How long before a change in soil organic carbon can be detected?

Authors


Pete Smith, tel. +44 01224 272702, fax +44 01224 272703, e-mail: pete.smith@abdn.ac.uk

Abstract

When planning sampling in an experiment where soil organic carbon (SOC) content is expected to change, it is necessary to know how many samples will need to be taken to demonstrate a change in SOC and after how long this change will be detectable. Much has been published on the number of samples required to demonstrate the minimum detectable difference in SOC, but less on how long it takes for this change to be detectable. In this paper, a model of SOC dynamics is used to estimate the minimum time taken for a change in total SOC content to become measurable under different carbon inputs, land uses and soil types.

For free air carbon dioxide enrichment (FACE), and other experiments in which SOC is expected to increase, relationships between the percentage change in C inputs and the time taken to measure a change in SOC are presented, for two levels of sampling intensity corresponding to the maximum that is practically possible in most experiments (∼100 samples) and that used regularly in field experiments (10–20 samples).

In FACE experiments, where C inputs increase by a maximum of about 20–25%, SOC change could be detected with 90% confidence after about 6–10 years if a sampling regime allowing 3% change in background SOC level (probably requiring a very large number of samples) were used, but could not be detected at all if a sampling regime were used that allowed only a 15% change in background SOC to be detected. If increases in C inputs are much below 15%, it might not be possible to detect a change in soil C without an enormous number of samples. Relationships between the change in C inputs and the time taken to measure a change in SOC are robust over a range of soil types and land uses.

The results demonstrate how models of SOC dynamics can be used to complement statistical power analyses for planning when, and how intensively, to sample soils during experiments. An advantage of the modelling approach demonstrated here is that estimates of the minimum time taken for a change in soil carbon to become detectable can be made, even before any detailed soil samples are taken, simply from estimates of the likely increase in carbon inputs to the soil (via expected changes in net primary production).

Ancillary