Data and code for: Fate and soil–plant partitioning of fertilizer-derived ¹⁵N and crop performance in tropical maize–forage intercropping systems

ABSTRACT This dataset and accompanying R code support the study “Fate and soil–plant partitioning of fertilizer-derived ¹⁵N and crop performance in tropical maize–forage intercropping systems”. A two-season field experiment conducted in Botucatu, São Paulo, Brazil, during 2017–2018 and 2018–2019 evaluated crop performance and the partitioning of fertilizer-derived nitrogen (N) among maize, forage, litter and soil pools on a clayey Rhodic Hapludox. The experiment used a split-plot randomized complete block design with four replicates. Guinea grass (Megathyrsus maximus), palisade grass (Urochloa brizantha) and ruzigrass (Urochloa ruziziensis) were assigned to main plots, and N rates of 0 and 180 kg N ha⁻¹ were assigned to subplots. Ammonium sulfate was applied at planting (30 kg N ha⁻¹) and at V5 (150 kg N ha⁻¹). Fertilized subplots contained unconfined microplots receiving ¹⁵N-labeled ammonium sulfate enriched at 4.5 and 3.9 atom % ¹⁵N excess in the first and second seasons, respectively. New microplots were established in the second season to trace the new fertilizer application. Maize, forage, surface litter and soil were sampled at maize physiological maturity (R6). Soil sampling covered five depth intervals to 80 cm. The records describe dry matter yield, N accumulation, grain and N harvest indices, maize grain yield, N derived from fertilizer, fertilizer N recovery and the unrecovered fraction. They support comparisons of fertilizer-N partitioning among forage species. Unrecovered N is a balance residual and does not directly measure N loss. Seasonal contrasts are descriptive because maize cultivar, plant population and growing conditions differed between seasons. TECHNICAL INFORMATION The deposit contains the experimental dataset in Experimental-data.xlsx, the statistical analysis script in R-code.R, and usage and methodological documentation in README.md. The Excel workbook contains three worksheets of plain, unformatted values, without styled tables, filters or frozen panes: Metadata: study information, worksheet-to-R object mapping, variable definitions, data types, measurement units, missing-value conventions, data provenance and import examples. Biomass: 48 subplot-season records and 18 columns containing the experimental design factors, dry matter yield of maize grains, stover, shoots, litter and forage, N accumulation in these pools, maize harvest indices and grain yield at 13% moisture. Isotopic: 24 fertilized microplot-season records and 42 columns containing design factors, fertilizer-derived N percentages and masses, recovery in plant, litter and soil pools, total recovery, and unrecovered N. Soil depths are 0–10, 10–20, 20–40, 40–60 and 60–80 cm, together with the 0–80 cm profile total. Data are organized with one variable per column and one observation per row. Variable names match the R script. Forage codes are mm (Guinea grass), ub (palisade grass) and ur (ruzigrass). The season codes first and second represent 2017–2018 and 2018–2019. The variable nrate is expressed in kg N ha⁻¹ and is treated as categorical. Dry matter and grain yield are in Mg ha⁻¹; N masses are in kg N ha⁻¹. Ndff percentages refer to N in each pool, whereas recovery percentages refer to the applied fertilizer N. Missing observations would be represented by blank cells; no experimental values are missing in this release. The Metadata worksheet provides the complete data dictionary. The deposited values include measured and derived responses used as statistical inputs. Instrument-level ¹⁵N measurements, individual row/interrow soil records, soil N concentrations and bulk densities used in the original calculations, and daily weather records are not included. Row and interrow soil positions were subsamples, averaged with equal weights within each depth and microplot. The R script reads the two data worksheets directly from Experimental-data.xlsx. It checks record counts, treatment combinations, cross-sheet identifiers and arithmetic identities, then fits 78 mixed-effects models using nlme. Biomass and other non-isotopic responses include forage species, N rate and their interaction as fixed effects, with random intercepts for block and main plot within block. Isotopic responses include forage species as a fixed effect and block as a random intercept. Models are fitted separately by season using restricted maximum likelihood. Fixed effects are evaluated with sequential F tests at α = 0.10 using unrounded p-values. Significant interactions are followed by simple effects; otherwise significant main effects are compared. Tukey-adjusted pairwise comparisons and compact-letter displays are generated within each comparison family. Cell means, descriptive marginal means, selected simple-effect means, standard errors, pointwise 90% confidence intervals, ANOVA tables, residual diagnostics, fitted models and session information are exported. Eleven Ndff-percentage columns are retained for reuse but are not included in the original set of modeled responses. The code requires readxl, nlme, emmeans, multcomp and multcompView. It was tested with R 4.3.3, readxl 1.4.3, nlme 3.1-164, emmeans 1.10.0, multcomp 1.4-25 and multcompView 0.1-9. No manual working-directory path is embedded in the script. Instructions for installation and execution are provided in README.md. FUNDING This work was undertaken as part of NUCLEUS, a virtual joint centre promoting enhanced N-use efficiency through an integrated soil–plant systems approach in the United Kingdom and Brazil. Support was provided by FAPESP (grant 2015/50305-8), FAPEG (grant 2015-10267001479), FAPEMA (grant RCUK-02771/16), and the Biotechnology and Biological Sciences Research Council (grant BB/N013201/1) under the Newton Fund scheme. Eduardo Mariano and Camila da Silva Grassmann received FAPESP scholarships under grants 2017/02517-1 and 2016/25253-7, respectively.