Data-driven analysis of soil quality indicators using limited data

摘要: Abstract The difficult question of which variables to include as a minimum data set soil quality (SQ) indicators may be simplified by statistical methods, allow working with databases including categorical and numerical commonly used for assessing SQ. aims this study were: i) identify structural related parameters that associate SQ at different geographic areas ii) test the potential power using decision trees in setting up framework assessment, determining properties, visually evaluated, could included estimation physical properties such saturated hydraulic conductivity (Ks). was evaluated visual assessment (VSA) field limited number chemical (bulk density (BD), air capacity, plant available water capacity (PAWC), (Ks), stable aggregates (WSA), particle size distribution, organic carbon (SOC) cation exchange (CEC)) determined laboratory. Using those physical, morphological soils both tropical temperate areas, classification model were grown. Parameters differed between areas. Ks strongest variable ‘tropical’ soils, but WSA, SOC PAWC also key differences For ‘temperate’ only selected tree building algorithm. SOC, clay, CEC discriminating constructed from combined set. Statistically significant relationships measured are promising demonstrating description required merging morphological, indicators. Thresholds predicting better established when frameworks involve VSA. We proved prediction KS more accurate predictor variables, case showed simpler structure compared built properties. In conclusion, encouraging selection Moreover, seems promising. VSA render response other developing (agricultural interest) capable representing dynamic.