Ahuja 1989

You might also like

Download as pdf
Download as pdf
You are on page 1of 8
(0038-078X,/89/1486-0404802.00/0 Sout Serence. December 1989 Vol. 148, No. 6 Printed in US.A. EVALUATION OF SPATIAL DISTRIBUTION OF HYDRAULIC CONDUCTIVITY USING EFFECTIVE POROSITY DATA L. R. AHUJA,' D. K. CASSEL? R. R. BRUCE? anp B. B. BARNES! The use of spatial distribution of effec- tive porosity (¢,) to estimate the distribu- tion of saturated hydraulic conductivity (K,) is evaluated on five new soils and on a combination of soils. The K, is related to 4. by a generalized Kozeny-Carman equa- tion. This equation is then combined with scaling theory to derive the frequency dis- tribution of K, scaling factors from ¢, dis- tribution. The results of fractile diagrams ‘and variances show that the use of the generalized Kozeny-Carman equation, with exponent taken equal to 4 or 5, gives, overall, good estimates of the distribution of K, scaling factors from ¢, measure- ments. The exponent value of 4 is more generally applicable. It is, interesting to find that the above method applies as well across soil types, i.e., when the data of different soils are combined. The empirical equation for combined soils may also be used directly to estimate mean K, over an area, as a first-order approximation. The theories of soil water flow and chemical transport have vital field applications for effi- cient management of resources and mainte- nance of environmental quality. A great deal of effort is being made at present to develop field- scale models of flow and transport throughout the soil-plant-atmosphere continuum to be used as research and management tools (Shaffer and Larson 1987; Lemmon 1986). The technological advancement and efficiency of today’s digital computers have helped make this possible. One obstacle that remains to be overcome for general, application of the models is the nonavailability of soil hydraulic property data for most soils. ‘The determination of soil hydraulic properties is time-consuming and expensive. A large num- ber of measurements are required to characterize ‘Water Quality and Watershed Research Labora- tory, USDA-ARS, Durant, Okla. * Dept. of Soil Science, North Carolina State Univ., Raleigh, N.C. * Southern Piedmont Conservation Research Cen- ter, USDA-ARS, Watkinsville, Ga. Received 7 June 1988; revised 29 Nov. 1988. their spatial and temporal variability and distri- butions, Therefore, it is prudent to develop sim- plified methods to characterize hydraulic prop- erties of field soils. ‘Two complementary approaches may be fol- lowed to achieve a more rapid and less expensive hydraulic characterization of soils. One is to use simpler field methods of measuring soil hy- draulic properties (e.g., Libardi et al. 1980; Chong et al. 1981; Ahuja et al. 1980; Dane and Hruska 1983). The second is to estimate soil hydraulic properties from other easily obtained soil properties and/or limited data (Rawls et al. 1982; Naney et al. 1983; Ahuja et al. 1985). Because hydraulic conductivity at saturation is an important hydraulic property that is difficult to obtain, Ahuja et al. (1984) proposed using the spatial distribution of effective porosity in a soil to estimate the distribution of soil’s saturated hydraulic conductivity. The effective porosity ¢ of a soil was defined as the total porosity minus the volumetric soil water content at 33-kPa suc- tion and was related to the soil’s saturated hy- draulic conductivity (K,) by a generalized Koz- eny-Carman equation. This equation was then combined with scaling theory (Miller and Miller 1956; Warrick et al. 1977) to derive the fre- quency distribution of K, scaling factors from distribution. Comparison of the results with ex- perimental data for three Hawaii soils grouped a set and for an Oklahoma soil appeared promising, but did not