ABSTRACT

Seven empirical models for calculating hydraulic conductivities in soils based on grain-size distribution were investigated in this study. The results were compared with hydraulic conductivity of soils computed using the constant head permeability test. Three samples were collected from three trial pits in different locations along the bank of the stream located downstream of National Root Crops Research Institute's earth dam, Umudike, Abia State Nigeria. The samples were subjected to sieve analysis and the constant head permeability tests using standard methods. Hydraulic conductivities in soils computed from the empirical formulae were each compared with hydraulic conductivity calculated using the constant head formula. Results showed that mean hydraulic conductivities for constant head, Hazen, Breyer, Kozeny-Carman, USBR, Kozeny, Terzaghi and Slitcher models were 18.16 m/d, 35.52 m/d, 34.80 m/d, 30.50 m/d, 25.86 m/d, 19.08 m/d, 15.66 m/d and 10.86 m/d respectively. ANOVA results for pairwise comparison indicated that Kozeny formula gave the best performance with a p-value of 0.78 at 0.05 critical value. This was followed by Terzaghi, USBR and Slitcher with p-values of 0.44, 0.11 and 0.059 respectively, while the Kozeny-Carman, Hazen and Breyer performed poorly with p-values of 0.03, 0.008 and 0.007 respectively. Confirmatory test carried out using the Dunnett simultaneous software package for level mean - control mean, produced an adjusted p-value which was highest at 1.000 for Kozeny model. In all the tests, Kozeny, Terzaghi, Slitcher and USBR performed well with p-values 1.000, 0.923, 0.117, and 0.092 above the critical value of 0.05, while the Breyer, Hazen, and Kozeny-Carman performed poorly with p-values 0.000, 0.000 and 0.004 below the same critical value. They result further showed that Slitcher model is the best for estimation of hydraulic conductivity with root mean square error (RMSE) of 6.78, mean absolute error (MAE) of 5.73, relative error (RE) of 26.71 and deviation time (DT) of 1.46, From the results of the adjusted p-value, Kozeny and Terzaghi were the best at 1.0 and 0.923 respectively while Breyer and Hazen were the worst at 0. There exists high level of inconsistencies in the findings from different researchers and therefore further researches are recommended.