Digital mapping of soil attributes using machine learning

Patrícia Morais da Matta Campbell, Márcio Rocha Francelino, Elpídio Inácio Fernandes Filho, Pablo de Azevedo Rocha, Bruno Campbell de Azevedo


Mapping the chemical attributes of the soil on a large scale can result in gains when planning the use and occupation of the land. There are different techniques available for this purpose, whose performance should be tested for different types of landscapes. The aim of this study was to spatialize chemical attributes of the soil, comparing eight methods of prediction. Forty morphometric attributes, generated from a digital elevation model, were used as independent variables, in addition to geophysical data, images from the Landsat 8 satellite and the NDVI. All possible combinations between the satellite bands were calculated, generating 28 new variables. Combinations between the Th, U and K bands obtained from the geophysical data were also calculated, generating a further three variables. The final variables to be calculated were the distances between the four points of the edges of the basin (d1, d2, d3 and d4). The dependent variables for the model were Al, Ca, Fe, K, Mg, Na, Si, Ti, Cr, Cu, Mn, Ni, P, Pb, V, Zn, Zr, S and Cl. A total of 200 soil samples were used, which were collected from 100 points at two depths (0-10 and 10-30 cm); the total elements were determined using an X-ray fluorescence analyzer. The Random Forest algorithm proved to be superior to the others in predicting the chemical attributes of the soil at both depths, and is suitable for predicting soil attributes in the study region. Spatial variables are essential, and should be considered when modelling chemical elements in the soil. Using the methods under test, it is possible to predict elements with R² values ranging from 0.32 to 0.62.


XRF; Spatial approach; Prediction models

Texto completo:



ADHIKARI, K. et al. Digital mapping of soil organic carbon contents and stocks in Denmark. Plos One, v. 9, n. 8, p. e105519, 2014.

AKPA, S. I. C. et al. Total soil organic carbon and carbon sequestration potential in Nigeria. Geoderma, v. 271, p. 202-215, 2016.

ALVES, E. E. N. et al. Determinação da massa por área mínima de amostras de solo e vegetal para análise no µ-EDX. In: SIMPÓSIO MINEIRO DE CIÊNCIA DO SOLO, 3., 2015, Viçosa, MG. Anais [...]. Viçosa, MG: UFV, 2015. p. 28-30.

ARRUDA, G. P. de; DEMATTÊ, J. A. M.; CHAGAS, C. da S. Mapeamento digital de solos por redes neurais artificiais com base na relação solo-paisagem. Revista Brasileira de Ciência do Solo, v. 37, p. 327-338, 2013.

BISHOP, T. F. A. et al. Modelling soil attribute depth functions with equal-area quadratic smoothing splines. Geoderma, v. 91, n. 1/2, p. 27-45, 1999.

BÖHNER, J.; MCCLOY, K. R.; STROBL, J. SAGA: analysis and modelling applications. Göttingen: Göttinger Geographische Abhandlungen, 2006. 130 p.

CAMERA, C. et al. A high resolution map of soil types and physical properties for Cyprus: a digital soil mapping optimization. Geoderma, v. 285, p. 35-49, 2017.

CHAGAS, C. da S. et al. Spatial prediction of soil surface texture in a semiarid region using random forest and multiple linear regressions. Catena, v. 139, p. 232-240, 2016.

DAVIES, B. E.; GAMM, S. A. Trend surface analysis applied to soil reaction values from Kent, England. Geoderma, v. 3, p. 223-231, 1969.

DHARUMARAJAN, S. et al. Spatial prediction of major soil properties using Random Forest techniques: a case study in semi-arid tropics of South India. Geoderma Regional, v. 10, p. 154-162, 2017.

GOMEZ, C. et al. Soil organic carbon prediction by hyperspectral remote sensing and field vis-NIR spectroscopy: an australian case study. Geoderma, v. 146, n. 3/4, p. 403-411, 2008.

GRIMM, T. et al. Soil organic carbon concentrations and stocks on Barro Colorado Island: digital soil mapping using Random Forests analysis R. Geoderma, v. 146, p. 102-113, 2008.

GROENEVELD, R. A.; MEEDEN, G. Measuring skwness and kurtosis. Jounal of the Royal Statistical Society, v. 33, n. 4, p. 391-399, 1984.

GUO, P-T. et al. Digital mapping of soil organic matter for rubber plantation at regional scale: an application of random forest plus residuals kriging approach. Geoderma, v. 237-238, p. 49-59, 2015.

