Abstract
A series of wind tunnel experiments were conducted to investigate the effect of surface roughness distribution on evaporation processes. In particular, the relationship between non‐homogeneous roughness distribution parameters and effective roughness parameters, which are often required to estimate the evaporation rate over a heterogeneous area, were examined closely. A number of sand ridges having various distributions were placed perpendicular to the main wind direction on a 3 m long lysimeter (1 m in width and 0·6 m in depth) filled with fine sand. The groundwater table in the lysimeter was maintained at a constant level ranging from 0·1 to 0·55 m below the surface. The lysimeter was fitted to the wind tunnel, in which incoming airflow was controlled to have a constant velocity, temperature, and humidity. An internal boundary layer was formed on the sand surface and the mean profiles of wind velocity, humidity, and temperature were observed within this layer. In total, 99 profiles over 33 different surfaces were obtained. The roughness parameters for each surface were evaluated by fitting the mean profile equations to the observed profiles. The following effects of roughness distribution on evaporation and the effective roughness parameters were found. (1) An increase in standard deviation of the roughness distribution, which is the variation in obstacle intervals, caused an increase in the evaporation rate when the average roughness spacing was constant and relatively large. (2) The interrelationships among effective roughness z~m~, z~v~, and z~h~ can be described by the same function that describes that of homogeneous surfaces. (3) Evaporation from rough surfaces was more sensitive to the depth of groundwater than that from the smooth surface within our experimental conditions, because the rough surfaces allow soil water over a broader range of depth to reach the surface. The first two results suggest that the effective roughness parameters are a function of the average and perturbation of roughness distribution on the surface and that the effective z~v~ and z~h~ values can be estimated from effective z~m~ value. Copyright © 2002 John Wiley & Sons, Ltd.