Abstract
Abstract. Collecting comprehensive knowledge about spectral signals in areas composed by complex structured objects is a challenging task in remote sensing. In the case of vegetation, shadow effects on reflectance are especially difficult to determine. This work analyzes a larch forest stand (Larix decidua MILL.) in Pinnis Valley (Tyrol, Austria). The main goal is extracting the larch spectral signal on Landsat 8 (LS8) Operational Land Imager (OLI) images using ground measurements with the Cropscan Multispectral Radiometer with five bands (MSR5) simultaneously to satellite overpasses in summer 2015. First, the relationship between field spectrometer and OLI data on a cultivated grassland area next to the forest stand is investigated. Median ground measurements for each of the grassland parcels serve for calculation of the mean difference between the two sensors. Differences are used as “bias correction” for field spectrometer values. In the main step, spectral unmixing of the OLI images is applied to the larch forest, specifying the larch tree spectral signal based on corrected field spectrometer measurements of the larch understory. In order to determine larch tree and shadow fractions on OLI pixels, a representative 3D tree shape is used to construct a digital forest. Benefits of this approach are the computational savings compared to a radiative transfer modeling. Remaining shortcomings are the limited capability to consider exact tree shapes and nonlinear processes. Different methods to implement shadows are tested and spectral vegetation indices like the Normalized Difference Vegetation Index (NDVI) and Greenness Index (GI) can be computed even without considering shadows.
Highlights
Spatial heterogeneity and the combination of geometrical views and shadowing are some of the main obstacles when trying to capture tree spectral signals
We aim at overcoming this issue by a simple geometrical modeling of the larch forest and a linear spectral unmixing of Landsat 8 (LS8) Operational Land Imager (OLI) surface reflectance values, similar to the method as described in Roberts et al (1998)
One endmember component in the spectral unmixing procedure is the understory, which is captured with the Cropscan MSR5, a sensor with OLI-like bands
Summary
Spatial heterogeneity and the combination of geometrical views and shadowing are some of the main obstacles when trying to capture tree spectral signals. Practical approaches are often based on field spectrometers positioned above the trees, for example on a 30 m high mast (Sukuvaara et al, 2007) or the bucket of a lift vehicle (Koch et al, 1990). The problem about these approaches is that they are cost-intensive and it is difficult to tell if the whole tree, more than that or just parts are captured.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
