Soil saturated hydraulic conductivity assessment from expert evaluation of field characteristics using an ordered logistic regression model
Derechos de accesoopenAccess
MetadataShow full item record
Cita bibliográficaIngelmo, Florencio, J. Molina, Ma, Miguel de Paz, J., Visconti, F. (2011). Soil saturated hydraulic conductivity assessment from expert evaluation of field characteristics using an ordered logistic regression model. Soil & Tillage Research, 115, 27-38.
The knowledge of the soil saturated hydraulic conductivity (K(s)) is essential for irrigation management purposes and for hydrological modelling. Several attempts have been done to estimate K(s) in base of a number of soil parameters. However, a reliable enough model for qualitative K(s) estimation based on the expert assessment of field characteristics had not been developed up to date. Five field characteristics, namely macroporosity (M), stoniness (S), texture (T), compaction (C) and sealing (L), in addition to tillage (G) were carefully assessed according to three classes each, in 202 sites in an agricultural irrigated area in Eastern Mediterranean Spain. After the evaluation of field characteristics, a single ring infiltrometer was used to determine the K(s) value as the solution of the infiltration equation when the steady state was reached. The distribution of the K(s) was assessed and five classes with 10-fold separations in class limits were defined accordingly. The relationships among site characteristics and K(s) were analyzed through a correspondence analysis (CA). Next, an ordered logistic regression model (OLRM) for the prediction of the K(s) class was developed. The CA revealed that, though tightly related, the set of six site characteristics should not be simplified into a smaller set, because each characteristic explains a significantly different aspect of K(s). Consequently, the OLRM was based on the six characteristics, which presented the following order of importance: L>M>G>T>C>S. According to the cross-validation of the OLRM the hit probability for the prediction of the K(s) class attained an average value of 50%, which increased to 63% for the highest class of K(s). Moreover, wrong estimation of the K(s) class exceeded the +/- 1 range only in 3% of sites. Therefore, a reliable enough assessment of K(s) can be based on the expert assessment of field characteristics in combination with an OLRM. (C) 2011 Elsevier B.V. All rights reserved.