IWA Publishing

Table 3

Results of downscaling dominant predictors by ANN

Feature extraction method	Variable	Model	RMSE^a train	RMSE^a verification	DC train	DC verification
WTC	Precipitation	BNU-ESM	0.23	0.25	0.74	0.72
		Can-ESM5	0.14	0.16	0.84	0.81
		INM-CM5	0.13	0.16	0.81	0.76
	Temperature	BNU-ESM	0.11	0.15	0.9	0.88
		Can-ESM5	0.05	0.08	0.94	0.90
		INM-CM5	0.05	0.11	0.94	0.92
CC	Precipitation	BNU-ESM	0.45	0.41	0.56	0.53
		Can-ESM5	0.35	0.39	0.62	0.65
		INM-CM5	0.22	0.26	0.71	0.65
	Temperature	BNU-ESM	0.47	0.43	0.56	0.51
		Can-ESM5	0.34	0.30	0.58	0.55
		INM-CM5	0.36	0.32	0.56	0.51

Feature extraction method	Variable	Model	RMSE^a train	RMSE^a verification	DC train	DC verification
WTC	Precipitation	BNU-ESM	0.23	0.25	0.74	0.72
		Can-ESM5	0.14	0.16	0.84	0.81
		INM-CM5	0.13	0.16	0.81	0.76
	Temperature	BNU-ESM	0.11	0.15	0.9	0.88
		Can-ESM5	0.05	0.08	0.94	0.90
		INM-CM5	0.05	0.11	0.94	0.92
CC	Precipitation	BNU-ESM	0.45	0.41	0.56	0.53
		Can-ESM5	0.35	0.39	0.62	0.65
		INM-CM5	0.22	0.26	0.71	0.65
	Temperature	BNU-ESM	0.47	0.43	0.56	0.51
		Can-ESM5	0.34	0.30	0.58	0.55
		INM-CM5	0.36	0.32	0.56	0.51

^aNormalized RMSE.