Easy access to DEMs and the development of technology treating DEMs make it easier to extract stream longitudinal profiles from DEMs than previously done. Since such profiles possess many problems such as artificial flats and steps, it should be required for them to be smoothed like natural profiles to estimate gradient values along those sections. However smoothing itself comes with much distortion of raw profile from original DEMs. There has been no research evaluating quantitatively the effects due to smoothing process. Here we attempt to quantify the effects of major smoothing methods on raw and real profiles, suggest a new method to overcome the limitations of them, and evaluate it. This study not only suggests a new smoothing method, but also provides a guideline for choosing a proper smoothing method.