Description / Table of Contents:
Arctic tundra ecosystems are experiencing warming twice the global average and Arctic vegetation is responding in complex and heterogeneous ways. Shifting productivity, growth, species composition, and phenology at local and regional scales have implications for ecosystem functioning as well as the global carbon and energy balance. Optical remote sensing is an effective tool for monitoring ecosystem functioning in this remote biome. However, limited field-based spectral characterization of the spatial and temporal heterogeneity limits the accuracy of quantitative optical remote sensing at landscape scales. To address this research gap and support current and future satellite missions, three central research questions were posed:
• Does canopy-level spectral variability differ between dominant low Arctic vegetation communities and does this variability change between major phenological phases?
• How does canopy-level vegetation colour images recorded with high and low spectral resolution devices relate to phenological changes in leaf-level photosynthetic pigment concentrations?
• How does spatial aggregation of high spectral resolution data from the ground to satellite scale influence low Arctic tundra vegetation signatures and thereby what is the potential of upcoming hyperspectral spaceborne systems for low Arctic vegetation characterization?
To answer these questions a unique and detailed database was assembled. Field-based canopy-level spectral reflectance measurements, nadir digital photographs, and photosynthetic pigment concentrations of dominant low Arctic vegetation communities were acquired at three major phenological phases representing early, peak and late season. Data were collected in 2015 and 2016 in the Toolik Lake Research Natural Area located in north central Alaska on the North Slope of the Brooks Range. In addition to field data an aerial AISA hyperspectral image was acquired in the late season of 2016. Simulations of broadband Sentinel-2 and hyperspectral Environmental and Mapping Analysis Program (EnMAP) satellite reflectance spectra from ground-based reflectance spectra as well as simulations of EnMAP imagery from aerial hyperspectral imagery were also obtained.
Results showed that canopy-level spectral variability within and between vegetation communities differed by phenological phase. The late season was identified as the most discriminative for identifying many dominant vegetation communities using both ground-based and simulated hyperspectral reflectance spectra. This was due to an overall reduction in spectral variability and comparable or greater differences in spectral reflectance between vegetation communities in the visible near infrared spectrum.
Red, green, and blue (RGB) indices extracted from nadir digital photographs and pigment-driven vegetation indices extracted from ground-based spectral measurements showed strong significant relationships. RGB indices also showed moderate relationships with chlorophyll and carotenoid pigment concentrations. The observed relationships with the broadband RGB channels of the digital camera indicate that vegetation colour strongly influences the response of pigment-driven spectral indices and digital cameras can track the seasonal development and degradation of photosynthetic pigments.
Spatial aggregation of hyperspectral data from the ground to airborne, to simulated satel-lite scale was influenced by non-photosynthetic components as demonstrated by the distinct shift of the red edge to shorter wavelengths. Correspondence between spectral reflectance at the three scales was highest in the red spectrum and lowest in the near infra-red. By artificially mixing litter spectra at different proportions to ground-based spectra, correspondence with aerial and satellite spectra increased. Greater proportions of litter were required to achieve correspondence at the satellite scale.
Overall this thesis found that integrating multiple temporal, spectral, and spatial data is necessary to monitor the complexity and heterogeneity of Arctic tundra ecosystems. The identification of spectrally similar vegetation communities can be optimized using non-peak season hyperspectral data leading to more detailed identification of vegetation communities. The results also highlight the power of vegetation colour to link ground-based and satellite data. Finally, a detailed characterization non-photosynthetic ecosystem components is crucial for accurate interpretation of vegetation signals at landscape scales.
Type of Medium:
vi, 126 Seiten
Dissertation, Universität Potsdam, 2019
Table of Contents
1.1 Scientific Background and Motivation
1.1.1 Arctic Tundra Vegetation
1.1.2 Remote Sensing of Arctic Tundra Vegetation
1.1.3 Hyperspectral Remote Sensing of Arctic Vegetation
1.2 Aims and Objectives
1.3 Study Area and Data
1.3.1 Toolik Lake Research Natural Area
1.3.2 In-situ Canopy-level Spectral Data
1.3.3 True-colour Digital Photographs
1.3.4 Leaf-level Photosynthetic Pigment Data
1.3.5 Airborne AISA Imagery
1.3.6 Simulated EnMAP and Sentinel-2 Reflectance Spectra
1.3.7 Simulated EnMAP Imagery
1.4 Thesis Structure and Author Contributions
1.4.1 Chapter 2 -A Phenological Approach to Spectral Differentiation of Low-Arctic Tundra Vegetation Communities, North Slope Alaska
1.4.2 Chapter 3 -Monitoring Pigment-driven Vegetation Changes in a Low Arctic Tundra Ecosystem Using Digital Cameras
1.4.3 Implications of Litter and Non-vascular Components on Multiscale Hyperspectral Data in a low-Arctic Ecosystem
2 A Phenological Approach to Spectral Differentiation of Low Arctic Tundra Vegetation Communities, North Slope Alaska
2.3 Materials and Methods
2.3.1 Study Site and Low Arctic Vegetation Types
2.3.2 Ground-Based Data and Sampling Protocol
2.3.3 EnMAP and Sentinel-2 Surface Reflectance Simulation
2.3.4 Stable Wavelength Identification Using the InStability Index
2.4.1 Spectral Characteristics by Phenological Phase
2.4.2 InStability Index and Wavelength Selection of Ground-based Spectra
2.4.3 InStability Index and Wavelength Selection of Simulated Satellite Reflectance Spectra
2.5.1 Phenological Phase and Wavelength Stability of Ground-based Spectra
2.5.2 Phenological Phase and Wavelength Stability of Satellite Resampled Spectra
2.5.3 Influence of Spatial Scale
2.8 Supplementary Material
2.8.1 Data Publication
3 Monitoring Pigment-driven Vegetation Changes in a Low Arctic Tundra Ecosystem Using Digital Cameras
3.3.1 Study Site
3.3.2 Digital Photographs
3.3.3 Field-based Spectral Data
3.3.4 Vegetation Pigment Concentration
3.3.5 Data Analyses
3.4.1 RGB Indices as a Surrogate for Pigment-driven Spectral Indices
3.4.2 RGB Indices as a Surrogate for Leaf-level Pigment concentration
3.7 Supplementary Material
3.7.1 Data Publication
4 Implications of Litter and Non-vascular Components on Multiscale Hyperspectral Data in a Low Arctic Ecosystem
4.3 Materials and Methods
4.3.1 Study Site
4.4 Remote Sensing Data
4.4.1 Ground-based Image Spectroscopy Data
4.4.2 Airborne AISA Hyperspectral Data
4.4.3 EnMAP Simulation
4.4.4 Spectral Comparison by Wavelength
4.4.5 Linear Mixture Analysis
4.5.1 Spatial Scaling of Spectral Signals
5 Synthesis and Discussion
5.1 Phenological Phase: does phenology influence the spectral variability of dominant low Arctic vegetation communities?
5.2 Vegetation Colour: How does canopy-level vegetation colour relate to phenological changes in leaf-level photosynthetic pigment concentration?
5.3 Intrinsic Ecosystem Components: How does spatial aggregation of high spectral resolution data influence low Arctic tundra vegetation signals?
5.4 Key Innovations
5.5 Limitations and Technical Considerations
5.6 Outlook: Opportunities for Future Research
AWI Reading room