cover a sufficient range of soil types to show the applicability of the technique in general. The purpose of this study was to further evaluate the method for five soils representing a broad range of soil physical prop- erties. CONCEPTUAL BACKGROUND We provide here a summary of the conceptual background detailed in Ahuja et al, (1984). The generalized Kozeny-Carman equation used in the analysis is K, = Bor a) where B and n are constants. The exponent n is, assumed to be equal to either 4 or 5; the value 404 DISTRIBUTION OF HYDRAULIC CONDUCTIVITY FROM EFFECTIVE POROSITY 405 of 5 gave slightly better results for the oxisols from Hawaii and the mollisol from Oklahoma. The intercept B may vary with soil type. The jal distribution of K, is characterized in terms of scaling factors based on the concept of similar media (Miller and Miller 1956; Warrick et al. 1977) Ku = Kimo? (2) where K, is the saturated hydraulic conductivity (K,) at a certain location i within a soil, Kim is, the scaled mean hydraulic conductivity over the whole soil, and a; is a scaling factor for location i. Setting the average of a; values for N locations within the soil to be 1.0, Kin is obtained as Kn = (& xe) / xf 3) and then the a, as a = N(Ka)? | & (Ku) ) When Eq. (1) is substituted in Eq. (4), the ‘equation for a; becomes x a = Noi? 2 on? 6) The constant B of Eq, (1) cancels out in Ea. (5), allowing scaling factors for different locations to be obtained without knowing the constant B. ‘A few field measurements of K, preferably at locations having a; values near 1.0, are required to determine an average Kin. MATERIALS AND METHODS ‘The frequency distribution of scaling factors for K, obtained from effective porosity measure- ments using Eq. (5) were compared with actual distribution of scaling factors for K, determined by Eq, (4) initially in three new data sets (four different soils). The soils were Cecil (clayey, kaolinitic, thermic, Typic Hapludults), Lake- land (sandy, thermic, coated, Typic Quartzip- samments), Norfolk (fine-loamy, siliceous, thermic, Typic Paleudults), and Wagram oamy, siliceous, thermic, Arenic Paleudults). ‘The experimental data used for the tests con- sisted of saturated hydraulic conductivity (K.) and soil water retention curves measured on replicated (4 to 10), undisturbed soil cores taken from different soil horizons at several sites for each soil. These data are published in Southern Cooperative Series Bulletins (Bruce et al. 1983; Dane et al. 1983; Quisenberry et al. 1987). For all data used in the present study, the soil cores were 6.0 to 8.5 em in diameter and 6.0 to 7.5 cm in length. The cores were from 10 to 14 depths between 0 and 2 m in Cecil soil, 6 or 7 depths between 0 and 1.63 m in Lakeland, and 7 depths between 0 and 1.52 m in Norfolk and Wagram. Constant-head methods were used to measure K, and hanging-water-column and pressure- plate procedures to measure water retention curves. The water retention at saturation and 33-kPa suction were used to obtain the effective porosity ¢.. For Cecil soil, water retention at saturation was not initially measured, but was calculated using measured soil bulk density and particle density data. The calculated value of the soil water content at saturation may be higher than the experimental saturation value even in soil cores wetted carefully in the labo- ratory, due to the existence of entrapped air. On the other hand, the experimental value of satu- ration may be underestimated if it was measured ata slight suction that may be enough to empty some macropores present in the soil core. The errors in values of the saturated soil water con- tent could cause significant errors in calculation of the effective porosity for fine-textured cores. Of course, the measured values of K, may also be subject to large error and uncertainty because of