Formalise list of BC assessment methods for CLIM-RECAL

[RuthBowyer]

To formalise and discuss with met office, potential list (to be updated) :

• Mean bias error (MBE), mean absolute error (MAR)
• 5th/9th percentile during reference period on a seasonal basis (absolute and fractional difference)
• Correlation (pearson/spearman) preds and obs
• Diagnostic plots (CDF and QQ)
• Root mean square error RMSE between model predictions and observed values
• SD comparison between raw and bias corrected vals during historical period

The key vars we are interested in:

• Seasonal max, av, min - all vars
• Monthly means and SDs
• Days over thresholds (ie heatwave, coldsnaps)

[RuthBowyer]

This approach might be of use (from Jaisal et al., 2021 https://doi.org/10.1029/2021EA001981)

Taylor's diagram is used to analyze the improvement in model performance after bias correction (Taylor, 2001). The Taylor diagram gives a summary about correlation (r) between observed, uncorrected, and bias-corrected model data, together with standard deviation (SD) and centered root mean square difference (RMSE)