Multivariate analysis and visualization of soil quality data for no-till systems.

To evidence the multidimensionality of the soil quality concept, we propose the use of data visualization as a tool for exploratory data analyses, model building, and diagnostics. Our objective was to establish the best edaphic indicators for assessing soil quality in four no-till systems with regard to functioning as a medium for crop production and nutrient cycling across two Illinois locations. The compared situations were no-till corn-soybean rotations including either winter fallowing (C/S) or cover crops of rye (Secale cereale; C-R/S-R), hairy vetch (Vicia villosa; C-R/S-V), or their mixture (C-R/S-VR). The dataset included the variables bulk density (BD), penetration resistance (PR), water aggregate stability (WAS), soil reaction (pH), and the contents of soil organic matter (SOM), total nitrogen (TN), soil nitrates (NO(3)-N), and available phosphorus (P). Interactive data visualization along with canonical discriminant analysis (CDA) allowed us to show that WAS, BD, and the contents of P, TN, and SOM have the greatest potential as soil quality indicators in no-till systems in Illinois. It was more difficult to discriminate among WCC rotations than to separate these from C/S, considerably inflating the error rate associated with CDA. We predict that observations of no-till C/S will be classified correctly 51% of the time, while observations of no-till WCC rotations will be classified correctly 74% of the time. High error rates in CDA underscore the complexity of no-till systems and the need in this area for more long-term studies with larger datasets to increase accuracy to acceptable levels.