the above factors. The readers should be aware of these uncertaintis ‘The number of data points available were 69 for Cecil, 79 for Lakeland, 21 for Norfolk, and 20 for Wagram soil. Because the number of data points for Norfolk and Wagram were not enough to obtain a good fractile diagram of scaling fac- tors, the data for these two similar soils were combined into one set. This is justified because Norfolk and Wagram often occur as a soils com- plex in the field. For each data set, log-log plots of K, verous . were made, and fractile diagrams or the cumulative frequency distributions (Hald 1952) of the logarithms of scaling factors were used to compare Eqs. (4) and (5). The data for all four soils of this study and four soils of the previous study (Ahuja et al. 1984) were then combined into one set, and used for the above tests, Combining the data allowed us to examine if the value of constant B in Eq. (1) can be aseumed identical for all soils within the varia- bility of data. 406 After the above anslyses were completed, a large data set of K,-¢, measurements on 180 undisturbed soil cores of Pima clay loam (Typic Torrifluvents) by Coelho was discovered,‘ The 7.6-cm-diameter cores were taken from five soil depths, 30 to 150 cm, and the methods noted above were used to measure K, and water reten- tion curves. This data set, containing 176 non- zero data pairs, was also subject to the above analyses. RESULTS AND DISCUSSION ‘The saturated hydraulic conductivities (K,) for all depths of Cecil soil are plotted as a function of effective porosity (4) in Fig. 1. A -squares line fitted through the log-log data tad lope of 2.58 and a highly significant (P <0.01) correlation coefficient, 7, of 0.824. The K, is strongly related to ¢,, even though other unaccounted factors contribute to its magnitude. ‘Measurement of K, on undisturbed soil cores is commonly subject to a large error, due to un- known effects of entrapped air and the presence of macropore channels, Several replicated-cores, data in our study showed as much as a 25-fold difference in K, values between replicates. The replicated data also showed an error of as much a8 0.04 cm*/cm? in measuring $,. These errors also contribute to the large scatter in the rela- tionship. The slope of the fitted line (n of Eq. (1)) is significantly different from the hypothe- sized value of 4 or 5 statistically (P < 0.01), as indicated by a t test, but the slope as obtained by a log-log regression has been shown to be very sensitive to scatter in the data and some outlier points even though the fractile results were better with n = 4 or 5 (Ahuja et al. 1984). As one of the reviewers of this paper pointed out, an eyeball evaluation of the data in Fig. 1a indicates that most of the data points seem to fall along a line steeper than the regression line. Recomputation proved the regression line shown in Fig. 1a to be correct. Figure 1b compares the cumulative frequency distribution (fractile diagram) of logarithms of sealing factors a obtained from measured K, data using Eq. (4) with the distribution of a values obtained from ¢, values using Eq. (5). For the latter, the results are given for exponent nin Eq. (6) set equal to 4 or 5. The scaling ‘M.A. Coelho, 1974, Spatial variability of water- related soil physical properties, Ph.D. thesis, Univ. of Arizona, Tucson, AHUJA, CASSEL, BRUCE, AND BARNES CECIL SOI, Ky ai8T hE rso8ea"* RMSE (Io9,9K,) +0649 HYDRAULIC CONDUCTIVITY, K .