Data and R code for: Radiocarbon evidence that tree mycorrhizal type modulates mean soil carbon age across depths but not organic matter fractions

Abstract

This record contains the data and R code for the paper "Radiocarbon evidence that tree mycorrhizal type modulates mean soil carbon age across depths but not organic matter fractions".Soils were sampled in June 2021 in Yale-Myers Forest, Connecticut, USA. Each of three stands contained one arbuscular mycorrhizal (AM)-dominated plot and one ectomycorrhizal (EcM)-dominated plot. Each plot contained two subplots, one with and one without the ericoid mycorrhizal shrub. In each subplot, three horizons were sampled: the Oa; the surface mineral horizon, 0–10 cm below the Oa; and the subsurface mineral horizon, to 30 cm cumulative depth including the Oa. Each sample was separated into particulate (POM, >53 µm) and mineral-associated (MAOM, <53 µm) organic matter. This gives 72 fraction-level observations from 36 soil samples.For each fraction: Δ14C, δ13C, δ15N, C and N concentrations (within the fraction and on a whole-soil basis), and C:N ratio. For each bulk soil sample: pH and 14-day C mineralisation. For each plot: tree basal area and the share of AM and EcM trees. For each subplot: soil moisture and C-stock-weighted whole-profile Δ14C. The data also include summary statistics of radiocarbon analytical uncertainty, standards and blanks. data_dictionary.csv defines every column, with units and derivations.run_all.R runs the full workflow in about 1 minute and writes every table, figure, fitted model and log to output/. It reproduces data-based figures and tables in the paper and its supplementary information. The deposited output/ was produced with R 4.4.1, and the workflow was also checked with R 4.3.3. README.md describes the sampling design, models, file layout and how to check a run. We used the large language model Claude to assist with code workflow editing during revisions; all code, analyses, and outputs were reviewed by the authors.