Table 6.

Comparison results after and before the saturation: Algorithm 2 with Poisson for f and identity for g, $t_{\mathrm{s}}=0$, $t_{\mathrm{i}}=5$, $t_{\mathrm{c}}=84$, $t_{\mathrm{e}}=108$ and $h=1$. We denote by w: the metric with saturation and wo: the metric without saturation, in the test period. Ratios of w/wo are reported. Note that, for the metrics such as RMSE, MASE, and MARE, a ratio lower than 1 implies an improvement of the model to make good forecasts with saturation. A ratio of $R_{\mathrm{COR}}^2$ higher than 1 indicates that applying the saturation improves the forecasts.

	RMSE w/wo	MASE w/wo	MARE w/wo	${\mathit{R}}_{\mathbf{COR}}^2$ w/wo
Dakar	0.95	0.97	0.96	1.01
Fatick	1	1	1	1
Kedougou	1.01	1.02	0.87	1.04