Statistical Consideration In Conducting Landscape Level Soil Studies.

Soil science studies at the systems level pose statistical problems not found in more controlled settings.  System comparison requires measurements of multiple soil properties and the application of combinations of conventional and multivariate statistical analyses.  An example of such a study is the evaluation of land use/management effects on management-dependent soil properties. These studies are essential for assessment of soil quality, C sequestration, and the anthropogenic component of soil change.  We will describe such an experiment from the Southeastern U.S. Coastal Plain (Alabama) and discuss possible data analyses approaches. Issues to be discussed are true vs. pseudo replication, correlated error structures that occur with sampling by depth, and correlated response variables.