In an effort to improve accuracy of current aircraft ice accretion prediction tools, experimental and analytical studies have been conducted on airfoils roughened by natural ice accretion. Surface roughness introduced by ice accretion and its effect on surface convective heat transfer have been tested and modeled, based on 10 experimental test cases. A novel scaling coefficient relating the Stanton and the Reynolds number (CSR) was introduced for heat transfer comparison and modeling in turbulent regime. By coupling the ice roughness and heat transfer models together with LEWICE ice accretion tool, an improved ice accretion model has been achieved. Four experimental ice shapes were obtained at the AERTS laboratory for model validation. The new surface roughness model had very good agreement in both overall ice shape and ice thickness at the stagnation line (within 5% discrepancy for four experimental cases) whereas LEWICE prediction constantly under-estimate the stagnation ice thickness by 30%. Over-prediction of ice horn lengths were also addressed by the proposed model. In one of the in glaze to rime regime cases, LEWICE over-predicted the upper and lower horn length by 32% and 22% respectively, whereas the new model prediction resulted in ±3% accuracy.