An observed per-plot frost offset, fitted from 3+ seasons of logged outcomes on that ground, supersedes all modelled climatology for that plot.
Well established · suboptimal if ignored → advise
How it works
Garden-scale microclimate (south wall, frost pocket, raised-bed drainage, urban heat island) lives below 800m resolution and below 80m. No gridded product resolves it. The season log is a direct measurement on the exact ground in question.
The evidence
Evidence: settled without a trial - measurement beats model on the measured point.