Soil organic carbon sequestration and tillage systems in the Mediterranean basin: a data mining approach

This study has reviewed 66 long-term experimental comparisons on Soil Organic Carbon (SOC) andtillage systems inMediterranean arable crops (from 15 sites located in Greece, Italy, Morocco and Spain), with the aim to identify the biophysical and agronomic variables most associated with C sequestration rate. Data were organized in a dataset containing basic environmental descriptors (elevation, temperature, rainfall), information on soil tillage system (conventional, minimum, no-tillage), soil attributes (ph, particle size distribution and texture), crop rotation, fertilization, time length of the experiment, initial and final SOC stocks. The collected information were analyzed using a data mining approach including Spearman non-parametric correlations, Principal Component Analysis (PCA), hierarchical clustering and step-wise multiple regression. Tillage, crop rotation, and fertilization were the most significant factors affecting C sequestration rate. Non-parametric correlations reported negative coefficients for initial SOC stock, length of the experiment, mineral fertilization, tillage and production system. C sequestration rate increased significantly under no-tillage. Hierarchical clustering indicates that geographical proximity reflects