Measuring canopy height with UAVSAR

Marc Simard, Caltech/Jet Propulsion Laboratory, marc.simard@jpl.nasa.gov (Presenter)
Scott Hensley, JPL, scott.hensley@jpl.nasa.gov
Marco Lavalle, JPL, marco.lavalle@jpl.nasa.gov
Ralph Dubayah, University of Maryland, dubayah@umd.edu
Naiara Pinto, University of Maryland, npinto@umd.edu

We used the UAVSAR, an airborne fully polarimetric L-band radar system, to estimate forest canopy height from repeat-pass interferometry. We present the UAVSAR campaign strategy and describe the methodology. Finally we validate the results with LVIS (Laser vegetation Imaging sensor) and field data.

The polinSAR (polarimetric interferometric synthetic aperture radar) model relies on the measurement of correlation and phase as different polarizations are sensitive to different scattering mechanisms. Essentially, cross-polarized signal (HV) is mainly sensitive to canopy while co-polarized signals (HH and VV) are strongly sensitive to ground. Thus, differences in the vertical location of the scattering phase centers (i.e. the height measurements) are used to infer canopy height. However, since UAVSAR is a repeat-pass interferometric system, slight changes within the canopy (e.g. due to weather and motion of scatterers) between radar acquisitions tend to decorrelate successive radar images. This effect can be taken into account within the polinSAR model but strongly depends on the assumed temporal decorrelation.

We designed the UAVSAR campaign to primarily quantify temporal decorrelation. Within a period of two weeks, we collected several days of data at different time intervals in order to sample several temporal baselines. Every flight, we collected two sets of interferometric pairs, a zero and 65m baseline, which should be affected by similar temporal decorrelation effects. The same set of tracks were repeated each flight day which enables us to form stacks of interferograms. This comprehensive experiment enabled quantification of the temporal decorrelation and its relationship to polarization, time intervals, weather patterns, terrain slope, forest type and canopy structure. We found the main factors influencing inSAR coherence are canopy structure and changes in weather.

The measurements of temporal decorrelation from the zero baseline pairs were used to correct the finite baseline interferograms and improve polinSAR canopy height estimates. We compared our results with LVIS estimates of canopy height and obtained an error of about 3m. These results are encouraging. However, we need to improve our algorithms with a priori knowledge of forest structure and include other potential decorrelation sources. We performed this analysis on a preliminary set of sites (boreal and temperate) and plan to expand our analysis to other temperate and tropical sites that were covered during the 2009-2010 UAVSAR campaign.

Presentation Type:  Poster

Session:  Other   (Tue 11:30 AM)

Associated Project(s): 

• Simard, Mac: 3D Vegetation Structure using L-band InSAR and Lidar ...details

Poster Location ID: 284