Investigation into the Effects of Climatic Change on Temperature, Rainfall, and Runoff of the Doroudzan Catchment, Iran, Using the Ensemble Approach of CMIP3 Climate Models

Table 8

Performances of different neural networks concerning the monthly rainfall-runoff model for the Doroudzan catchment.

Ann type	Input	Output	Learning algorithm	Network architecture	Transfer function		Data percentage			RMSE			r
					First	Second	Learning	Validation	Test	Learning	Validation	Test
GFF	Tave,P,E	Q	CG	3-4-4-1	TANH	LTANH	65	15	25	0.15	0.32	19.13	0.840
	Tave,P,E	Q	CG	3-15-8-1	TANH	LTANH	65	15	25	0.15	0.23	38.66	0.719
	Tave,P,E	Q	CG	3-4-4-1	TANH	LTANH	65	15	25	0.12	0.65	35.69	0.586
	Tave,P,E	Q	CG	3-4-4-1	TANH	LTANH	80	10	10	0.10	1.18	34.99	0.600
	Tave,P,E	Q	CG	3-4-4-1	TANH	LTANH	70	10	20	0.19	0.20	44.98	0.597
	P,E	Q	CG	2-4-4-1	TANH	LTANH	70	10	20	0.15	0.28	27.18	0.670
	P,Tave	Q	CG	2-4-4-1	TANH	LTANH	70	10	20	0.15	0.28	34.18	0.61
	P-3,P-2,P-1,P	Q	CG	4-4-4-1	TANH	LTANH	75	10	15	0.0080785	0.02670124	17.50	0.726
	P-3,P-2,P-1,P	Q	CG	4-14-10-1	TANH	LTANH	75	10	15	0.00249497	0.00467218	46.12	0.238
	P-3,P-2,P-1,P	Q	CG	4-4-4-1	TANH	LTANH	75	10	15	0.00268782	0.00762632	20.11	0.911
	P-2,P-1,P	Q	CG	3-4-4-1	TANH	LTANH	75	10	15	0.0176	0.0232	54.23	0.530
	P-1,P	Q	CG	2-4-4-1	TANH	LTANH	75	10	15	0.0324	0.02670124	17.5	0.48