A stream's temperature is a major factor in its ability to support fish life and to be utilized for other beneficial purposes. The approach most generally used for stream temperature prediction is the Energy-Budget method, which involves the inventory of all the energy entering and leaving the stream. A temperature prediction study on the coast fork of the Willamette River was conducted in 1963 using this method. Subsequent analysis of the data revealed a poor correlation between predicted and measured stream temperatures at night. Due to the relative magnitudes of the terms in the Energy-Budget, it became evident that the term for evaporative heat loss was in error. Since the evaporative heat loss can be computed directly from a value of evaporation rate, one can state that an accurate determination of evaporation rate is essential to the successful analysis of any Energy-Budget Equation applied to streams. Several types of mass transfer equations, both theoretical and experimental, have been developed for predicting evaporation from lakes and reservoirs. There has not yet been developed, . however, a method which relates specific stream parameters, such as surface configuration and stream turbulence, to evaporation. Therefore, as an initial step in the search for needed knowledge, this thesis undertook as its purpose the evaluation of the effect of surface configuration on the magnitude of evaporation from a modeled stream surface. There are three types of mass transfer equations which are commonly used for the prediction of evaporation. The most common type is a general equation which employs Dalton's relationship between evaporation and vapor pressure differences together with a correction for wind velocity. The other two types (the discontinuous and continuous mixing approaches) are based upon the structure of the turbulent boundary layer above the evaporation surface. The equations of Norris and of Thornthwaite and Holzman, based on the discontinuous mixing approach, and the equations of Sutton, based on the continuous mixing approach, were evaluated using experimental data to establish their utility as tools for predicting stream evaporation. The research involved the measuring of evaporation rates from porous stream models of five surface configurations placed in a low velocity wind tunnel and the measurement of physical parameters which are included in various evaporation equations. Several test conditions were used for each surface configuration. The analysis of the data had two parts. First, the evaporation rates computed from several existing equations were compared to measured evaporation, and second, an effort was made to establish a relationship between surface configuration and evaporation. Several statistical tests were used in these analyses, and the following conclusions were reached: 1. Equations based on the structure of the overlying air may be used to predict evaporation rates, without any specific knowledge concerning the stream surface configuration. 2. Of the three theoretical methods tested, Sutton's method gave the best correlation between computed and measured evaporation rates. The method of Norris was next, followed by the equation of Thornthwaite and Holzman. 3. The mass transfer coefficient, B', in an experimentally developed equation of the form E = B' u(e[subscript s] - e[subscript a]) is related to model surface configuration. Using wave steepness, H/L, as a characteristic of model surface configuration, B' increases as H/L increases. Therefore, when an equation of this form is used, evaporation rate increases as wave steepness increases. 4. The failure during the 1963 study made on the coast fork of the Willamette River to compute accurate stream temperatures during nighttime periods can be attributed, at least in part, to errors in underestimating the evaporation rates. 5. Surface configuration does affect evaporation rates from streams, but its full quantitative evaluation awaits further research.