This branch:
- solves #33 (closed)
- fixes a bug in compute_probabilities() where the time dimension was dropped from the arrays when it was equal to 1.
- fixes a bug where save_metrics() failed if
Corr
was the only metric requested in the recipe.
Re: #33 (closed), replacing NAs with missing values seemed to mess with the continent borders in the observations and calibrated data, so I left those saving functions untouched.