Prediction of Dispersivity for Undisturbed Soil Columns from Water Retention Parameters E. Perfect,* M. C. Sukop, and G. R. Haszler ABSTRACT ity   D/v (L), that can be related solely to characteris- tics of the porous medium (Fried and Combarnous, Dispersivity () is a required input parameter in solute-transport 1971). Dispersivity is a required input parameter in con- models based on the advection-dispersion equation (ADE). Normally  is obtained from miscible-displacement experiments. This depen- taminant transport models based on the ADE (Zheng dency on inverse procedures imposes a severe limitation on our pre- and Bennett, 1995). Except for a few simple systems such dictive capability. If solute breakthrough curves and soil hydraulic as packed beds of uniformly sized particles,  cannot be properties were measured simultaneously, pedotransfer functions obtained from independent measurements (e.g., Koch could be developed to predict  from independent measurements. In and Flu ¨ hler, 1993). Since most natural porous media are this study, short (6 cm long) undisturbed columns were employed to heterogeneous, soil physicists are currently unable to pre- investigate the relationship between  and the water-retention curve dict solute dispersion in undisturbed soil without first as parameterized by the air-entry value ( a ) and Campbell exponent conducting a miscible-displacement experiment to mea- (b ). We worked with 69 columns from six soil types ranging in texture from loamy sand to silty clay, conventional-till and no-till management sure it. This dependency on inverse procedures imposes practices, steady-state saturated flow conditions, and a step decrease a severe limitation on our predictive capability. in CaCl 2 concentration from 0.009 to 0.001 M. Breakthrough curves If solute breakthrough curves and static physical were measured by monitoring changes in effluent electrical conductiv- properties were determined on the same sample, empiri- ity using a computerized data acquisition system. Estimates of  (cal- cal relations could be developed to predict  from inde- culated using the method of moments) ranged from 1 to 192 mm pendent measurements. However, such studies are re- for the six soil types. Stepwise multiple-regression analysis explained markably rare, and most of them have used packed beds 50% of the total variation in , and indicated that dispersion in- of disturbed media (Passioura and Rose, 1971; Han et creased as  a and b increased. Since both  a and b increase with increasing clay content,  also increases moving from coarse- to fine- al., 1985; Xu and Eckstein, 1997). Under saturated con- textured soils. Our regression equation can be used as a pedotransfer ditions, the magnitude of solute dispersion at any given function to predict  from existing databases of soil hydraulic proper- flow rate is controlled by the pore-space geometry (Per- ties. Further research is needed to independently validate its predic- fect and Sukop, 2001). Since it is the geometrical char- tive capability, and to develop strategies for upscaling the model pre- acteristics of solids or aggregates rather than the pore dictions. space that are measured in the packed bed approach, the resulting relationships are not directly applicable N umerous mathematical models are available for to undisturbed soil. While studies involving artificial macropores (Kanchanasut et al., 1978; Li and Ghodrati, describing solute transport in soil. Of these, the 1997) may provide more information on the relationship ADE is the most widely used. For steady state, one- between solute spreading and pore-space geometry, dimensional water flow, the ADE for a nonreactive they are subject to similar criticisms regarding their ap- solute is given by (Fried and Combarnous, 1971): plicability to natural systems. In terms of heterogeneous systems, Anderson and C t =  2 C Dx 2 - v C x [1] Bouma (1977) observed greater Cl dispersion in undis- turbed soil samples with subangular blocky structure as where C is concentration of solute in the soil water (M compared with those with prismatic structure. Walker L -3 ), t is time (T), x is distance (L), D is the dispersion and Trudgill (1983) reported significant correlations be- coefficient (L 2 T -1 ), and v is the mean pore-water veloc- tween solute transport parameters and several pore- ity (L T -1 ). Equation [1] is normally applied to large geometry variables measured by image analysis of soil volumes of homogeneous soil. Parker and van Genuch- thin sections. Gist et al. (1990) showed that tracer disper- ten (1984) have shown that, with appropriate boundary sion in consolidated rocks was a function of the width conditions, it is able to reproduce the highly asymmetric of the pore-size distribution determined by Hg poro- breakthrough curves obtained from short laboratory simetry. Network model simulations by Bruderer and columns, even when continuous macropores are pres- Bernabe ´ (2001) indicate that for a given flow rate, solute ent. However, the use of Eq. [1] under such conditions dispersion increases logarithmically as the normalized remains a contentious issue (Germann, 1991; Feyen et geometric standard deviation for a log-normal distribu- al., 1998). tion of pores increases. Because D in Eq. [1] depends upon v, it is desirable Several authors have investigated the relationship be- to define an alternative mixing parameter, the dispersiv- tween pore structures revealed by dye-staining patterns and solute-transport parameters. Seyfried and Rao (1987) and Vervoort et al. (1999) report increasing solute dis- E. Perfect, Dep. of Geological Sciences, Univ. of Tennessee, Knox- ville, TN 37996; M.C. Sukop, Dep. of Plants, Soils and Biometeorol- ogy, Utah State Univ., Logan, UT 84322; G.R. Haszler, Dep. of Agron- Abbreviations: ADE, advection-dispersion equation; b, Campbell ex- omy, Univ. of Kentucky, Lexington, KY 40546. Received 24 Apr. ponent; D, dispersive coefficient; v, mean pore-water velocity; , dis- 2001. *Corresponding author (eperfect@utk.edu). persivity; , total porosity;  a , air-entry value; **, significant at the 0.01 probability level. Published in Soil Sci. Soc. Am. J. 66:696–701 (2002). 696