P = accumulated precipitation during time interval, cm T M U = absolute diurnal temperature, (Table 1) 5 K (Table 1) mean moisture fraction in upper 2.5 cm soil (Table 1) = mean diurnal windspeed, cm/sec (Table 1) or by a relationship containing some combination of these independent variables. A modified model would be possible if significant covariance existed between parameters as might be expected for those factors relating to soil moisture. Subsequent analysis revealed that the independent variables (pt) provided the best fit of the data for total (C,), saltation (C_), and ground creep (C.) processes. Ground creep gave the best overall fit: C. = 7.32 x 10 8prt* (2) (r = .950, F = 131, p < 0.01) where r = correlation coefficient F = Statistic for testing significance of the regression P = probability of obtaining a value of F this large by chance alone A relationship of the form: in (C.) = byin (M) + boln (U) + In (P) + 41n (T), was fitted to obtain estimates for soil moisture and windspeed exponents for an unconstrained model fit. Only windspeed gave a significant partial regression coefficient (bz = 2, p < 0.01), and incorporation of windspeed into the model gave: C, = 2.50 X 107!2prty2 (3) (r = .966, F = 200, p < 0.01). Addition of this term decreased the sum of squared residuals (SSR) from the regression line by 32%. Calculation of the percentage reduction in SSR when eq. 3 ts used to replace eq. 2 was used to determine whether the former model fit the data significantly better. estimating the following statistic: This is inferred by F = [(SSR2~SSR3)/(Do--D3)] / (SSR3/D3), where SSRo, SSR3 = sum of squared residuals for eq. 2, eq. 3, respectively D2, D3 = degrees of freedom for eq. 688 2, eq. 3, respectively.