Thomas Hilker

Simulation of multiangular remote sensing products using small satellite formations

To completely capture the multiangular reflectance of an opaque surface, one must estimate the bidirectional reflectance distribution function (BRDF), which seeks to represent variations in surface reflectance as a function of measurement and illumination angles at any time instant. The gap in angular sampling abilities of existing single satellites in Earth observation missions can be complemented by small satellites in formation flight. The formation would have intercalibrated spectrometer payloads making reflectance measurements, at many zenith and azimuthal angles simultaneously. We use a systems engineering tool coupled with a science evaluation tool to demonstrate the performance impact and mission feasibility. Formation designs are generated and compared to each other and multisensor single spacecraft, in terms of estimation error of BRDF and its dependent products such as albedo, light use efficiency (LUE), and normalized difference vegetation index (NDVI). Performance is benchmarked with respect to data from previous airborne campaigns (NASAs Cloud Absorption Radiometer), and tower measurements (AMSPEC II), and assuming known BRDF models. Simulations show that a formation of six small satellites produces lesser average error (21.82%) than larger single spacecraft (23.2%), purely in terms of angular sampling benefits. The average monolithic albedo error of 3.6% is outperformed by a formation of three satellites (1.86%), when arranged optimally and by a formation of seven to eight satellites when arranged in any way. An eight-satellite formation reduces albedo errors to 0.67% and LUE errors from 89.77% (monolithic) to 78.69%. The average NDVI for an eight satellite, nominally maintained formation is better than the monolithic 0.038.

Modeling Gross Primary Production for Sunlit and Shaded Canopies Across an Evergreen and a Deciduous Site in Canada

Light use efficiency (LUE) models offer an effective way for regional gross primary productivity (GPP) estimation. However, LUE is not easily determined at the landscape level due to its complexity and dependence on various environmental factors. One possible strategy to avoid the requirement for assessing environmental stressors is using the photochemical reflectance index (PRI) to determine LUE via the epoxidation state of the xanthophyll cycle. Integration of such measurements into GPP models could lead to more realistic GPP estimates of landscape level. Conventional, “one-leaf” LUE models, however, seem less suitable for integration of such remote sensing observations, as optically derived estimates are dependent on the shadow fraction viewed at a given time. Here, we utilize the two-leaf LUE (TL-LUE) model to parameterize LUE from multiangle PRI observations and compare it with MOD17 approach. Significant relationships were found between LUE (LUE, LUEsun, and LUEshaded) and PRI (PRI, PRIsun, and PRIshaded) over 8- and 16-day time steps. Similarly,

Remote sensing of seasonal light use efficiency in temperate bog ecosystems

R^{2}

An Approach for Determining Relationships Between Disturbance and Habitat Selection Using Bi-weekly Synthetic Images and Telemetry Data

values for the relationships between modeled GPP and observed GPP (EC derived measurements of GPP) were 0.87 (TL-LUE) and 0.81 (MOD17) at deciduous forest and 0.54 (TL-LUE) and 0.46 (MOD17) at evergreen forest for eight-day periods, as well as 0.84 (TL-LUE) and 0.74 (MOD17) at deciduous forest and 0.49 (TL-LUE) and 0.46 (MOD17) at evergreen forest for 16-day periods. Our results are relevant when planning potential future satellite missions to help constrain existing GPP models using remotely sensed data, as such observations will likely be affected by canopy shading effects at the time of observation.

Spectral analysis of amazon canopy phenology during the dry season using a tower hyperspectral camera and modis observations

Despite storing approximately half of the atmosphere’s carbon, estimates of fluxes between wetlands and atmosphere under current and future climates are associated with large uncertainties, and it remains a challenge to determine human impacts on the net greenhouse gas balance of wetlands at the global scale. In this study we demonstrate that the relationship between photochemical reflectance index, derived from high spectral and temporal multi-angular observations, and vegetation light use efficiency was strong (r2u2009=u20090.64 and 0.58 at the hotspot and darkspot, respectively), and can be utilized to estimate carbon fluxes from remote at temperate bog ecosystems. These results improve our understanding of the interactions between vegetation physiology and spectral characteristics to understand seasonal magnitudes and variations in light use efficiency, opening new perspectives on the potential of this technique over extensive areas with different landcover.

New approaches in multi-angular proximal sensing of vegetation: Accounting for spatial heterogeneity and diffuse radiation in directional reflectance distribution models

Ecological studies can be limited by the mismatch in spatial-temporal scales between wildlife GPS telemetry data, collected sub-hourly, and the large-area maps used to identify disturbances, generally updated annually. Recent advancements in remote sensing, data fusion modeling, mapping, and change detection approaches offer environmental data products representing every 16-day period through the growing season. Here we highlight opportunities and challenges for integrating wildlife location data with high spatial and temporal resolution landscape disturbance data sets, available from remotely sensed imagery. We integrated 16-day outputs from the Spatial Temporal Adaptive Algorithm for mapping Reflectance Change (STAARCH) disturbance maps with grizzly bear (Ursus arctos) telemetry data. Our results indicate that males and females avoided same-year disturbances, while male bears were most likely to avoid recently disturbed areas in summer. When intra-year (disturbances mapped at a 16-day time-step) analysis of disturbance was compared to traditional annual time-step analysis, annual aggregation of disturbance data resulted in an increase in the observed selection of same-year disturbed habitat, although change was not statistically significant (α 0.05). We caution the use of low-temporal resolution disturbance data to evaluate short-term impacts on wildlife and highlight the need for further development of probabilistic- and model-based techniques for overcoming spatial-temporal differences between datasets.