Table 1. Summary of real-time prediction accuracy, by prediction horizon.

These results are aggregated across all provinces. The R² and 95% PI coverage columns present the overall correlation coefficient and prediction interval coverage. The relative MAE columns show five quantiles of the distribution of province-level relative MAEs comparing the real-time model at the given horizon to a seasonal baseline model at the given horizon: Q₅ (the 5th percentile), Q₂₅ (25th percentile), Q₅₀ (median), Q₇₅ (75th percentile), and Q₉₅ (the 95th percentile). The relative MAEs were calculated as the MAE from the real-time predictions divided by the MAE from the seasonal average predictions, therefore values larger than 1 indicate that the real-time models showed more absolute error on average than the seasonal models.

			relative MAE (real-time vs. seasonal baseline)
horizon (h)	R 2	95% PI coverage	Q 5	Q 25	Q50 (median)	Q 75	Q 95
1	0.89	0.78	0.13	0.25	0.38	0.75	1.26
2	0.83	0.92	0.19	0.30	0.47	0.84	1.68
3	0.73	0.95	0.27	0.38	0.65	0.93	1.79
4	0.59	0.97	0.33	0.48	0.78	1.15	1.91
5	0.48	0.97	0.37	0.57	0.98	1.51	2.14
6	0.41	0.97	0.44	0.73	1.15	1.82	2.69
7	0.36	0.97	0.53	0.90	1.32	2.09	3.24
8	0.34	0.95	0.55	0.96	1.55	2.29	4.07
9	0.33	0.93	0.66	0.97	1.77	2.78	4.88
10	0.33	0.93	0.70	1.07	1.95	3.08	5.45