Estimating Tropical-Forest Biomass from the Fourier Transform of Lidar or InSAR Profiles: Improved Performance and a Paradigm for Lidar-InSAR DESDynI Fusion
Robert
Treuhaft, Jet Propulsion Lab, California Institute of Technology, robert.treuhaft@jpl.nasa.gov
(Presenting)
Fabio
Gonçalves, Jet Propulsion Laboratory, California Institute of Technology,, fabio.goncalves@jpl.nasa.gov
Jason
Drake, USDA Forest Service, jasondrake@fs.fed.us
Bruce
Chapman, Jet Propulsion Laboratory, California Institute of Technology, bruce.chapman@jpl.nasa.gov
João
Roberto
dos Santos, Instituto Nacional de Pesquisas Epaciais, jroberto@ltid.inpe.br
Luciano
Dutra, Instituto Nacional de Pesquisas Epaciais, dutra@dpi.inpe.br
Paulo
Graça, Instituto Nacional de Pesquisas de Amazônica, pmlag@inpa.gov.br
George
Purcell, Jet Propulsion Laboratory, California Institute of Technology, george.purcell@jpl.nasa.gov
Biomass based on remotely sensed forest structure from lidar or InSAR is usually estimated by constructing a model of the form
'Measured structural parameters' in the above expression have often been based on mean forest height, height of median energy, total height, or standard deviation. This poster introduces the possibility of using the Fourier transform of vertical vegetation profiles instead of height in biomass regression, while showing the formal similarity of the lidar Fourier transform and the complex coherence of interferometric SAR (InSAR), with each vertical spatial frequency in lidar corresponding to an InSAR baseline. It is first shown that the use of average height or standard deviation is tantamount to using only very low Fourier frequencies.
The performance of height-based estimation is compared to that using 3 Fourier frequencies for the tropical wet forests of La Selva, with LVIS data. The use of Fourier transforms improves the scatter of model biomass about field-measured biomass by about 40%. For InSAR C-band, multibaseline data taken at La Selva, Fourier methods improve performance over that of average-height methods by about 25%. It is also argued that C-band InSAR profiles, presented either in the space domain (the usual presentation) or in the frequency domain, are close to those of wide-footprint (25 m) lidar.
The abovementioned similarity of the Fourier transform of lidar and the complex coherence of InSAR suggests a mode of data fusion in the frequency domain. In this mode, Fourier transforms of lidar would be obtained at spatial frequencies different than those of InSAR. A study is currently underway to characterize the spatial correlation of different lidar Fourier transforms. If correlation lengths do have spatial-frequency dependences, then those with the highest correlation lengths could, for example, come from the lidar, while the shorter ones could come from InSAR. In this way, the dense Fourier frequency content of lidar, with its sparse spatial coverage, might be merged with the dense spatial coverage of InSAR, with its sparse number of Fourier frequencies (baselines).
Presentation Type: Poster
Poster Session: Carbon Cycle Science
NASA TE Funded Awards Represented:
Treuhaft, Robert
Toward Regional 3-Dimensional Tropical-Forest Structure and Biomass from Interferometric SAR
Treuhaft, Robert
The Performance of Structure and Biomass Estimation from InSAR 3-D Vegetation Missions at L-band over Tropical Forests