Articles | Volume 13, issue 13
https://doi.org/10.5194/bg-13-3863-2016
https://doi.org/10.5194/bg-13-3863-2016
Technical note
 | 
06 Jul 2016
Technical note |  | 06 Jul 2016

Technical note: A bootstrapped LOESS regression approach for comparing soil depth profiles

Aidan M. Keith, Peter A. Henrys, Rebecca L. Rowe, and Niall P. McNamara

Abstract. Understanding the consequences of different land uses for the soil system is important to make better informed decisions based on sustainability. The ability to assess change in soil properties, throughout the soil profile, is a critical step in this process. We present an approach to examine differences in soil depth profiles between land uses using bootstrapped LOESS regressions (BLRs). This non-parametric approach is data-driven, unconstrained by distributional model parameters and provides the ability to determine significant effects of land use at specific locations down a soil profile. We demonstrate an example of the BLR approach using data from a study examining the impacts of bioenergy land use change on soil organic carbon (SOC). While this straightforward non-parametric approach may be most useful in comparing SOC profiles between land uses, it can be applied to any soil property which has been measured at satisfactory resolution down the soil profile. It is hoped that further studies of land use and land management, based on new or existing data, can make use of this approach to examine differences in soil profiles.

Download
Short summary
It is important to be able to assess whether changes in land use or management affect soil properties, but testing differences between soil profiles can be challenging. We developed a straightforward approach that makes use of soil core data (which are not spatially independent) to compare soil properties at specific locations down a depth profile. We hope that this approach can be used more widely, based on new or existing data, to examine the impacts of land use and management on soils.
Altmetrics
Final-revised paper
Preprint