Code and outputs: provenance-gated causally-informed XAI for crop yield prediction (T-PGCE)

Analysis code, simulation code and reference outputs (main and supplementary tables and figures) accompanying the manuscript "A Provenance-Gated Framework for Causally-Informed Explainable AI in Crop Yield Prediction: Identification Diagnostics for Hypothesis-Generating Causal Explanation". The archive contains a single entry point (RUN_ALL.py), nine seeded analysis stages (preprocessing, causal discovery, predictive models, exposure effect, explanations, validation, rating analysis, semi-synthetic simulation, tables and figures), and supplementary checks (clustered missingness test, Moran's I). Third-party input data (CYCleSS, CropClimateX) are not redistributed; see README.md for sources. Anonymized expert ratings are available on reasonable request.