Analysis and characterization of olive tree cultivation system in Granada province (South of Spain) with optimal scaling and multivariate techniques

A survey of olive grove cultivation in Granada province (South of Spain) was undertaken during the 1999-2000 harvest. The interrelationships between 30 survey variables of nominal, ordinal and continuous nature, including agronomic and soil variables, were analysed using Optimal Scaling and a Principal Component Analysis (PCA) to uncover the structure of the variables in the system and detect the relationships between them. 210 olive groves were surveyed in the study period. In this work, we used the Categorical Principal Component Analysis (CATPCA) as an Optimal Scaling procedure to quantify nominal and ordinal variables, and also, as an exploratory method for selecting the most representative variables of the system. Subsequently, the factorial model was refined by means of a classical PCA. Seven components were extracted, and these accounted for 74% of the total variance. PC1, 2, 5, 6 and 7 are related to agronomic and climatic parameters, and PC3 and 4 are related to soil parameters. PCA confirmed the expert impression of high consistency with farmer's knowledge (agronomic and pedological) reflected in surveys. The Clustering of the soils based on user knowledge gave rise to 3 large clusters or user groups. The latter are differentiated by several user soil parameters: soil depth, hardened layer depth, stoniness and mean slope. Factorial Correspondence Analysis (FCA) showed that farmers discriminated the most contrasted soil map units as Litosols (that farmers perceived as shallow) and Fluvisols and Luvisols (that farmers perceived as flat). This suggests that user knowledge could be used by soil scientists.