Artificial Neural Networks for Estimating Soil Water Retention Curve Using Fitted and Measured Data

Table 2

Statistical analysis of artificial neural networks developed in this study for predicting the soil water retention curve.

Matric potentials (−cm)
Index	0		10		20		30		50		100		200		500		1000		Mean; ANN 1	Mean; ANN 2	ANNo
	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2	ANN; 1	ANN; 2
RMSE	0.064	0.062	0.057	0.054	0.056	0.055	0.051	0.055	0.050	0.060	0.069	0.075	0.082	0.098	0.083	0.087	0.078	0.084	0.066	0.070	0.088
	0.697	0.712	0.727	0.756	0.755	0.763	0.800	0.773	0.818	0.740	0.699	0.645	0.606	0.433	0.559	0.508	0.543	0.479	0.689	0.645	0.640
GMER	1.062	1.066	1.041	1.036	1.032	1.026	1.027	1.035	1.025	1.055	1.077	1.106	1.136	1.076	1.228	1.278	1.272	1.346	1.100	1.114	0.923
Target mean	0.428		0.407		0.388		0.368		0.339		0.299		0.268		0.236		0.216		0.328^a	0.328^a	0.331^b
Estimated; mean	0.448	0.444	0.414	0.412	0.388	0.387	0.365	0.367	0.334	0.339	0.300	0.303	0.271	0.247	0.238	0.242	0.213	0.219	0.330	0.329	0.273

In this case, target mean refers to water contents fitted by van Genuchten parameters for each matric potential. ^bThe mean value of observed water contents.