SOIL PROPERTIES MAPPING USING PROXIMAL AND REMOTE SENSING AS COVARIATE

Pusch, Maiara; Oliveira, Agda L. G.; Fontenelli, Julyane V.; Amaral, Lucas R. do

doi:10.1590/1809-4430-Eng.Agric.v41n6p634-642/2021

Maiara Pusch

School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil.

Agda L. G. Oliveira

School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil.

Julyane V. Fontenelli

School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil.

Lucas R. do Amaral ^**Corresponding author. School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil. E-mail: lucas.amaral@feagri.unicamp.br

School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil.

https://orcid.org/0000-0001-8071-4449

*Corresponding author. School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil. E-mail: lucas.amaral@feagri.unicamp.br

Area Editor: Teresa Cristina Tarlé Pissarra

Figures (5)
Tables (1)

Thumbnail

FIGURE 1
Location of the study site and representation of the sampling points, indicating the soil samples.

Thumbnail

FIGURE 2
Apparent soil electrical conductivity data (spatial density from ∼272 samples ha⁻¹) (a) and bare soil reflectance in the SWIR2 band of a composition image of a historical series between 1984 and 2020 years (30 m spatial resolution) (b).

Thumbnail

FIGURE 3
Local and Global bivariate Moran index between soil properties and covariates. OM - organic matter; CEC – cation exchange capacity; V – base saturation; K – potassium; P – phosphorus; ECa – soil apparent electrical conductivity; SWIR2 – shortwave infrared band. ^ns – not significant correlation (p> 0.05), ^*p-value < 0.05, ^**p-value < 0.01, ^***p-value < 0.001.

Thumbnail

FIGURE 4
Semivariograms for all five soil properties (principal variables) according the geostatistical method: UGM - Ordinary Kriging; KED(ECa) – Residual Semivariograms when Kriging with External Drift was used with apparent soil electrical conductivity as auxiliary variable; KED(SWIR2) - Residual Semivariograms when Kriging with external Drift was used with the Short Infrared band SWIR2 as auxiliary variable; KED(ECa+SWIR2) - Residual Semivariograms when Kriging with external Drift was used with both covariates. OM – soil organic matter; CEC - cation exchange capacity; V - base saturation; P – available phosphorus, K - available potassium. The “circle” symbol is the experimental semivariogram and the solid lines are the theoretical semivariograms adapted by residual maximum likelihood (REML). NOTE: KED_ECa+swir2 are overlapping the KED_ECa.

Thumbnail

FIGURE 5
Spatial distribution of soil properties for the different geostatistical methods: UGM - Ordinary Kriging; KED(ECa) – Kriging with External Drift using apparent soil electrical conductivity as covariate; KED(SWIR2) - Kriging with external Drift with the Short Infrared band (SWIR2) as covariate; KED(ECa+SWIR2) - Kriging with external Drift with both covariates. OM - organic matter; CEC- cation exchange capacity; V - base saturation; P – available phosphorus; K – available potassium.

Thumbnail

TABLE 1
Semivariogram parameters (c - partial sill; c_o - nugget; a – range; and theoretical model) and statistics of the prediction performance in the leave-one-out cross validation (LOOCV) of soil properties for the different geostatistical methods (UMG and KED scenarios), using ECa and SWIR2 as covariates.

Soil property	Geostatistical method	Semivariogram parameters				LOOCV
Soil property	Geostatistical method	c	c_o	a	Model^ Semivariogram model: Sph – Spherical; Exp – Exponential.	EM	RMSE	NSE
OM (g kg⁻¹)	UGM ^* * when trend was modeled by coordinates, we used universal kriging as UGM method.	5.60	1.08	469.15	sph	0.00	1.76	0.84
	KED(ECa)	5.28	1.09	467.08	sph	0.00	1.75	0.84
	KED(SWIR2)	6.05	0.97	470.52	sph	−0.01	1.77	0.84
	KED(ECa + SWIR2)	5.72	0.99	468.16	sph	-0.01	1.77	0.84
CEC (mmol_c dm⁻³)	UGM ^* * when trend was modeled by coordinates, we used universal kriging as UGM method.	225.15	8.89	931.87	sph	0.09	6.39	0.85
	KED(ECa)	26.95	0.11	137.70	sph	−0.01	5.01	0.91
	KED(SWIR2)	107.95	7.77	428.32	sph	0.09	6.44	0.85
	KED(ECa + SWIR2)	27.11	0.18	138.50	sph	-0.01	5.03	0.91
V (%)	UGM	148.53	27.48	899.35	sph	0.07	7.19	0.52
	KED(ECa)	28.05	30.35	456.73	sph	0.00	6.57	0.60
	KED(SWIR2)	142.56	28.49	899.37	sph	0.08	7.21	0.51
	KED(ECa + SWIR2)	24.34	31.32	432.78	sph	-0.01	6.64	0.59
K (mmol_c dm⁻³)	UGM ^* * when trend was modeled by coordinates, we used universal kriging as UGM method.	0.30	0.03	205.10	exp	0.00	0.48	0.62
	KED(ECa)	0.14	0.09	385.44	sph	0.00	0.47	0.64
	KED(SWIR2)	0.21	0.05	319.90	sph	0.00	0.49	0.61
	KED(ECa + SWIR2)	0.15	0.09	385.37	sph	0.00	0.48	0.63
P (mg dm⁻³)	UGM	0.03	0.04	271.56	sph	0.04	5.71	0.02
	KED(ECa)	0.03	0.04	282.31	sph	0.04	5.74	0.01
	KED(SWIR2)	0.03	0.04	277.95	sph	0.04	5.75	0.01
	KED(ECa + SWIR2)	0.04	0.04	290.47	sph	0.04	5.80	−0.01

[1] *Corresponding author. School of Agricultural Engineering, University of Campinas/ Campinas - SP, Brazil. E-mail: lucas.amaral@feagri.unicamp.br

Brasil

Brasil

SOIL PROPERTIES MAPPING USING PROXIMAL AND REMOTE SENSING AS COVARIATE

ABSTRACT