Kravchenko, A. N. and G. P. Robertson. 2015. Statistical challenges in analyses of chamber-based soil CO2 and N2O emissions data. Soil Science Society of America Journal 79:200-211.
Measurements of soil greenhouse gas (GHG) emissions have gained a lot of attention in an effort to potentially increase agriculture’s role in mitigating climate effects. However, it seems not well recognized that the nature of chamber-based GHG data is such that analyses require advanced statistical techniques to fully explore experimental treatment effects. Moreover, for soil GHG data some experimental design approaches can enhance while others can weaken a study’s ability to detect treatment differences. Here we identify and explore the implications of key choices in experimental design and statistical analyses relevant to chamber-based soil GHG studies. In particular, we discuss (i) relative contributions of different sources of random variability in GHG field studies, (ii) relative benefits of increasing the numbers of samples at different replication levels to increase statistical power, and (iii) benefits of accounting for heterogeneous variances and using repeated measures analysis in GHG studies. Emissions data for CO2 and N2O collected from three experimental sites in Michigan demonstrated high spatial and temporal variability for CO2 and N2O fluxes. For both gases the total variability is dominated by small-scale spatiotemporal variability sources, which constituted 55% of the total variability for CO2 and 95% for N2O fluxes. While increasing the number of replicate plots is the main route of rising statistical power, increasing the number of subsamples (chambers and gas samples) per replicate plot can also provide substantial gains. Judicious repeated measures analysis and especially accounting for heterogeneous variances are important strategies for the efficient analysis of chamber-based GHG data.
Associated Treatment Areas:
Methods T1 T2 T3 T4 T5 T6 T7
Download citation to endnote bibtexSign in to download PDF back to index