Differentially private federated crop-yield benchmarking across sovereign agricultural registries: a privacy-utility frontier analysis for climate analytics competition policy

Sovereign agricultural registries hold the crop-yield records that climate-analytics competition policy needs, yet the same records are personal data whose release is constrained by national privacy law. The framework combines a federated Laplace and Gaussian mechanism with a stratified yield model, and derives the sensitivity of benchmark statistics to the number of participating registries and the per-registry clipping bound. The framework is illustrated with a reanalysis of NASA POWER monthly air-temperature reanalysis for Accra, Ghana (2000-2023, 312 monthly observations, mean 26.521 C, warming trend 0.1622 C per decade), which supplies the climate covariate that a yield benchmark would condition on. Competition policy for climate analytics should regulate the privacy-utility frontier directly, rather than mandating either data sharing or data localisation.