Prediction of soil classes in a complex landscape in Southern Brazil

Moura-Bueno, Jean Michel; Dalmolin, Ricardo Simão Diniz; Horst-Heinen, Taciara Zborowski; Cancian, Luciano Campos; Schenato, Ricardo Bergamo; Dotto, André Carnieletto; Flores, Carlos Alberto

doi:10.1590/S1678-3921.pab2019.v54.00420

Acessibilidade / Reportar erro

Brasil

Pesquisa Agropecuária Brasileira

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

SOIL SCIENCE • Pesq. agropec. bras. 54 • 2019 • https://doi.org/10.1590/S1678-3921.pab2019.v54.00420 copy

Prediction of soil classes in a complex landscape in Southern Brazil

Predição de classes de solo em uma paisagem complexa no Sul do Brasil

Authorship SCIMAGO INSTITUTIONS RANKINGS

Abstract:

The objective of this work was to evaluate the use of covariate selection by expert knowledge on the performance of soil class predictive models in a complex landscape, in order to identify the best predictive model for digital soil mapping in the Southern region of Brazil. A total of 164 points were sampled in the field using the conditioned Latin hypercube, considering the covariates elevation, slope, and aspect. From the digital elevation model, environmental covariates were extracted, composing three sets, made up of: 21 covariates, covariates after the exclusion of the multicollinear ones, and covariates chosen by expert knowledge. Prediction was performed with the following models: decision tree, random forest, multiple logistic regression, and support vector machine. The accuracy of the models was evaluated by the kappa index (K), general accuracy (GA), and class accuracy. The prediction models were sensitive to the disproportionate sampling of soil classes. The best predicted map achieved a GA of 71% and K of 0.59. The use of the covariate set chosen by expert knowledge improves model performance in predicting soil classes in a complex landscape, and random forest is the best model for the spatial prediction of soil classes.

Index terms:
digital soil mapping; pedometry; predictive covariates; predictive models; soil-landscape relationship

Embrapa Secretaria de Pesquisa e Desenvolvimento; Pesquisa Agropecuária Brasileira Caixa Postal 040315, 70770-901 Brasília DF Brazil, Tel. +55 61 3448-1813, Fax +55 61 3340-5483 - Brasília - DF - Brazil
E-mail: pab@embrapa.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] *Corresponding author: bueno.jean1@gmail.com

Symbol	Brazilian soil classification system⁽¹⁾	Soil taxonomy⁽²⁾	Sampling points (%)
RL	Neossolo Litólico	Orthent	36
PBAC	Argissolo Bruno-Acinzentado	Udult	26
PVA	Argissolo Vermelho-Amarelo	Udult	14
CX	Cambissolo Háplico	Udept	9
RR	Neossolo Regolítico	Orthent	8
SX	Planossolo Háplico	Aqualf	7
Total			100

Covariate set⁽¹⁾	Prediction model⁽²⁾	Class accuracy (%)⁽³⁾					General accuracy (%)	Kappa index
Covariate set⁽¹⁾	Prediction model⁽²⁾	PBAC	PVA	RL	RR+CX	SX	General accuracy (%)	Kappa index
Training set
A	DT	94	88	98	83	100	93	0.91
	RF	65	50	88	17	67	63	0.50
	SVM	54	89	83	60	0	69	0.56
	MLoR	84	56	90	39	100	77	0.68
	Mean	74	71	90	50	67	76	0.66
B	DT	94	88	100	78	100	93	0.91
	RF	61	50	83	11	67	60	0.46
	SVM	54	73	83	50	0	67	0.54
	MLoR	66	62	84	42	80	71	0.61
	Mean	69	68	87	45	62	73	0.63
C	DT	97	93	88	82	100	90	0.87
	RF	63	45	77	13	100	60	0.46
	SVM	56	73	78	50	100	67	0.54
	MLoR	62	54	78	43	89	68	0.56
	Mean	70	66	80	47	97	71	0.61
Validation set
A	DT	75	43	80	25	100	65	0.52
	RF	83	29	90	0	100	65	0.50
	SVM	52	0	75	0	0	61	0.43
	MLoR	58	29	85	25	100	61	0.46
	Mean	67	25	83	13	75	63	0.48
B	DT	75	29	90	25	100	67	0.54
	RF	47	75	79	50	100	67	0.53
	SVM	48	33	83	50	0	63	0.47
	MLoR	50	40	86	75	100	69	0.57
	Mean	55	44	85	50	75	67	0.53
C	DT	50	23	72	25	34	51	0.32
	RF	56	40	87	100	100	71	0.59
	SVM	40	0	80	0	0	55	0.34
	MLoR	53	50	87	50	50	67	0.53
	Mean	50	28	82	44	46	61	0.44