%0 Conference Proceedings
%4 dpi.inpe.br/sbsr@80/2008/11.18.01.33
%2 dpi.inpe.br/sbsr@80/2008/11.18.01.33.45
%@isbn 978-85-17-00044-7
%T Sensoriamento remoto no mapeamento digital da fertilidade do solo: solucionando um grande inconveniente em agricultura de precisão
%D 2009
%A Ramírez-López, Leonardo,
%A Demattê, José Alexandre Melo,
%A Terra, Fabrício da Silva,
%A Bortoletto, Marco Antonio Melo,
%@affiliation Universidade de São Paulo/ESALQ
%@affiliation Universidade de São Paulo/ESALQ
%@affiliation Universidade de São Paulo/ESALQ
%@affiliation Universidade de São Paulo/ESALQ
%@electronicmailaddress lrlopez@esalq.usp.br
%@electronicmailaddress jamdtmat@esalq.usp.br
%@electronicmailaddress mamborto@esalq.usp.br
%@electronicmailaddress mamborto@esalq.usp.br
%E Epiphanio, José Carlos Neves,
%E Galvão, Lênio Soares,
%B Simpósio Brasileiro de Sensoriamento Remoto, 14 (SBSR)
%C Natal
%8 25-30 abr. 2009
%I Instituto Nacional de Pesquisas Espaciais (INPE)
%J São José dos Campos
%P 363-370
%S Anais
%1 Instituto Nacional de Pesquisas Espaciais (INPE)
%K espectroscopia de solos, quimiometría, variabilidade espacial, amostragem estratificada e em grade.
%X Precision agriculture requires high soil sampling density to assess the soil spatial variability and therefore the cost for asses this variability is significantly high. In this way, the aim of this study was to demonstrate that with a small number of soil samples (with low costs) statistical models for the quantification of soil attributes in a large number of foil samples based on the soil spectral reflectance can be obtained with suitable accuracy for the description of the spatial variability of attributes related to the soil fertility. In the study area of 473 ha 900 soil samples were collected in two depths. A soil analysis was carried out and were obtained Ca, Mg, K and CIC, was also calculated the base sum and soil textural fractions were also determined. For each soil sample a spectral reflectance was obtained. Models for the quantification of soil attributes were calibrated by partial least squares regression (PLS). These models were calibrated with three different quantity of soil samples, for the selection of these samples was used two different methods. The performance each model was evaluated by cross validation. Finally these models were applied to all spectral data and was obtained the predicted soil attribute values in each soil sample. The predicted soil attributes (by remote sensing) and the soil attributes form conventional soil analysis were mapped using geoestatística methods. This study demonstrates that the remote sensing can be solve a great problem in precision agriculture and digital soil mapping.
%9 Agricultura
%@language pt
%3 363-370.pdf