innovative building envelope prototype, while 8 different approximations have been implemented in the model in order to evaluate a variety of available correlations on the subject. Formerly the numerical approach presented here used available experimental data for the same prototype. The air's temperature experimental data served as input to the model boundaries in order to exactly impose the same initial conditions for obtaining the results. Furthermore, experimental data regarding the temperature on the sample's interfaces were used for validating the proposed theoretical model. Satisfactory enough agreements are observed between the theoretically simulated and experimentally measured temperatures and the model is afterwards evaluated to determine the most accurate approximation. Finally, the numerical study presented in this paper concluded in the proposition of 2 new updated correlations to model natural convection across vertical cavities in the vicinity of low emissivity surfaces with the perspective to further support their implementation into Computational Fluid Dynamic (CFD) codes. (C) 2016 Elsevier B.V. All rights reserved.