(cm/h) oor 008 Oi 04 EFFECTIVE POROSITY,$¢, (m°/m) Fic. 1a. Saturated hydraulic conductivity as a function of effective porosity for Cecil soil. The symbol ** denotes statistical significance at the 1% level of probability. Ee ara Fe Kp F nb. ig 3 UN 5 Fe © SCALING FACTOR a ons5 cECIL se Oe PROBABILITY VARIABLE ,(Y-YI/oy Fic. 1b, Fractile diagram of the logarithms of scal- ing factors obtained from K, data (Eq. (4)) compared with the fractile diagram for K, factors obtained from 4. data (Eq. (5)) for Cecil soil. The symbol Y in the theoretical probability variable stands for the trans- formed value of a (e.g, log a or others) that would have @ normal distribution. The symbols ¥ and ¢, represent the mean and standard deviation of Y, re- spectively DISTRIBUTION OF HYDRAULIC CONDUCTIVITY FROM EFFECTIVE POROSITY 407 factors calculated from ¢, using n = 4 are closer to the measured K, factors than the factors for n = 5, the root-mean-squares deviations (RMSD) between the calculated and measured factors being 0.179 and 0.393, respectively. The difference between the measured and estimated factors is larger near the lower end of the distri- bution, where the errors in measurements of K, and ¢, and the calculation of probability values are also expected to be large (as much as two orders of magnitude; see Hald 1952; pp. 138- 139). For Lakeland soil, the measured soil K, values are plotted as a function of ¢. in Fig. 2a. The K, values are within a much narrower range com- pared with those of Cecil soil (Fig. 1a). The slope of the fitted line through the log-log data is 3.09 with an r = 0.58, still highly significant statis tically (P< 0.01). The slope value is not signi icantly different from 4 (P > 0.05), but is differ- ent from 5 (P < 0.01). The fractiles (Fig. 2b) indicate, however, that the scaling factors esti- to L NORFOLK SOIL to) 9 WAGRAM SOIL (+), 10-4] zone ~ Krb96age: rs0832** RMSEtioggK . +0378 HYDRAULIC CONDUCTIVITY, K(em/h) 004 or 0204 EFFECTIVE POROSITY, ¢, (m?/m") Fic. 3a, Saturated hydraulic conductivity as a function of effective porosity for Norfolk and Wagram soils, The symbol ** denotes statistical significance at the 1% level of probability. 10 LAKELAND SOIL —x 204054," © rs0.5e*" FMSEWOOgK,) J+ us +019 Kg(cm/h) 3 8 10! HYDRAULIC CONDUCTIVITY, 0.25 03 Oa EFFECTIVE POROSITY, ¢g(m’/m”) 035 Fic. 22, Saturated hydraulic conductivity as & function of effective porosity for Lakeland soil. The symbol ** denotes statistical significance at the 1% level of probability. é 8 Fe aemigi Engh wore ® Bl eel Be ein 5 wa oe sa Fo ee PROBABLITY VARIABLE ,(Y-T)/ry Fic. 2. Comparison of the fractile diagrams of K. sealing factors derived from ¢. and measured K, for Lakeland soil. 0 wae Bos be antic) ZZ ky) eg a ' at g my ose E ee Ds © Sl a, ve a orntei 79 tet 7 § [oe ons wt? NORFOLK @ WAGRAM rr PROBABILITY VARIABLE ,(Y-¥)/ey Fic. 3b, Comparison of the fractile diagrams of K, scaling factors derived from ¢, and measured K, for Norfolk and Wagram soils. mated from ¢., using n equal to either 4 or 5, are essentially identical to the actual sceling factors for measured K,. Using n = 5 provides slightly better estimates (RMSD = 0.016) than ing n = 4 (RMSD = 0.032). ‘The plot of K, versus ¢. for Norfolk and Wagram soils is shown in Fig. 3a. The slope of the log-log relationship is 2.064, with an r = 0.832. The slope value is significantly different from 4 or 5 (P < 0.01). The fractiles (Fig. 3b) show that the distribution of scaling factors estimated from 4, using n = 4 is not adequate, but is closer to the measured K, factor distribu- tion than obtained using n = 5 (RMSD = 0.279 408 HAWAII (3 soled RENFROW 6 10oF cech. + LAKELAND * NORFOLK waoram = En 5 = HYDRAULIC CONDUCTIVITY 0080040 02 0a EFFECTIVE POROSITY, ¢g(m’ mi) Fic. 4a. K, versus 4, relationship for eight differ- ent soils combined. The symbol ** denotes statistical significance at the 1% level of probability. ant Aiba ane ones SCALING FACTOR a 8 SOILS COMBINED a a PROBABILITY VARIABLE ,(Y-Y)/oy Fic. 4b. Comparison of the fractile diagrams of K, scaling factors derived from 4 and measured K, for eight soils combined. AHUJA, CASSEL, BRUCE, AND BARNES and 0.466, respectively). It appears that using n= 