Development of Pedotransfer Functions for Estimation of Soil Hydraulic Parameters using Support Vector Machines

1443 Soil Sci. Soc. Am. J. 73:1443-1452 doi:10.2136/sssaj2008.0021 Received 23 Jan. 2008. *Corresponding author ([email protected]). © Soil Science Society of America 677 S. Segoe Rd. Madison WI 53711 USA All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher. A number of hydrological models used to simulate situations ranging from fi eld-scale water fl ow to global climate change rely on numerical techniques that simulate heat, water, and solute fl uxes in the vadose zone. Th e use of fl ow models for variably saturated conditions requires accurate estimates of the hydraulic characteristics that govern water retention and water fl ow in soils (Wösten et al., 2001). Th e hydraulic characteristics of soils vary spatially from one location to another, and are also scale-dependent (Hopmans et al., 2002). Th e temporal variability can also occur as a result of various biological and human activities, such as root-growth, soil management and agricultural practices, or physical processes, such as soil shrinking-swelling, soil crusting, and/or water repellence (Wösten et al., 2001). Th erefore, it is necessary to characterize the hydraulic characteristics of soils while keeping their spatial and temporal variability in perspective. Ideally, it would be best to measure soil hydraulic parameters in the laboratory or fi eld so that the variability in space and time can be suffi ciently characterized. However, this diffi cult task is rarely accomplished because of the signifi cant fi nancial and time investments required for such measurements. Additionally, the spatial variability of soil hydraulic properties, their scale-dependency and possibly large modeling domains can make such characterization diffi cult. To circumvent these practical diffi culties with direct methods, researchers have shown keen interest in developing indirect methods beginning as early as 1912 (Briggs and Shantz, 1912). Th e basic idea behind these indirect methods, oft en called pedotransfer functions (PTFs aft er Bouma and van Lanen, 1987), is to predict hard-to-measure soil hydraulic properties (such as retention parameters and hydraulic conductivities) using easily obtainable input information (such as soil texture, bulk density, and particle-size distribution). In essence, PTFs represent predictive functions that translate the data we have (input) into the data we need (output) (Wösten et al., 2001) Navin K. C. Twarakavi* Auburn Univ. 201 Funchess Hall Auburn, AL 36849