اندازه گيري و تخمين ويژگي هاي هيدروليكي خاك با نفوذ سنج مكشي و شبكه هاي عصبي مصنوعي در مقياس حوضه ي آبخيز

چكيده انگليسي :

134 Measuring and Predicting Soil Hydraulic Properties with Tension Infiltrometer and Artificial Neural Network in Watershed Scale Hamid kelishadi hamidgolshadi@yahoo com 19 September 2012 Department of Soil Science College of Agriculture Isfahan University of Technology Isfahan 84156 83111 IranDegree M Sc Language FarsiSupervisor s Assoc Prof Mohammad Reza Mosaddeghi mosaddeghi@basu ac ir Prof Mohammad Ali Hajabbasi hajabbas@cc ac iut irAbstractNear saturated soil hydraulic properties are needed to study and to model water andcontaminant transport processes in the vadose zones Soil management and land use viatheir effects on soil properties such as texture bulk density structure and organic carboncould indirectly change soil hydraulic properties However measurements of soil hydraulicproperties in the field and lab are time consuming and costly Moreover the results mightnot be reliable due to high spatial and temporal variabilities of soil physical and hydraulicproperties Attempts have been made to predict these properties indirectly using basic andeasily available soil data by pedotransfer functions PTFs In recent years use of artificialneural networks ANNs has becomes common for deriving PTFs The objective of thisstudy was to measured and predicted near saturated hydraulic properties in pasture andarable lands of Farsan and Koohrang cities in Chaharmahal va Bakhtiari province Themajor land uses in the area were pasture dryland farming irrigated farming and fallow Unsaturated water infiltration was measured at consecutive inlet suctions of 15 10 5 and 2cm using a tension infiltrometer at 100 locations Then soil water retention was measuredin the lab at suctions of 0 5 10 15 20 25 30 40 50 60 70 80 90 100 330 500 2000and 5000 cm on undisturbed soil sample which was taken from the soil under the disc ofinfiltrometer Saturated hydraulic conductivity was also measured on the same sampleusing constant head method Easily available soil properties of texture organic matter andcarbonate contents sodium adsorption ratio dry bulk density at field water condition air dry and dry bulk density at suction of 2 cm were measured The infiltration data wasmodeled using Wooding 1968 analytical method and the best fit values of Gardner 1958 parameters were calculated The van Genuchten 1980 parameters were predictedusing the cumulative infiltration data by DISC software and using the lab water retentiondata by RETC software The field and lab methods for measuring and predicting soilhydraulic properties were compared using the paired t test Regression equations forpredicting soil hydraulic properties and parameters were derived using stepwise scheme For the neural network analysis feed forward back propagation network with Marquardt Levenberg training function and TANSIG transfer function was used for predicting soilhydraulic parameters Sensitivity analysis in the neural network analysis was done byStatSoft method Land use effect on soil properties was investigated using GLM methodand LSD mean comparison Results showed that lab measured saturated hydraulicconductivity and water content were significantly greater than those predicted using thefield data Predicted shape n and scale partameters of van Genuchten 1980 modelusing the field data were significantly greater than those predicted using the lab data too Unsaturated hydraulic conductivity values predicted by Wooding s analytical analysis weresignificantly greater than those predicted by DISC numerical analysis The PTFs of theliterature were not able to predict soil hydraulic parameters accurately Neural networkspredicted soil hydraulic parameters better than linear and non linear multiple regressions Sensitivity analysis of neural network identified relative compaction and dry bulk density

كليشادي، حميد

اندازه گيري و تخمين ويژگي هاي هيدروليكي خاك با نفوذ سنج مكشي و شبكه هاي عصبي مصنوعي در مقياس حوضه ي آبخيز

هدايت هيدروليكي غير اشباع , توابع انتقالي خاك , كاربري زمين , طول درشت مويينگي , قابليت جذب