Department of Civil Engineering, Kansas State University, Manhattan, KS, 66506, 913-532-1585
Soil hydraulic properties are needed for modeling water movement in soils. Past attempts at relating hydraulic properties to easily measurable quantities like pore size distribution have relied on empirical and semi-empirical relationships. This study proposes to use Artificial Neural Networks (ANNs) for predicting soil hydraulic properties (hydraulic conductivity, saturated water content and the alpha parameter of Gardner's relationship) from percentages of sand, silt and clay and bulks density of the soil. The backpropagation algorithm is used for this purpose. The performance of ANNs will be compared with some existing models.
soil hydraulic properties, particle size distribution, neural networks, backpropagation algorithm
This paper is from the Proceedings of the 10th Annual Conference on Hazardous Waste Research 1995, published in hard copy and on the Web by the Great Plains/Rocky Mountain Hazardous Substance Research Center.