| # Persistence, 1 degree | |
| Holds the most recent analysis constant at every lead hour, on a 1 degree | |
| global grid (181 x 360, pole to pole inclusive, longitude 0-360). | |
| Input is a 0.25 degree analysis; the 1 degree grid is a strict subset of it, so | |
| the model selects the collocated points rather than interpolating. No values are | |
| invented. | |
| Output is `temperature_2m` in Kelvin, shaped `(lead, lat, lon)`, at 6-hourly | |
| steps to +168h. | |
| ## Why persistence | |
| Persistence is the standard reference forecast in meteorology. It is trivial to | |
| compute and its error curve is known in advance -- small at short leads, growing | |
| steadily -- which makes it useful as a skill floor and as a check that a scoring | |
| pipeline is behaving. | |
| ## Files | |
| - `predict.py` -- `run(inputs, declaration, init_time)` | |
| - `earthboi.yaml` -- the declaration: variables, lead hours, grid | |
| ## Licence | |
| MIT. | |