Statistical analysis
All statistical analyses were conducted using PSAW Statistics (version 18.0) and PC-ORD (version 5.32) statistical software. For Experiment I, at each time interval, a Dunnett's test (0.05 P level) was used to determine if the application of glyphosate caused a significant deviation in respiration rate and total CO2-C evolved from the untreated control. Linear regression was used to assess relationships between carbon added as glyphosate and CO2-C evolved. For Experiment II, EL-FAMEs were calculated using both absolute concentration (nanomoles per gram soil) and relative concentration (mole percent of total moles of EL-FAME). Prior to data analysis, EL-FAMEs that were present in fewer than 6.5% of all samples were discarded (McCune and Grace 2002). Forty-six EL-FAMEs were used for analysis.
A Kruskal–Wallis nonparametric one-way analysis of variance was used to evaluate the significance of glyphosate application on each EL-FAME and the EL-FAME taxonomic groupings. Non-metric multidimensional scaling (NMS) was performed using PC-ORD based on Sørensen distance. Two hundred and fifty runs were made with real data and compared to 250 randomized runs. A stability criterion of 0.00001 was used. NMS was performed on both the absolute and relative concentration data sets (nanomoles per gram soil and mole percent of total EL-FAME). Before NMS, the data were transformed using a monotonic square root transformation to improve normality and reduce the coefficient of variation among EL-FAMEs. A general linear model univariate analysis of variance (ANOVA) was used to evaluate the significance of glyphosate application on microbial and exchangeable K.