In recent years there has been a great increase in the availability of novel techniques for forest monitoring purposes. On the one hand, there are ground-based systems like Terrestrial Laser Scanning (TLS). TLS allows the capture of the canopy structure in three dimensions with a precision not seen so far. This makes it possible to build precise forest representations. These models can be exploited to retrieve novel structural metrics like plant area profiles (Jupp et al., 2009). They can also deliver inputs for radiative transfer models, with which the forest's radiative response can be studied.
On the other hand, optical sensors carried by Unmanned Aerial Vehicles (UAV) produce ultra-high resolution images that can be processed to Digital Surface Models (DSM) of the top of canopy. New hyperspectral sensors for UAVs provide high spectral resolution observations of the canopy (Suomalainen et al., 2014). In combination, TLS and UAV permit detailed observations of forest biophysical parameters.
In addition to these novel techniques, traditional ones like digital hemispherical photography are available, which are widely used in forest monitoring especially in the tropics.
The first work package includes the establishment of a database with diverse datasets for a deciduous forest site and to monitor leaf senescence during autumn.
Validation of Top Of Canopy reflectance factors
Validation of satellite products is a key requirement to make products useful for further applications. Especially land applications require a reliable atmospheric correction. In this context, UAVs offer unique validation opportunities. UAVs operate in the lower part of the troposphere, which results in observations that are hardly influenced by atmospheric components. This makes it possible to validate the radiometric accuracy and in particular the atmospheric correction of satellite products.
The UAV based hyperspectral mapping system in Laboratory of Geo-information Science and Remote Sensing provides these advantages. Additionally, its hyperspectral capabilities allow flexible modelling of different satellite spectral bands such as those of Sentinel-2 and Landsat 8.
The second work package aims to compare atmospherically corrected Sentinel-2 and Landsat 8 time series with UAV based time series on both reflectance factor and vegetation index level.
Structural parameter retrieval from Sentinel-2 and Landsat 8 reflectance factors
Biophysical and chemical vegetation properties are most commonly derived from VIS/NIR observations. There are two most commonly used methodologies for this: inversion of vegetation radiative transfer models (VRTM) and empirical calibration of regression methods based on vegetation indices (VI). VRTMs like PROSAIL (Jacquemoud et al., 2009) are based on radiative transfer and therefore adaptable to many types of vegetation. However, they typically require extensive sets of input parameters, which result in ill-posed problems when it comes to the inversion of the model. VIs do not require further parameters, but their calibration with local data makes them site specific.
The aim of the third work package is to test approaches based on VRTM inversion and VI calibration for retrieval of structural properties, especially LAI from Sentinel-2 and Landsat 8 time series.