Canopy Vertical Spatial Scales which Constrain Biomass in a Tropical Forest at the Plot Level: Unifying Lidar and InSAR for Biomass Estimation
Abstract
Structural remote sensing of forest biomass, using lidar and/or interferometric synthetic aperture radar (InSAR), often involves regressing field measured biomass against remotely sensed characteristics of the vertical density profile. Because spaceborne lidar or InSAR sensors will estimate structural characteristics averaged at the plot level (0.04-1 hectare), and because tropical forests contain 40% of the Earth’s forested biomass, this study focuses on the scales of vertical characteristics which best correlate with tropical forest biomass. This work suggests that the structural characteristics used in both lidar and InSAR biomass estimation, such as mean height or total height or height of median energy, are based on the behavior of Fourier vertical frequency components of vegetation density near zero frequency; that is, they are very low-spatial frequency characteristics of the vertical vegetation distribution. In this work, we ask which other vertical Fourier frequencies in lidar- or InSAR-produced structure metrics can best correlate with field biomass. Using lidar (LVIS) data from La Selva Biological Station, Costa Rica, taken in 2005, lidar canopy observations are Fourier transformed in the vertical direction to decompose into vertical frequency components. Each baseline of an InSAR observation, the complex coherence, is this Fourier transform of the canopy, if the ground contribution can be neglected. Using the qualitative similarity in vertical profiles seen by lidar, InSAR (at C-band, from AirSAR in 2004), and field measurements in the La Selva data, we produce the equivalent many (1000’s of) InSAR baselines from the lidar data and, using the lidar-simulated InSAR, determine the optimal spatial frequencies—baselines at DESDynI orbital altitudes for InSAR—which would estimate biomass in this wet tropical forest most accurately for either technique. For biomass ranging from 39-490 Mg/ha, regressing field biomass against some function of height (average or total) yields RMS prediction scatters of 90 Mg/ha from either lidar or InSAR. By using higher spatial Fourier frequencies with vertical wavelengths between 100-200 m and 10-20 m (at the smallest), with up to 8 parameters describing the Fourier amplitudes and phases at each frequency, the RMS scatters decrease to about 60 Mg/ha from lidar, and 75 Mg/ha from the available InSAR baselines. Error budgets for each technique will attempt to account for the performance difference in biomass estimation. DESDynI relevant performance enhancements include optimal lidar design or operation modes to capture the 100-10m spatial frequencies. Optimal InSAR baselines will be considered for biomass estimation based on vertical Fourier decomposition. We also explore the potential of unified, quantitative lidar-InSAR fusion schemes which cull optimal, and possibly different, vertical frequency components from each technique.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2009
- Bibcode:
- 2009AGUFM.B24A..06T
- Keywords:
-
- 0426 BIOGEOSCIENCES / Biosphere/atmosphere interactions;
- 0428 BIOGEOSCIENCES / Carbon cycling;
- 0439 BIOGEOSCIENCES / Ecosystems;
- structure and dynamics;
- 0480 BIOGEOSCIENCES / Remote sensing