Numerical evaluation of inverse modelling methods for 1D and 3D water infiltration experiments in homogeneous soils

L. Lassabatère (1) , D. Yilmaz (1) , Rafaël Angulo-Jaramillo (2) , J. Soria Ugalde (3) , Isabelle Braud (4) , J. Simunek (5)

1 LCPC/EAU - Division Eau et Environnement
2 ENTPE - École Nationale des Travaux Publics de l'État
3 Universidad de Guanajuato
4 UR HHLY - Hydrologie-Hydraulique
5 University of California

Modelling and understanding water fluxes in the vadose zone are important in regards to water management and require appropriate characterization methods of soil hydraulic properties. The presented work studies three common methods for characterization of soil hydraulic properties based on the inverse modelling of Beerkan water infiltration experiments: the CI method for Cumulative Information method and two BEST methods for Beerkan Estimation Soil pedotransfer methods. These methods estimate the soil sorptivity and saturated hydraulic conductivity by fitting infiltration data using infiltration models. The CI method directly fits the experimental cumulative infiltration to the usual short time expansion of the complete analytical model proposed by Haverkamp et al. (1994). The BEST methods are based on a specific algorithm that splits the experimental curves into two parts; the first part being fitted to the short time expansion and the second part to the long time expansion. To test the methods, several subsets of infiltration data were generated using the complete analytical model for several radii of the disc source and for times ranging from zero to several truncation times. The methods were then applied and the ratio between their estimations and the target values were evaluated to quantify their related accuracy. The results clearly demonstrated that the CI method must be used only to short time infiltration data. Yet, this method is usually used without any truncation of the experimental data, whereas the truncation should certainly be required. The BEST methods proved efficient and robust, provided the steady state was reached at the end of the infiltration experiment and both short and long time data solutions were used. The gain in accuracy of the BEST methods was all the more important when the disc radius was small.

Mots clés: MODELISATION; INFILTRATION; PROPRIETE DU SOL; MILIEU NON SATURE

Domaines: Sciences de l'environnement

L. Lassabatère, D. Yilmaz, Rafaël Angulo-Jaramillo, J. Soria Ugalde, Isabelle Braud, et al.. Numerical evaluation of inverse modelling methods for 1D and 3D water infiltration experiments in homogeneous soils. 19th World Congress of Soil Science, Soil solutions for a changing world, Aug 2010, Brisbane, Australia. 4 p. ⟨hal-00537732⟩