Parameterization and identification of the best-fit model for the soil water retention curve with multivariate analysis

The soil water retention curve (SWRC) assesses soil water flow. This study used soil-depth and texture-specific point data to select the best-fit SWRC model using multivariate statistical criteria, improving accuracy for different Brazilian soils. Models evaluated included van Genuchten (with Mualem or Burdine restrictions), Dexter, Ross, and Groenevelt and Grant. We analysed average volumetric water content and dispersion of experimentally measured point data. Principal component analysis (PCA) assessed model fit, validated by linear discriminant and cluster analyses. Statistical indicators included correlation, determination, index of agreement, root mean square error (ERMS), ERMS-to-standard deviation ratio, and normalized ERMS (ENRMS). Using experimental point data is important, since this resulted in higher ENRMS values but offered more reliable SWRC characterization due to wider confidence intervals. PCA and discriminant analysis identified ENRMS as the most effective fit criterion. Cluster analysis classified ENRMS performance as “very good” (0% to <18%), “good” (18% up to 35%), and “satisfactory” (35% up to 70%).