TECHNICAL ARTICLES BALANCED ENTROPY INDEX TO CHARACTERIZE SOIL TEXTURE FOR SOIL WATER RETENTION ESTIMATION M. A. Martin 1 , Y. A. Pachepsky 2 , J.-M. Rey 3 , J. Taguas 1 , and W. J. Rawls 4 Soil hydraulic parameters are needed in most projects on transport and fate of pollutants. Pedotransfer procedures are often used to estimate soil hydraulic properties from basic soil data available from soil surveys. Soil particle size distribution, or texture, is known to be a leading soil property affecting soil ability to retain and transmit water and solutes. A substantial effort has been put in searching for texture parameters useful for estimating soil hydraulic properties. Recently a new, entropy-based index has been proposed that characterizes the non-evenness of particle size distributions. This index called balanced entropy has a potential to reflect probable packing of soil particles. Our objective was to see whether the balanced entropy can serve along with other basic soil properties as one of variables-predictors of soil water retention. We computed the balanced entropy for 9871 soil samples in the NRCS soil characterization database and applied the data mining tools to estimate water retention from soil textural composition, organic carbon content, and bulk density. The balanced entropy was the best single predictor and the most important predictor of volumetric water contents at j33 kPa, which are notoriously difficult to estimate. Using the balanced entropy is a promising approach to improve the accuracy of estimated soil hydraulic properties. (Soil Science 2005;170:759–766) Key words: Soil water, pedotransfer functions, balanced entropy, texture, soil water retention. A GRICULTURAL and environmental model- ing and assessment have many uses for soil parameters governing retention and transport of water and chemicals in soils. These properties are notorious for the difficulties and high labor costs involved in measuring them. Often, there is a need to resort to estimating modeling- related soil parameters from other readily avail- able data with empirical equations called pedo- transfer functions (Pachepsky and Rawls, 2004). A need in such estimations exists in modeling at various scales ranging from general atmospheric circulation simulations to the decision support in fine precision agriculture. Particle size distributions (PSD) are used by almost all pedotransfer functions (van Genuchten and Leij, 1992; Wo¨sten et al., 2001). Particle size classes differ among national and interna- tional classifications, and this affects the number and the size of classes used in PTFs (Nemes et al., 1999). Using sand, silt, and clay contents is the most common approach. Some PTF authors take advantage of the presence of additional textural classes in their data sets (Shein et al., 1999; Williams et al., 1992). 759 0038-075X/05/17010-759–766 October 2005 Soil Science Vol. 170, No. 10 Copyright * 2005 by Lippincott Williams & Wilkins, Inc. Printed in U.S.A. 1 Escuela Te´ cnica Superior de Ingenieros Agro´ nomos, Universidad Polite´ cnica de Madrid, 28040 Madrid, Spain. E-mail: miguelangel.martin@upm.es 2 USDA-ARS Environmental Microbial Safety Laboratory, Beltsville, MD. 3 Universidad Complutense, Madrid, Spain. 4 USDA-ARS Hydrology and Remote Sensing Laboratory, Beltsville, MD. Received Sept. 27, 2004; accepted May 23, 2005. DOI: 10.1097/01.ss.0000190507.10804.47 Copyr ight © Lippincott Williams & Wilkins. Unauthor iz ed reproduction of this article is prohibited.