3 would appreciably improve the estimates of scaling factors in this soil ‘The combined plot of K, ve the above soils and four soils of the previous study (three Hawé and Renfrow soil; Ahuja et al. 1984) is presented in Fig. 4a, The total number of data points is 297. The slope of the log-log relationship is 3.35, with an r = 0.844. ‘The slope value is significantly different from 4 ‘or 5 (P<0.01). Once again, the position of many of the data points suggests a line steeper than the regression line, but computations show this not to be so. The intercept of the linear regres- sion, log (1058.4), is not significantly different from the intercept of any of the individual soils (P > 0.05), except from that of the Norfolk- Wagram data set (P < 0.01). The combined correlation coefficient is as high as the coeffi- cient for any of the component soils individually. ‘These observations, as well as the fractiles pre- sented below, support the idea that Eq. (1) may be applied across soils, ‘The fractiles for the combined data are shown in Fig. 4b, The fractiles estimated from ¢, using either n = 4 or n = 5 are close to the measured K, fractile. Overall, the fractile for n = 4 is slightly better (RMSD = 0.152) than that for n= 6 (RMSD = 0.242). However, the differences, among the three sets of values are probebly within the errors of measurement of K, and ¢.. It can be inferred from the graph that using n = 3 would result in a fractile that is much worse than the experimental fractile. For Pima clay loam soil, a log-log fit to the data of K, versus 4. had a slope of 2.98, with an 7 = 0.670 (Fig. 5a). The r value is statistically significant (P < 0.01), but the slope is different from 4 or 5 (P < 0.01). The fractile estimated from 4, using n = 5 is closer to the measured K, fractile for the most part than that using n = 4 (Fig. 5b); overall, however, the fractile for n = 4 is slightly closer to the K, fractile (RMSD = 0.079 and 0.084, respectively). In Table 1, we present the mean and variance of the scaling factors (a) and variance of the logarithms of scaling factors determined by each of the three ways—from experimental K, values (a,) and from ¢. values using n = 4 (a) and n= 5 (as) for individual soils and their combi- nations. The mean scaling factor was forced to equal 1.0 in all cases by Eqs. (3) to (5). The fractiles of the logarithms of scaling factors pre- 4. data for all DISTRIBUTION OF HYDRAULIC CONDUCTIVITY FROM EFFECTIVE POROSITY 409 PIMA SOIL Kg 3i.53¢,8" | 10.670** RMSE log) 10.444 Genarot vias tor 8 | HYDRAULIC CONDUCTIVITY, K<(cm/h) oa Oo] 02 34 EFFECTIVE POROSITY,¢, (m’/m") Fic. 5a, Saturated hydraulic conductivity as a function of effective porosity for Pima soil. The sym- bol ** denotes statistical significance at the 1% level of probability. sented above and those presented earlier (Ahuja et al. 1984) indicate that the distributions are close to lognormal. Assuming lognormal distri- butions, we compared the variance of log ay and log ag against the variance of experimental val- ues log a, by an F test. The results are shown in Table 1. For the Norfolk-Wagram soil group, both the variances (log a and log as) were significantly greater than the variance of log a. ‘Among the rest, variance of log a, is not signif- icantly different from that of log a, in all cases except Lakeland. For Lakeland, the variance of log as is statistically the same as that of log a. For the Hawaii group, Renfrow, and Pima soils, the variance of log as is also not significantly different from that of log az, as is the variance of log a. For the combination of soils, the vari- ance of log az is not significantly different from the variance of log a., but the variance of log as is. These results of the variance comparisons agree with the results of fractile comparisons based on RMSD values given