Forest canopy nitrogen concentration retrieval using hyperspectral remote sensing

Figure 1: Spectral signatures of deciduous and needle-leaved sample plots derived from APEX data. Click to enlarge image

Figure 2: Percent dry weight nitrogen concentrations in the canopy of individual tree species (left) and forest types (right). Click to enlarge image

Figure 3: Percent dry weight nitrogen canopy concentrations of deciduous and needle-leaved trees using two statistical techniques. After the classification to needle-leaved and deciduous species each group is assigned the mean nitrogen concentration valued estimated from lab measurements (broadleaved = 2.623 %N, needle-leaved 1.380). Click to enlarge image

The development of imaging spectrometers, the availability of hyperspectral observations from air- and spaceborne platforms as well as the facility of their processing has remarkably evolved over the last decade. This trend leads to a level of maturity, which makes imaging spectrometry accessible and useful for a larger research and user community. Against this background Hyper-Swiss-Net aims at developing a range of prototype application products drawing from the diverse expertise present in the project consortium and the respective user community. The development and implementation of the different products will be based on dedicated flight experiments with the airborne ESA imaging spectrometer APEX and will directly build on and link into the operational capabilities of the APEX processing and archiving facility (PAF). The scientific expertise gathered during the project will be further disseminated within the Swiss research community by integrating the developed capabilities into specific teaching modules.

Anthropogenic disturbances have significantly affected the biological diversity, the carbon cycle and sequestration dynamics and the overall services and structure of terrestrial ecosystems. Use of remote sensing and in particular of imaging spectroscopy (IS) data can assist greatly towards improved ecological analyses. A number of studies exist where key remotely-sensed structural and physiological parameters of vegetation directly linked to underlying ecological processes can be estimated. These are (a) LAI (b) aboveground biomass (c) APAR and (d) land use classifications since vegetation cover types differ significantly in morphological and ecophysiological characteristics that drive certain ecological processes. However, another very important aspect of plant ecophysiology that controls terrestrial biogeochemical processes is foliar canopy chemistry. Canopy concentrations of chlorophyll, nitrogen and lignin drive key processes like rate of photosynthetic production. Nitrogen is linked to the availability of nutrients and to carbon allocation and lignin to the rate of nutrient cycling via their influence on decomposition rates. Additionally, the concentration of nitrogen and carbon in plants expressed as foliar Carbon/Nitrogen ratio (C:N) drive processes such as decomposition and mineralization and thus strongly influence soil organic matter concentrations and turnover rates. Consequently, foliar C:N is a core variable of a number of ecosystem process models.

In the last decades a lot of research has been carried out using IS data and statistical methods for the assessment of suitable wavelengths for N content determination. Hyperspectral sensors, have been developed that make use of the unique spectral characteristics of vegetation and offer the potential to estimate canopy biochemical concentrations. Studies using hyperspectral remote sensing data to estimate canopy biochemical concentrations have been carried out in different ways: under controlled laboratory conditions, in the field using a spectroradiometer and using data collected from airborne and spaceborne hyperspectral sensors. All of the above studies have successfully developed statistical (empirical) methodologies and accomplish high accuracies in predicting foliar biochemical concentration.

For this project, the Landuse Dynamics module from WSL implements an algorithm that makes use of the statistical method (or empirical approach) and specific absorption features that might not be implemented in RTM’s for generating algorithms to estimate canopy nitrogen content of forest habitats. We attempt to establish such robust statistical relationships by performing several measurements under varying field conditions and for different forest habitat mixtures and species to sample as much of the nitrogen variability present within the study area. Field sampling of canopy nitrogen will primarily be required for the first years of IS APEX data acquisition to establish relationships between spectral reflectance and nitrogen concentration values. Field sampling would be required until a large sample size that covers the observed variability of nitrogen concentrations driven by forest species, age and mixtures will be collected. In 2004, 2008 and 2009 WSL has performed extensive field campaigns collecting these data and will continue to extent canopy nitrogen concentrations measurements over different IS APEX data acquisition sites across Switzerland, thus making the use of WSL algorithm operational.

Contact

• Achilleas Psomas