Espectroscopia de reflectância vis-nir na predição de produtividade da soja e definição de áreas de manejo do solo
Carregando...
Arquivos
Data
2020-02-28
Autores
Barbosa, Priscila de Miranda
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A cor do solo é reflexo, em parte, de seus constituintes minerais, que pode ser utilizado como indicador de nível, grau e intensidade de manejo do solo. A geoestatística, alinhada à modelagem matemática são capazes de auxiliar na tomada de decisões no campo. O Random Frost é uma linguagem de aprendizagem de máquina, na qual é possível delimitar áreas e estimar a produção. O objetivo deste estudo foi utilizar a espectroscopia de reflectância VIS-NIR na predição da produtividade de soja e no mapeamento de áreas de manejo. O estudo foi realizado em Alpinópolis – MG, no ano de 2016. Foram coletadas 100 amostras de solos na profundidade de 0-0,20 m para análises, em que foram quantificados os teores de areia, silte, argila, pH, matéria orgânica, saturação por bases e os teores de ferro cristalino e não cristalino. Os minerais da fração argila foram caraterizados por difração de raios - X (DRX). Por espectroscopia de reflectância VIS-NIR, obtiveram - - se os índices de cor (matiz, valor, croma, RGB e o índice de avermelhamento). A partir dos parâmetros de cor e da análise de Random Forest, estimou-se a produtividade da soja. Os resultados foram avaliados por estatística descritiva; e o padrão espacial, por análise geoestatística, para a elaboração dos mapas. As curvas da encosta convexa apresentam as concavidades mais acentuadas, indicando que, nessas regiões, há maior ocorrência de óxidos de ferro, pois estão bem representados os vales característicos dos minerais que são expressos na faixa do visível. A cor do solo obtida por espectroscopia de reflectância VIS-NIR foi eficiente na identificação de áreas com diferentes potenciais de produtividade da cultura da soja. A produtividade da soja respondeu ao comportamento das variáveis croma e IAV, sendo possível caracterizar 5 ambientes distintos de produção. A espectroscopia de reflectância VIS-NIR mostra-se viável como técnica indireta para definir zonas mínimas de manejo e auxilia nas tomadas de decisões táticas e operacionais que buscam incrementos na produtividade da soja.
The soil color is partly a reflection of its mineral constituents, which can be used as an indicator of soil management level, degree and intensity. Geostatistics, aligned with mathematical modeling, are able to assist in decision making in the field. Random Frost is a machine learning language, in which it is possible to define areas and estimate production. The aim of this study was to use the VIS-NIR reflectance spectroscopy to predict soybean yield and to map management areas. The study was carried out in Alpinópolis - MG, in 2016. 100 soil samples were collected at a depth of 0-0.20 m for analysis, in which the contents of sand, silt, clay, pH, organic matter, base saturation and the levels of crystalline and non-crystalline iron. The clay fraction minerals were characterized by X-ray diffraction (XRD). By VIS-NIR reflectance spectroscopy, the color indices were obtained (hue, value, chroma, RGB and the reddening index). Based on the color parameters and the Random Forest analysis, soybean productivity was estimated. The results were evaluated by descriptive statistics; and the spatial pattern, by geostatistical analysis, for the elaboration of maps. The curves of the convex slope show the most pronounced concavities, indicating that, in these regions, there is a greater occurrence of iron oxides, as the characteristic valleys of the minerals that are expressed in the visible range are well represented. The soil color obtained by VIS-NIR reflectance spectroscopy was efficient in identifying areas with different soybean productivity potentials. The soybean productivity responded to the behavior of the chroma and IAV variables, being possible to characterize 5 different production environments. VIS-NIR reflectance spectroscopy is shown to be viable as an indirect technique for defining minimum management zones and assists in making tactical and operational decisions that seek increases in soybean productivity.
The soil color is partly a reflection of its mineral constituents, which can be used as an indicator of soil management level, degree and intensity. Geostatistics, aligned with mathematical modeling, are able to assist in decision making in the field. Random Frost is a machine learning language, in which it is possible to define areas and estimate production. The aim of this study was to use the VIS-NIR reflectance spectroscopy to predict soybean yield and to map management areas. The study was carried out in Alpinópolis - MG, in 2016. 100 soil samples were collected at a depth of 0-0.20 m for analysis, in which the contents of sand, silt, clay, pH, organic matter, base saturation and the levels of crystalline and non-crystalline iron. The clay fraction minerals were characterized by X-ray diffraction (XRD). By VIS-NIR reflectance spectroscopy, the color indices were obtained (hue, value, chroma, RGB and the reddening index). Based on the color parameters and the Random Forest analysis, soybean productivity was estimated. The results were evaluated by descriptive statistics; and the spatial pattern, by geostatistical analysis, for the elaboration of maps. The curves of the convex slope show the most pronounced concavities, indicating that, in these regions, there is a greater occurrence of iron oxides, as the characteristic valleys of the minerals that are expressed in the visible range are well represented. The soil color obtained by VIS-NIR reflectance spectroscopy was efficient in identifying areas with different soybean productivity potentials. The soybean productivity responded to the behavior of the chroma and IAV variables, being possible to characterize 5 different production environments. VIS-NIR reflectance spectroscopy is shown to be viable as an indirect technique for defining minimum management zones and assists in making tactical and operational decisions that seek increases in soybean productivity.
Descrição
Palavras-chave
Cor do solo, Randon Forest, Indice de avermelhamento, Pedometria, Óxidos de ferro