above. ‘The collective results presented above show that the generalized Kozeny-Carman equation is generally successful in estimating spatial dis- tribution of K, from effective porosity measure- in several widely different layered soils, individually and in combination. The exponent value of 4 is more widely applicable than the value of 5. Although the slope (n) values of log K, versus log ¢, regressions presented in this paper were all less than 3.35, the fractiles cal- culated from Eq. (5) are generally better with n= 40r5, The slope of the log-log regression is influenced very much by scatter in the data and outlier points, as explained earlier, whereas the fractiles seem to be influenced more by slope in the data from middle to high values of K,, where this slope was greater than the fitted value (see Figs. 1a, 4a, and 5a). It should be noted that we calculated and used only scaling factors for K, for individual loca- tions within a soil or a group of soils, and not the actual K, values at these locations. In theory only one actual measurement of K, at one loca- tion, preferably at the site where « = 1.0, will be needed to find K, values at all locations using their scaling factors and Eq, (2). Such a site can be determined based on a values found from de PIMA SOIL 4b + apn Kg 7EK NeIT6 SCALING FACTOR a BUNS Eb ona ones a PROBABILITY VARIABLE ,(Y-Y)/oy Fic. 5b, Comparison of the fre diagrams of K, sealing factors derived from @, and measured K, for Pima soil. 410 AHUJA, CASSEL, BRUCE, AND BARNES TABLE 1 ‘Mean and variance of Ke scaling factors (a) determined from three sources: experimental Kg values (a), effective porosity values using n = 4 (ay), and effective porosity values using n= 5 (as) Si Novof deta ——MO*__ Red ees <8 8 2 = 4m «== Cecil @ 10 10 10 0925 1.745 2455 om17 0098 0622" Lakeland 7 10-10 10 0.078 0.042 0.066 0.014 0.008" 0012 Norfolk-Wogram 41.1.0 «1.01.0 0.560 1.108 1.544 0.193 0294" 0.459 Hawaii (3 types) 55-10-1010 0586 0301 0445 0.123 0072 O42 Renfrow 5 «10-10-10 0833 0.822 0689 0314 0.304 0475 BSoilscombined 297-10 :L0-1.0 1.162 0.800 1.061 0.399 0.404 0.632" Pima 176 L010 «1.00545 0.286 0439 0.089 0.072 O13. Secilscombined __473_—=—«.0.0.—1.0_0.931_ 0.607 0.828 0.207 0.292 _O46I* “*, Significantly different from the experimental value (P < 0.05), as indicated by the F test. data. In practice more than one measurement ‘would be better. In some cases, such measured values may already be available among a group of soils, Ttis encouraging that one equation of the form of Eq. (1) may apply across many soil types, illustrated in Fig. 4. Such an equation, for example, might be used to determine the K, values at many locations in a watershed having several different soils. For the entire area the mean and variance of K, can be estimated from 4» values, even though there will be considerable error at any one location. To see the general applicability of the regression equation used to describe the K, data in Fig. 4a, we used this equation to calculate K, values from ¢, values for Pima clay loam soil, whose data were not used in the above regression equation. The sults are shown by a broken line in Fig. 58. Thi calculated line is a fair first-order approximation to the experimental data. Similar calculation of K, from ¢, values for a group of coils from South Korea (Dr. Ki Tae Um personal communic tion) also showed a fair agreement with expel mental K, values (Fig. 6). For general applica- tion, we combined the 297 data points of 8 soils shown in Fig. 4a with 176 data points of the Pima clay loam (Fig. 5a) to derive a new regres- sion equation: K, = 764.5 4°, with an r= 0.818 and RMSE (logioK,) = 0.613. Again, the inter- cept of the log-log regression was not signifi- cantly different (P > 0.05) from the intercept of any of the individual soils, except for the