In terms of wind profiles over 39 surfaces covered with slender, porous and stocky roughness elements respectively at different density that were observed in a blown-sand wind tunnel under wind velocity around 4-20 m·s^-1, aerodynamic roughness length is redefined as value estimated from whole wind velocity profile following the log law rather than that from inertial sub-layer so as to understand further drag effect on airflow and to reduce uncertainty of estimation about aerodynamic roughness. These whole wind profiles with the log law (here called as WWPL) extend from 0.1-0.3 h to the top of boundary layer except profiles over stocky roughness elements in wake flows which extend upward from top of the roughness elements, indicating extension of WWPL in street flows being stable in contrast with extension of inertial sub-layer. The trends of aerodynamic roughness lengths to observed heights are viewed as increase-decrease (probability 70%), decrease (21%) and increase types (9%) at the range around 0.01-1 mm. The trends can be explained as variations of wind velocity gradients with height due to dissipation of airflows' momentum at the bottom and restoration above top of roughness elements. As a result, adoption of roughness length from WWPL resulting from vertical average of wind velocity gradient, is necessary for expressing drag effect of roughness surfaces on airflow. Increase of aerodynamic roughness length from WWPL with roughness elements' density as a power function shows further the index being better indicator of the resistance effect compared to the traditional roughness length. On average, aerodynamic roughness length from WWPL in wake flow being about 1-5 times higher than that from WWPL in street flow, indicates aerodynamic roughness from WWPL in street flow is a better parameter for predication saltation threshold.