The spatial distribution of the leaf-area-index (LAI) and its seasonality in forest stands is of special interest for phenological ecosystem studies. As a measure of energy, gas and water exchanges with plants it serves as surrogate for e.g. biomass and timber volume models, carbon sequestration, and forest health status. LAI is sensitively responding to seasonal temperature changes and acts as a suitable descriptor for the characteristics and stability of regional and micro climate. This is of particular interest for the highly sensitive ecosystems of mountain forests. LAI derived from vegetation indices (VI) computed by the combination of spectral bands (e.g. normalized differenced vegetation index, NDVI), has proven a standard product from space- and airborne optical sensors in various resolutions.
However, in complex environments, such as mountainous regions, topography and structural differentiation of forest stands are diminishing the quality of LAI-estimations from optical remote sensing sensors. To date accuracy limitations are due to mixed pixel effects combining spectral properties of different objects and understory and shadowing effects of topography. Sensor-, sun- and ambient-occlusion effects within the canopy also lower the reliability of LAI-estimations. In order to compensate the current problems, multivariate relationships (including forest structure) are required for a more reliable LAI-estimation in complex forest stands.
The increasing availability of spatial and temporal (in terms of revisiting time) high resolution satellite images allows the worldwide application of methods developed for airborne remote sensing sensors. This opens new possibilities for the detection and characterization of forest structure by texture signatures, considering spatial variations of relevant spectral bands (red, NIR, panchromatic). The extension of spectral resolution, adding e.g. middle infrared channel (MIR), is also adding further possibilities in analysis. Additionally, the dynamic acquisition of the space borne imagery allows the characterization of high frequent temporal spectral signatures (in the order of several days).
The project phenoSAT-α (Area-wide mapping of phenology using optical high resolution satellite imagery) – as a feasibility study – aims at linking spectral and texture signatures to discrete forest geometries and spatial distributions of LAI. By deriving forest structure related indices, specific multivariate relationships between LAI, vegetation indices and texture indices are suggested. By texture and signature-based LAI-functions a more differentiated LAI-estimation is aimed in terms of tree species, tree age, spatial resolution and time resolution.
Relationships between textural and spectral signatures and forest structure will be derived, calibrated and validated using detailed Radiative Transfer Modelling (RTM) for a selected set of detailed ground plot reconstructions. Using terrestrial laser scanning point clouds for selected reference plots, tree architectures will be reconstructed, allowing a dynamic modelling of foliage for various time steps and densities.
The coupling of reconstructed plot models (branch density, crown coverage, etc), simulated foliage densities (LAI) and simulated image textures will provide a deeper insight into texturestructure relationships. As object-based reconstructions, the datasets include object classifications. This allows the separation of respective classes such as foliage, branches and understory in the simulations, providing detailed insight into the dependencies between various indices and single stand components.