Ali Akbar Moosavi; Mohammad Omidifard
Abstract
Introduction: Saturated hydraulic conductivity and the other hydraulic properties of soils are essential vital soil attributes that play role in the modeling of hydrological phenomena, designing irrigation-drainage systems, transportation of salts and chemical and biological pollutants within the soil. ...
Read More
Introduction: Saturated hydraulic conductivity and the other hydraulic properties of soils are essential vital soil attributes that play role in the modeling of hydrological phenomena, designing irrigation-drainage systems, transportation of salts and chemical and biological pollutants within the soil. Measurement of these hydraulic properties needs some special instruments, expert technician, and are time consuming and expensive and due to their high temporal and spatial variability, a large number of measurements are needed. Nowadays, prediction of these attributes using the readily available soil data using pedotransfer functions or using the limited measurement with applying the geostatistical approaches has been receiving high attention. The study aimed to determine the spatial variability and prediction of saturated (Ks) and near saturated (Kfs) hydraulic conductivity, the power of Gardner equation (α), sorptivity (S), hydraulic diffusivity (D) and matric flux potential (Фm) of a calcareous soil.
Material and Methods: The study was carried out on the soil series of Daneshkadeh located in the Bajgah Agricultural Experimental Station of Agricultural College, Shiraz University, Shiraz, Iran (1852 m above the mean sea level). This soil series with about 745 ha is a deep yellowish brow calcareous soil with textural classes of loam to clay. In the studied soil series 50 sampling locations with the sampling distances of 16, 8 , and 4 m were selected on the relatively regular sampling design. The saturated hydraulic conductivity (Ks), near saturated hydraulic conductivity (Kfs), the power of Gardner equation (α), sorptivity (S), hydraulic diffusivity (D) and matric flux potential (Фm) of the aforementioned sampling locations was determined using the Single Ring and Droplet methods. After, initial statistical processing, including a normality test of data, trend and stationary analysis of data, the semivariograms of each studied hydraulic attributes were calculated in various directions and their surface semivariograms were also prepared to determine the isotropic or anisotropic behavior of each studied soil attributes. Since all of studied soil hydraulic attributes were isotropic variables, therefore, the omnidirectional semivariograms were calculated and different theoretical models were fitted to them. The best fitted semivariogram models were determined using the determination coefficient, R2, and the residual sum of the square, RSS. The parameters of the best fitted models to the experimental semivariograms were also determined. The prediction of study hydraulic attributes was carried out using the parameters of semivariogram models by applying the ordinary Kriging approach. Predictions were also carried out using the Inverse Distance Weighing approach. The results of predictions were compared to each other using the Jackknifing evaluation approach and the suitable prediction method was determined and zoning was performed using the results of introducing prediction method. All of the semivariogram calculations and modeling, prediction of zoning of study hydraulic attributes were performed using the GS+ 5.1 software packages.
Results and Discussion: Results indicated that all of the studied soil hydraulic attributes belonged to the weak to moderated spatial correlation classes and the spherical model was the best fitted model for their semivariograms (except for Kfs and D that their best semivariogram models were exponential). The sill of all semivariograms ranged between 0.0003 to 0.419 for the S and Kfs, respectively. The nugget effects and the Range parameter of all semivariograms were located between 0.00015 to 0.108 for the S and Фm, and 211 to 6.4 m for Ks and D, respectively. Results also indicated that 3.5 and 50% of total variation of D and Ks was spatially structured and the other was random, respectively. The spatial correlation classes of near saturated soil hydraulic conductivity and soil hydraulic diffusivity were week, whereas, the spatial correlation classes of the other studied soil hydraulic attributes were moderate. Results revealed that the Inverse Distance Weighting method was the most suitable approach for the prediction of all studied soil hydraulic attributes in the present study. Comparison of the calculated statistical evaluation measures (i.e. Determination coefficient, R2, Mean residual error, MRE, mean square error, MSE, Normalized mean square error, NRMSE and geometric mean error ratio, GMER) and the final determined order of precision showed that the most and the least accurate predictions were obtained for Ks and Фm, respectively.
Conclusion: It is suggested in the cases that we need to map the hydraulic attributes or need their quantities in a large number; geostatistical prediction be performed using the limited measurements to reduce the needed time and costs.
