Quantitative remote sensing of Norway spruce (Picea abies (L.) Karst.): spectroscopy from needles to crowns to canopies

Mountain ecosystems represent nearly one fourth of the Earth's land surface, and provide (ecosystem) services to a significant part of the world's human population. As was noted in the 1992 United Nations Conference on Environment and Development (UNCED) inRio de Janeiro(Agenda 21), these ecosystems are experiencing rapid degradation due to environmental and human impact at the local scale having, however, a global spread. This work contributes to the spatial monitoring of mountain forest ecosystems dominated byNorwayspruce trees ( Picea abies (L.) Karst.) through developing new quantitative approaches using optical hyperspectral remote sensing.A common denominator for all experiments described and performed throughout this work is spectroscopy, used in combination with ecological reasoning and physically based bottom-up scaling approaches. The spectral information acquired at the level of leaves (needles) is scaled up to the level of tree crowns and then to the level of forest canopies by means of radiative transfer modelling. This up-scaling method is based fully on cause-effect relations of vegetation-photon interaction allowing for its subsequent numerical inversion. We have employed radiative transfer models to develop an inversion routine for the retrieval of quantitative forest canopy biochemical (concentration of chlorophyll) and biophysical (leaf area index) parameters based on hyperspectral data acquired by theAirborne Imaging Spectroradiometer (AISA)Eagle.The starting point of this work was the calibration and validation of the leaf radiative transfer model PROSPECT (leaf optical PROperties SPECTra) for Norway spruce needles at wavelengths ranging from 450-1000 nm. Simultaneously, an investigation on the variability in hemispherical-directional reflectance, transmittance and absorptance of sunlit spruce needle samples was carried out for environmental stress resistant and stress resilient spruce crowns. Recalibration of the PROSPECT chlorophyll and dry matter specific absorption coefficients kab (l) and km (l) resulted in close agreement of the PROSPECT simulated needle optical properties with the spectral measurements of three investigated needle age-classes. The root mean square error (RMSE) between simulated and measured needle reflectance, transmittance, and absorptance signatures was 1.74%, 1.53%, and 2.91%, respectively. Testing the adjusted PROSPECT model for total chlorophyll concentration, dry matter content, and leaf water content retrieved simultaneously from laboratory based spectrometer measurements of independent needle samples showed an improved performance.When comparing sunlit needles of primary shoots of the two environmental stress response classes, significant quantitative differences were discovered at red edge (710-715 nm) and green (540-565 nm) wavelengths with respect to the variability in needle optical properties. These differences are related to the changes in the concentration of the foliar pigments, which proved to be significantly lower for needles of stress resilient trees. High variability in qualitative spectral characteristics was found at wavelengths relating to far-red (740 nm) and green (530 nm) fluorescence emissions. No significant spectral variation was found within the near infrared wavelengths of the tested needles. Consequently, we conclude that sunlit needle samples of a specific age-class collected from a randomly selected branch within two tested crown parts were representative for both of them, but they are optically different if collected from stress resistant or stress resilient crowns.The needle optical properties were up-scaled to the crowns and to the canopies by the 3-dimensional Discrete Anisotropic Radiative Transfer (DART) model. A detailed sensitivity analysis investigating the effect of woody elements introduced into the DART model on the nadir bidirectional reflectance factor (BRF) was performed at a very high spatial resolution (0.4 m) before its direct use for the Norway spruce canopy. The sensitivity analysis was performed separately for both sunlit and shaded parts of the simulated forest canopy and it was validated against BRF measurements at the same spatial resolution acquired over the simulated forest stand by the AISA sensor. Results showed a nadir BRF for the Norway spruce canopy modelled as pure foliage to be similar to the one for foliage including only robust woody constituents (i.e., trunks and branches of first order). The incorporation of small woody parts in DART caused the nadir canopy reflectance to decrease in the near-infrared (NIR) about 4%, in the red edge about 2%, and in the green bands less than 1%. These findings stressed the importance of including fine woody elements in radiative transfer based retrievals of the forest canopy properties at very high spatial resolution, especially if the NIR wavelengths are exploited.The coupled radiative transfer models PROSPECT and DART were employed to develop and test the sensitivity of a robust chlorophyll estimating optical index for a heterogeneous coniferous forest canopy. A newly proposed index named 'Area under curve Normalized to Maximal Band depth between 650-725nm' (ANMB 650-725 ) is based on the continuum removal method applied on reflectance spectra. The approach is taking advantage of the fine spectral resolution and sampling interval of hyperspectral images to isolate the chlorophyll absorption feature between 650 and 725 nm. The results, obtained from simulated hyperspectral data with a pixel size of0.9 m, showed a strong linear relationship of the ANMB 650-725 with spruce crown C ab concentration (r 2 = 0.9798) and an insensitivity for varying canopy structural features such as LAI and canopy closure. Chlorophyll concentration retrieval using the ANMB 650-725 index remained stable also after introduction of the spectral information of epiphytic lichen ( Pseudevernia sp.) and an increased sensor noise (signal to noise ratio equal to five).It can be concluded from this work that up-scaling procedures from needles to crowns to canopies can be successfully applied in mountain ecosystems. In particular the quantitative, physical based approaches proved to be robust and further helped to refine radiative transfer based models. In the near future, we will be seeing a number of spaceborne missions being realized (e.g., either of hyper- or super-spectral nature), where the presented up-scaling approaches can significantly contribute to increased variable retrieval accuracy. The integration of such derived products into an assessment programme for monitoring and forecasting the development of forest ecosystem services (e.g., Global Monitoring for Environment and Security (GMES)) is a logical consequence of this work.