Statistical modelling of analytical and spectral soil measurement errors

Summary

Soil is a vital resource for plants, animals and humans. Over the last decades, human activities have degraded soils worldwide at an alarming rate. Various regional and global initiatives were launched to improve soil quality and promote sustainable land management. Comparable and quality-assessed soil data are needed at an appropriate scale level for these initiatives to succeed. However, laboratory soil data are known to contain measurement errors. Such analytical measurement errors can be caused by a sloppy analyst or indoor environmental conditions that affect the experiment’s outcome. In this thesis, analytical and spectral measurement errors in soil data were quantified, emphasizing the importance of informing data users about these errors before calibrating and validating models. Measurement errors propagate through soil models and can drastically reduce model performance and prediction uncertainty. In this thesis, the sensitivity of widely employed models within the soil science community to analytical and spectral measurement errors was studied.