ArticlesESTIMATING SATURATED HYDRAULIC CONDUCTIVITY AND AIR PERMEABILITY FROM SOIL PHYSICAL PROPERTIES USING STATE-SPACE ANALYSISPoulsen, Tjalfe G.1; Moldrup, Per1; Wendroth, Ole2; Nielsen, Donald R.3Author Information 1Dept. of Environmental Engineering, Institute of Life Sciences, Aalborg University, Sohngaardsholmsvej 57, DK-9000 Aalborg, Denmark. 2ZALF Institute for Soil Landscape Research, Eberswalder Str., 84 D-15374 Müncheberg, Germany. 3Dept. of Land, Air and Water Resources, University of California, Davis, CA 95616 Dr. Poulsen is corresponding author. E-mail: [email protected] Received Sept 3, 2002; accepted Jan. 24, 2003. Soil Science: May 2003 - Volume 168 - Issue 5 - p 311-320 doi: 10.1097/01.ss.0000070906.55992.75 Buy Metrics Abstract Estimates of soil hydraulic conductivity (K) and air permeability (ka) at given soil-water potentials are often used as reference points in constitutive models for K and ka as functions of moisture content and are, therefore, a prerequisite for predicting migration of water, air, and dissolved and gaseous chemicals in the vadose zone. In this study, three modeling approaches were used to identify the dependence of saturated hydraulic conductivity (KS) and air permeability at −100 cm H2O soil-water potential (ka100) on soil physical properties in undisturbed soil: (i) Multiple regression, (ii) ARIMA (autoregressive integrated moving average) modeling, and (iii) State-space modeling. In addition to actual soil property values, ARIMA and state-space models account for effects of spatial correlation in soil properties. Measured data along two 70-m-long transects at a 20-year old constructed field were used. Multiple regression and ARIMA models yielded similar prediction accuracy, whereas state-space models generally gave significantly higher accuracy. State-space modeling suggested KS at a given location could be predicted using nearby values of KS, ka100 and air-filled porosity at −100 cm H2O soil-water potential (ε100). Similarly, ka100 could be predicted from nearby values of ka100 and ε100. Including soil total porosity in the state-space modeling did not improve prediction accuracy. Thus, macro-porosity (ε100) was the key porosity parameter for predicting both KS and ka100 in undisturbed soil. © 2003 Lippincott Williams & Wilkins, Inc.