This repository contains the Python analysis scripts used in: Inatomi, M.: A robustness diagnostic framework for NMIP ensembles: Application to NMIP2 soil N₂O emission estimates, Geosci. Model Dev., [DOI], [year]. The scripts implement a four-class robustness diagnostic framework (Robust-increase, Robust-decrease, Divergent, Uncertain) and apply it to NMIP2 soil N₂O ensemble outputs. Scripts included: zenodo_a0004_calc_robustness.py — Compute robustness metrics and four-class classification (Fig. 2a) zenodo_a0005_stratify_by_landuse.py — Stratify by land-use type and N input quintiles (Fig. 3, Fig. 4) zenodo_s0001_sensitivity_thresholds.py — Sensitivity analysis of classification thresholds (Table S2) Input data must be prepared independently. See README.md for preprocessing instructions and data sources. The repository enables reproduction of the main figures and tables presented in the manuscript.