Constructing satellite-derived hyperspectral indices sensitive to canopy structure variables of a Cordilleran Cypress (Austrocedrus chilensis) forest

Abstract

Satellite hyperspectral data were used to construct empirical spectral indices related to the canopy structure of a Cordilleran Cypress (Austrocedrus chilensis) forest located in the Andes of central Chile. Measurements of tree diameter at breast height (DBH) and tree height (TH) were performed for a set of plots located within a pure and unevenly aged stand of A. chilensis with moderate cover. Normalized difference vegetation indices (NDIs) related to DBH and TH were constructed from the corresponding hyperspectral data in Hyperion imagery. NDIs construction utilized the original spectral reflectance curve, its first derivative, and the continuum-removed reflectance in a two-step procedure that ranks NDIs based on their Spearman correlation with the response variable while controlling the false discovery rate. Several reflectance-based NDIs as well as a larger group of derivative-based NDIs were significantly related to DBH or TH (rho > 0.70). The NDIs most strongly related to the field variables were based on derivative bands located within the same spectral regions used by the broadband greenness index known as green normalized difference vegetation index. Most other significant NDIs used NIR bands, which are well-known for their sensitivity to foliage amount changes. The results obtained in this exploratory study mostly agreed with the spectral regions expected to be most sensitive to changes in the canopy structure of vegetation. Further research in other A. chilensis forests subject to different site and environmental conditions is needed in order to assess the applicability of the NDIs over a wider range of this endemic species.

Satellite hyperspectral data were used to construct empirical spectral indices related to the canopy structure of a Cordilleran Cypress (Austrocedrus chilensis) forest located in the Andes of central Chile. Measurements of tree diameter at breast height (DBH) and tree height (TH) were performed for a set of plots located within a pure and unevenly aged stand of A. chilensis with moderate cover. Normalized difference vegetation indices (NDIs) related to DBH and TH were constructed from the corresponding hyperspectral data in Hyperion imagery. NDIs construction utilized the original spectral reflectance curve, its first derivative, and the continuum-removed reflectance in a two-step procedure that ranks NDIs based on their Spearman correlation with the response variable while controlling the false discovery rate. Several reflectance-based NDIs as well as a larger group of derivative-based NDIs were significantly related to DBH or TH (rho > 0.70). The NDIs most strongly related to the field variables were based on derivative bands located within the same spectral regions used by the broadband greenness index known as green normalized difference vegetation index. Most other significant NDIs used NIR bands, which are well-known for their sensitivity to foliage amount changes. The results obtained in this exploratory study mostly agreed with the spectral regions expected to be most sensitive to changes in the canopy structure of vegetation. Further research in other A. chilensis forests subject to different site and environmental conditions is needed in order to assess the applicability of the NDIs over a wider range of this endemic species.