Transformation of categorical field soil morphological properties into numerical properties for the study of chronosequences

During field soil surveys various soil morphological properties such as texture, structure, consistence, cutans and color are evaluated. However, these soil properties are described using nominal or ordinal variables. This makes it difficult to assess these variables objectively using numerical methods such as clustering or chronofunctions. Recently, principal component analysis (PCA) with non-linear optimal scaling transformation of the nominal and ordinal variables has been employed. We used this approach on the morphological properties of a Quaternary soil chronosequence from the fluvial terraces of the River Guadalquivir (Southern Spain) between the towns of Andújar and Villanueva de la Reina. The study comprised five pedons and 35 horizons (including the parent material from the current course of the river) in the following order: Haplic Palexeralf (1st terrace), Calcic Haploxeralf (2nd terrace), Haplic Palexeralf (3rd terrace), Typic Calcixerept (4th Terrace) and Typic Xerofluvent (Floodplain). A total of 17 field morphological variables were evaluated and transformed. The nominal variables were: structure type, texture, cutan morphology, hue dry and hue moist. The ordinal variables were: structure size, structure grade, consistence when moist, consistence when dry, stickiness, plasticity, cutan thickness, cutan frequency, value dry, value moist, chroma dry and chroma moist. The scaling obtained for variables showed a systematic trend with the soil age sequence, growing numerically with the expression of the property. The most important principal component of the non-linear PCA, PC1 (56% of the variance) was closely associated with the soil processes of illuviation of clay and rubefaction, which dominate the preholocenic soils of the chronosequence, and, consequently, was positively correlated with increase in hue, chroma, consistency when moist, structure grade and presence of cutans. The optimal scalings obtained from the morphological properties permitted the application of statistical methods such as cluster analysis, which grouped the five pedons and the parent material (PM) of the chronosequence in strict chronological order. Finally, we obtained chronofunctions of the transformed morphological variables which fitted the logarithmic models satisfactorily.