Spatial prediction of basal area and volume in Eucalyptus stands using Landsat TM data: an assessment of prediction methods

• DOI: 10.1186/s40490-017-0108-0
• Publication Date: 2018-01-16T11:46:51Z
• Authors: Aliny Aparecida d...; Mônica Canaan Car...; José Marcio de Mello; Lucas Rezende Gomide; Antônio Carlos Fe...; Fausto Weimar Ace...
• Keywords: Basal area; Eucalyptus camaldulensis; Forest Inventory

In fast-growing forests such as Eucalyptus plantations, the correct determination of stand productivity is essential to aid decision making processes and ensure efficiency wood supply chain. past decade, advances in remote sensing computational methods have yielded new tools, techniques, technologies that led improvements forest management assessments. Our aim was estimate map basal area volume stands through integration inventory, sensing, parametric, nonparametric spatial prediction. This study conducted 20 5-year-old clonal (362 ha) urophylla S.T.Blake x camaldulensis Dehnh. The are located northwest region Minas Gerais state, Brazil. Basal data were obtained from inventory operations carried out field. Spectral collected a Landsat 5 TM satellite image, composed spectral bands vegetation indices. Multiple linear regression (MLR), random (RF), support vector machine (SVM), artificial neural network (ANN) used for estimation. Using ordinary kriging, we spatialised residuals generated by prediction correction trends estimates more detailing behaviour volume. ND54 index variable had best correlation values with (r = − 0.91) 0.52) also most contributed MLR RF methods. algorithm presented smaller errors when compared other learning algorithms MLR. addition residual kriging did not necessarily result relative estimations these Random method mapping area. combination improvement accuracy all assessed this study, there always dependency structure method. approaches provide framework integrating field multispectral data, highlighting greatly improve estimation stands. has potential fast growth plantation monitoring, offering options robust analysis high-dimensional data.