Modeling the spatial distribution of wood volume in a cerrado stricto sensu remnant in Minas Gerais state, Brazil

• Autor(a) principal: Reis, Aliny Aparecida dos
• Data de Publicação: 2020
• Outros Autores: Diniz, Juliana Maria Ferreira de Souza, Acerbi Júnior, Fausto Weimar, Mello, José Márcio de, Batista, Anderson Pedro Bernardina, Ferraz Filho, Antonio Carlos
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/42417
• topic: Geostatistical models; Landsat 5 TM imagery; Multiple linear regression; Regression kriging; Brazilian savanna; Modelos geoestatísticos; Imagens Landsat 5 TM; Regressão linear múltipla; Krigagem com regressão; Cerrado

The Brazilian Savanna, the second largest biome in the country, has scarce information about its wood volume production. Since our aim was to contribute to the better wood volume characterization in Brazilian Savanna vegetation, we conducted a case study in a Cerrado Sensu Stricto remnant in Minas Gerais state, Brazil, using different approaches and datasets to model the spatial distribution of wood volume, including forest inventory data, remotely-sensed imagery, and geostatistical models. Wood volume data were obtained from a forest inventory carried out in the field. Spectral data were collected from a Landsat 5 TM satellite image, composed of spectral bands and vegetation indices. Ordinary kriging, multiple linear regression analysis, and regression kriging methods were used for wood volume estimation. Ordinary kriging resulted in estimates closer to each other in non-sampled areas (less variability) than the other methods for not considering information from these areas in the interpolation process. As multiple linear regression and regression kriging take into account the spectral data from remotely-sensed images, these methods provide higher discrimination potential for wood volume estimate mapping when vegetation presents high spatial heterogeneity, as in the Cerrado Sensu Stricto. Integration between field data, remotely-sensed imagery and geostatistical models provides a potential approach to spatially estimate wood volume in native vegetation.