Application of this technique to the models shows that eq. 3 is a significantly better fit (F = 6.7, p < 0.05) to the data than is eq. 2. Note that eq. 3 has one degree of freedom less than eq. 2 because of the additional change in the unconstrained equation. A regression of mean diurnal temperatures for each time interval with the normalized sinusoidal solar radiation pattern for the western hemisphere present at GZ gave the fit: T = 271.9 + 14. 28q9 (4) (r = .965, F = 187.7, p < 0.01) where T = absolute mean diurnal temperature, °K, and Sgg = (Sin [27 (JD ~ 90)/365] + 1), where JD = Julian day number, 1,2,3,....... 365. The pattern yields a maximum value on a Julian day corresponding to the summer solstice in the western hemisphere. Since the pattern of absolute temperature to the fourth power also produced a similar fit (r = .960, p < 0.01) with radiation pattern, a new model for predicting dust flux was obtained by substitution of Sg for T* in eq. 2: C. = 337.5 PS9q (r = .970, F = 221, p < 0.01), or to replace eq. C, = 1.12 x 10~*PSggU* (rx = .979, F = 323. p < 0.01). 3: (5) (6) The windspeed term reduced the SSR by 30%, and a percentage reduction in SSR by using eq. 6 over eq. 5 was significant (F = 6.05, p < 0.01) in providing a better fit to the dust flux data. A comparison of all models (eqs. 2, 3, 5, and 6) shows declining SSRs of (2,565,603), (1,734,597), (1,579,617), and (1,103,166), respectively. Since eq. 3 fits better than eq. 2, and eq. 6 fits better than eq. 5, then eq. 6 fits significantly better than any of the other models. A plot of observed dust flux for the soil creep compartment C_ is compared with calculated values from eqs. 5, 6 is presented in Figure 1. The explained variation in dust flux in all model equations is primarily explained by precipitation: C. = 550. 8P (r = .940, F = 107, p < 0.01); (7) a discussion of the possible causes for this effect is presented later in this report. Complementary equations for dust flux considering 689