Nor- folk-Wagram data, or from the intercept of 8 soils combined (Fig, 4a). The data for Korean soils were not included, because these data were obtained on rather small (5-cm-diameter, 3-cm- long) cores, which we considered less accurate [_ KOREAN Sons 10" F tkwang u,Ulsan and Ta —Fited ratation General relation for 8 other sols Kglemsh) 3 3 HYDRAULIC CONDUCTIVE t! '° 0 Or 10 EFFECTIVE POROSITY, $9 (m'/m) Fic. 6, Saturated hydraulic conductivity as a fune- tion of effective porosity for a group of South Korean shed line represents the regression line of than the other data. This equation should be useful for estimating K, directly over an area for use in resource management models where measured K, data are not available and the direct K, measurements are not feasible. If some DISTRIBUTION OF HYDRAULIC CONDUCTIVITY FROM EFFECTIVE POROSITY 411 direct measurements are available, the estima- tions can be further improved by using the scal- ing-factor approach. REFERENCES Abuja, L. R, RB. Green, 8. K. Chong, and D. R. ‘Nielaen. 1980. A simplified functions approach for determining soil hydraulic conductivities and water characteristics in situ. Water Resour. Res. 16:947-958. Ahuja, L. R., J. W. Naney, R. E. Green, and D. R. ‘Nielsen. 1984. Macroporosity to characterize apa. tial variability of hydraulic conductivity and ef- fects of land management, Soil Sci. Soe. Am. J. 48:699-702 ‘Ahuja, L.R., J. W. Naney, and R. D. Williams. 1985, Estimating soil water characteristics from simpler properties or limited data. Soil Sei. Soc. Am. J. 49:1100-1108. Bruce, R. R., J. H. Dane, V. L. Quisenberry, N. L. Powell, and A. W. Thomas. 1983. Physical char- acteristics of soils in the Southern Region: Cecil Southern Coop. Ser. Bull. 267. Georgia Agric. Exp. Stn., Univ. of Georgia, Athens, Ga. Chong, 8. K., R. E. Green, and L. R. Ahuja, 1981. Simpie in-situ determination of hydraulic con- ductivity by power-function description of drain- age. Water Resour. Res. 17:1109-1114. Dane, J. H., D. K. Cassel, J. M. Davidson, W. L. Pollans, and V. L. Quisenberry. 1983. Physical characteristice of soil in the Southern Regio ‘Troup and Lakeland series. Southern Coop. Ser. Bull. 262. Alabama Agric. Exp. Stn., Auburn, Ala Dane, J. H., and J. Hruska 1983. In situ determination of soil hydraulic properties during drainage. Soil Hald, A. 1952. Statistical theory with engineering ‘applications. Wiley, New York. di, P. L,, K. Reichardt, D. R. Nielsen, and J. W. Biggar. 1980. Simple field methods of estimating soil hydraulic conductivity. Soil Sei. Soc. Am. J. 44:37, Lemmon, H. 1988. Comax: An expert system for cot- ton crop management. Science 233:29-33. Miller, E. E., and R. D. Miller. 1956. Physical theory for capillary flow phenomena. J. Appl. Phys. 27:324-232, Naney, J. W., L. R. Ahuja, and B, B. Barnes. 1983. Variability and interrelation of soil-water and some related soil properties in a small watershed. In Advances in infiltration, Proc. National Con- ference on Advances in Infiltration, Chicago, Ill. 12-13 Dec. 1983. Am. Soc. Agric. Eng., pp. 92- 101. Quisenberry, V. L., D. K. Cassel, J. H. Dane, and J. C, Parker. 1987. Physical characteristics of soils in the Southern Region: Norfolk, Dothan, Golds- boro, Wagram. Southern Coop. Ser. Bull. 263. South Carolina Agric. Exp. Stn., Clemson Univ., Clemson, 8.C. Rawls, W. J., D. L. Brakensiek, and K. E. Saxton. 1982. Estimation of oil water properties. Trans. ‘Am. Soc. Agric. Eng. 25:1316-1320; 1328, Shaffer, M. J., and W. E. Larson (eds.). 1987. NTRM: ‘Asoil-crop simulation model for nitrogen, tillage, ‘and crop residue management. USDA-ARS, Con servation Res. Rept. 34-1. USDA, Agric. Res. Service, Washington, D.C. Warrick, A. W.,G. J. Mullen, and D. R. Nielsen. 197. Scaling of field-measured soil properties. Water Resour. Res. 13:355-362.

You might also like