Improving the performance of a spectral model to estimate total nitrogen content with small soil samples sizes

Chemical and Biological Technologies in Agriculture

Table 3 Result comparison of using soil decomposed data and particle size of 0.15 mm\(-\)0.25 mm to build the model

Data size	Data category	Regressor	Calibration		Validation		Tenfold validation
Data size	Data category	Regressor	\(R^2_c\)	\(RMSE_c\,((g.kg^{-1}))\)	\(R^2_v\)	\(RMSE_v\,((g.kg^{-1}))\)	\(R^2_{cv}\)	RPIQ
111	Pristine mixed soil data^[1]	UVE-PLSR	0.514	0.181	0.457	0.118	0.433	2.105
		GPR	0.737	0.132	0.543	0.108	0.522	2.603
		SPA-MLR	0.374	0.198	0.418	0.132	0.385	2.032
	0.15-0.25mm data^[2]	UVE-PLSR	0.645	0.156	0.597	0.113	0.550	2.937
		GPR	0.695	0.140	0.627	0.107	0.598	2.895
		SPA-MLR	0.429	0.198	0.553	0.117	0.484	2.501
	0.15-0.25mm data	Bagging	0.568	0.172	0.608	0.101	0.579	2.514
		Stacking	0.625	0.161	0.548	0.103	0.510	2.227
		Cubist	0.719	0.139	0.631	0.098	0.589	2.631
666	Six groups of particle sizes data^[3]	PLSR	0.643	0.167	0.637	0.097	0.600	2.650
		Stacking	0.773	0.125	0.784	0.075	0.720	3.511
		Bagging	0.715	0.140	0.737	0.083	0.693	3.324

\([1]\) Pristine soil data refer to using a mixture soil sample data with particle size \(< 2\) mm for modelling
\([2]\) 0.15–0.25 mm data are for modelling based on the soil data of particle size ranging from 0.15 mm to 0.25 mm, which superior to the rest five particle sizes group
\([3]\) Six groups of particle sizes data means using the whole particle sizes data to build the model