Abstract
Abstract. Accurate mapping of forest species composition is an important aspect of monitoring and management planning related to ecosystem functions and services associated with water refinement, carbon sequestration, biodiversity, and wildlife habitats. Although different vegetation species often have unique spectral signatures, mapping based on spectral reflectance properties alone is often an ill-posed problem, since the spectral signature is as well influenced by age, canopy gaps, shadows and background characteristics. Thus, reducing the unknown variation by knowing the structural parameters of different species should improve determination procedures. In this study we combine imaging spectrometry (IS) and airborne laser scanning (ALS) data of a mixed needle and broadleaf forest to differentiate tree species more accurately as single-instrument data could do. Since forest inventory data in dense forests involve uncertainties, we tried to refine them by using individual tree crowns (ITC) position and shape, which derived from ALS data. Comparison of the extracted spectra from original field data and the modified one shows how ALS-derived shape and position of ITCs can improve separablity of the different species. The spatially explicit information layers containing both the spectral and structural components from the IS and ALS datasets were then combined by using a non-parametric support vector machine (SVM) classifier.
Highlights
Forests cover almost one third of the total land surface of the Earth and play an important role in the global energy and matter fluxes between atmosphere and the land surface
Imaging spectrometry (IS) data are better suited for classification of forest species than ordinary multispectral remotely sensed data (Ustin and Xiao 2001), But discrimination of the different species with similar spectral response is still difficult task (Koetz et al, 2008)
We introduced a combination method for using biophysical parameters of the forest to improve species detection on the imaging spectrometry (IS) datasets
Summary
Forests cover almost one third of the total land surface of the Earth and play an important role in the global energy and matter fluxes between atmosphere and the land surface. The tree species composition is an important aspect of forest monitoring as well as for management planning. Assessing of tree species composition with traditional fieldwork is labor-intensive, time-consuming and mostly limited by spatial extent; remote sensing data enables to overcome these limitations. Imaging spectrometry (IS) data are better suited for classification of forest species than ordinary multispectral remotely sensed data (Ustin and Xiao 2001), But discrimination of the different species with similar spectral response is still difficult task (Koetz et al, 2008). Airborne laser scanning (ALS) systems are promising tool for providing forest biophysical information in both horizontal and vertical directions due to the penetration of the transmitted laser signal through the canopy
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have