Continuous wavelet analysis for spectroscopic determination of sub-surface moisture and water-table height in northern peatland ecosystems

Michael Falkowski, University of Minnesota, mfalkows@umn.edu (Presenter)
Asim Banskota, University of Minnesota, banskota@umn.edu
Evan S Kane, USFS Northern Research Station, eskane@mtu.edu
Alistair Matthew Stuart Smith, University of Idaho, alistair@uidaho.edu
Laura Louise Bourgeau-Chavez, Michigan Tech Research Institute, lchavez@mtu.edu
Nancy HF French, Michigan Tech Research Institute (MTRI), nhfrench@mtu.edu

Remote sensing of near-surface hydrological conditions within northern peatlands can provide important information in understanding carbon cycling, fire vulnerability, and climate change impacts in these ecosystems. Climate change is altering water table height and near surface moisture conditions in northern peatlands, which in turn both increases the susceptibility to fire and reduces the carbon sink capacity of these ecosystem. Thus, it is imperative to develop tools for monitoring peatland hydrological parameters such as surface moisture content and water-table position. In an effort to further develop remote sensing based measurements of peatland moisture characteristics, we employed coincident surface reflectance and moisture measurements in two Sphagnum moss-dominated peatland sites (an experimentally manipulated site and an un-manipulated natural peatland).We applied a continuous wavelet transform using a Mexican hat wavelet to the measured spectra to generate wavelet features and coefficients across a range of scales. Genetic algorithm and correlation scalogram analysis were subsequently used to select meaningful wavelet features for characterizing peatland moisture content. We cross-compared the wavelet derived moisture estimates to estimates derived from two narrowband spectral indices and to ground truth data. Overall, wavelet analysis was improvement over the previously tested spectral indices at both study sites. Linear mixed effect models for water table height using wavelet features accounted for more of the variance with both an improved marginal R2 (29% greater) and a larger conditional R2 (21% greater) as compared to the best performing spectral index. Similarly, wavelet based models for surface moisture content provided lower errors than the previously tested spectral indices. The current study also revealed the advantage of selecting best subsets of wavelet features over a more widely used technique that selects features based upon correlation scalograms.

Presentation Type:  Poster

Session:  General Contributions   (Tue 4:35 PM)

Associated Project(s): 

• Bourgeau-Chavez, Laura: Vulnerability of North American Boreal Peatlands To Interactions Between Climate, Hydrology, and Wildland Fires ...details
• Falkowski, Mike: Fuel Consumption and Carbon Cycling in Northern Peatland Ecosystems: Understanding Vulnerability to Burning, Fuel Consumption, and Emissions via Remote Sensing of Fuel Moisture and Fire Radiative Energy ...details

Poster Location ID: 143