ILOVECLIM model outputs for ocean oxygen isotopes: PCA, process deconvolution, and model–data comparison

This dataset contains selected outputs from the iLOVECLIM Earth system model for the industrial period. All variables were averaged over the last 100 years of each simulation in order to ensure statistical stability. The dataset includes all variables required to reproduce the figures presented in the associated study. The file Fig2_PCA.csv contains the results of a principal component analysis (PCA) performed on key environmental predictors, including temperature, salinity, oxygen saturation, dissolved oxygen (O₂), dissolved organic carbon (DOC), dissolved inorganic carbon (DIC), slow dissolved organic carbon (DOCs), nutrients (phosphate and nitrate), phytoplankton biomass, net community production (NCP), and remineralization. The file Fig3_deconvolution.csv provides the quantitative contribution of individual processes to δ¹⁸O-O₂ and ¹⁷Δocean anomalies relative to the CTRL reference simulation. Results are reported for three oceanic layers: the global ocean, the surface layer (upper 100 m), and the deep ocean (> 100 m). The file Measurements_compilation.xlsx gathers published observational datasets of δ¹⁸O-O₂ and ¹⁷Δocean compiled from the literature. The files Fig4_Violons.xlsx and Fig6-7_modelled_data_comparison.xlsx contain paired model-data comparisons. Simulated values of δ¹⁸O-O₂ and ¹⁷Δocean were extracted at the same locations (longitude, latitude, depth) as the observational data. Note that data from Hendricks et al. were excluded due to access restrictions. Finally, the file Latin_Hypercube.xslx contains the parameter sets used to generate the 300 sensitivity simulations, based on a Latin hypercube sampling approach.