Identification and mapping of associations among soil variables

Assessment and management of soil quality for agricultural land use planning in Northern Ghana is important as a consequence of increasing competition for land among many land uses. It requires information about soil properties that are typically intercorrelated. This study identifies associations among soil variables and determines the effects of land use on the variables. We applied a canonical correlation to investigate many-to-many relationships among soil variables, regression kriging to analyze spatial variability of the canonical variates, and indicator kriging to estimate probabilities of occurrence of natural vegetation and cropland. Three pairs of canonical variates have been identified (P < 0.05): soil organic C-ECEC, clay-pH, and drainage-chroma interactions, accounting for 58% and 73% of the variability in independent and dependent variables, respectively. The first pair was also the most important within natural vegetation and cropland, accounting for at least 23% of the variability. In natural vegetation, the second pair was the clay-pH association; in cropland it was the sand-chroma interaction. This study shows how canonical correlation methods revealed relationships between cultivation and soil variables, whereas geostatistical methods further complement the study of relationships between soil properties and land use.