HENGL, T. et al. Mapping soil properties of Africa at 250 m resolution: Random Forests significantly improve current predictions. Plos One, v. 25, p. 1-26, 2015.

HEUVELINK, G. B. M. et al. Geostatistical prediction and simulation of European soil property maps. Geoderma Regional, v. 7, n. 2, p. 201-215, jun. 2016.

KABATA-PENDIAS, A. Trace elements in soils and plants. 4. ed. New York: CRC, 2010. 533 p.

KISS, J. J. et al. The distribution of fallout Cesium-137 in southern Saskatchewan, Canada. Journal of Environmental Quality, v. 17, p. 445-452, 1988.

LAGACHERIE, P. Digital soil mapping: a state of the art. In: HARTEMINK, A. E.; McBRATNEY, A. B.; MENDONÇA-SANTOS, M. de L. Digital soil mapping with limited data. New York: Springer, 2008. cap. 1, p. 3-14.

LIMA, J. S. de S. et al. Análise espacial de atributos químicos do solo e da produção da cultura pimenta-do-reino (piper nigrum, l.). Idesia, v. 28, n. 2, p. 31-39, 2010.

MA, Y. et al. Mapping key soil properties to support agricultural production in Eastern China. Geoderma Regional, v. 10, p. 144-153, 2017.

MALONE, B. P. et al. Using model averaging to combine soil property rasters from legacy soil maps and from point data. Geoderma, v. 232/234, p. 34-44, 2014.

McBRATNEY, A. B. et al. On digital soil mapping. Geoderma, v. 117, n. 2, p. 3-52, 2003.

MENEZES, M. D. et al. Solum depth spatial prediction comparing conventional with knowledge-based digital. Scientia Agricola, v. 71, n. 4, p. 316-323, 2014.

OLIVEIRA, S. et al. Modeling spatial patterns of fire occurrence in Mediterranean Europe using multiple regression and Random Forest. Forest Ecology and Management, v. 275, p. 117-129, 2012.

PINHEIRO, H. S. K. et al. Modelos de elevação para obtenção de atributos topográficos utilizados em mapeamento digital de solos. Pesquisa Agropecuária Brasileira, v. 47, n. 9, p. 1384-1394, 2012.

POGGIO, L. et al. Regional scale mapping of soil properties and their uncertainty with a large number of satellite-derived covariates. Geoderma, v. 209/210, p. 1-14, 2013.

RAMIFEHIARIVO, N. et al. Mapping soil organic carbon on a national scale: towards an improved and updated map of Madagascar. Geoderma Regional, v. 9, p. 29-38, 2017.

ROUSE J. W. et al. Monitoring the vernal advancement and retrogradation (Green wave effect) of natural vegetation. Greenbelt: National Aerospace Spatial Adminitration, 1973. 371 p.

RUSSELL, J. S.; MOORE, A. W. Comparison of different depth weightings in the numerical analysis of anisotropic soil profile data. In: International Congress of Soil Science, 9., 1968, Adelaide. Transactions […]. Adelaide: ISSS, 1968. v. 4, p. 205-2013. 1968.

RYAN, P. J. et al. Integrating forest soils information across scales: spatial prediction of soil properties under Australian forests. Forest Ecology and Management, v. 138, p. 139-157, 2000.

SREENIVAS, K. et al. Digital mapping of soil organic and inorganic carbon status in India. Geoderma, v. 269, p. 160-173, 2016.

VÅGEN, T-G. et al. Mapping of soil properties and land degradation risk in Africa using MODIS reflectance. Geoderma, v. 263, p. 216-225, 2016.

VAYSSE, K.; LAGACHERIE, P. Evaluating digital soil mapping approaches for mapping GlobalSoilMap soil properties from legacy data in Languedoc-Roussillon (France). Geoderma Regional, v. 4, p. 20-30, 2015.

WIESMEIER, M. et al. Digital mapping of soil organic matter stocks using Random Forest modeling in a semi-arid steppe ecosystem. Plant Soil, v. 340, p. 7-24, 2011.

YANG, R-M. et al. Comparison of boosted regression tree and random forest models for mapping topsoil organic carbon concentration in an alpine ecosystem. Ecological Indicators, v. 60, p. 870-878, 2016.

Revista Ciência Agronômica ISSN 1806-6690 (online) 0045-6888 (impresso), Site:, e-mail: - Fone: (85) 3366.9702 - Expediente: 2ª a 6ª feira - de 7 às 17h.