Natascha Kljun. Photo.

Natascha Kljun

Professor

Natascha Kljun. Photo.

Estimating forest canopy parameters from satellite waveform LiDAR by inversion of the FLIGHT three-dimensional radiative transfer model

Author

  • I.J. Bye
  • Peter R. J. North
  • S O Los
  • Natascha Kljun
  • Jean de la Rosette
  • C. Hopkinson
  • Laura Chasmer
  • Craig Mahoney

Summary, in English

The Geoscience Laser Altimeter System (GLAS) has the potential to accurately map global vegetation heights and fractional cover metrics using active laser pulse emission/reception. However, large uncertainties in the derivation of data products exist, since multiple physically plausible interpretations of the data are possible. In this study a method is described and evaluated to derive vegetation height and fractional cover from GLAS waveforms by inversion of the FLIGHT radiative transfer model. A lookup-table is constructed giving expected waveforms for a comprehensive set of canopy realisations, and is used to determine the most likely set of biophysical parameters describing the forest structure, consistent with any given GLAS waveform. The parameters retrieved are canopy height, leaf area index (LAI), fractional cover and ground slope. The range of possible parameters consistent with the waveform is used to give a per-retrieval uncertainty estimate for each retrieved parameter. The retrieved estimates were evaluated first using a simulated data set and then validated against airborne laser scanning (ALS) products for three forest sites coincident with GLAS overpasses. Results for height retrieval show mean absolute error (MAE) of 3.71 m for a mixed temperate forest site within Forest of Dean (UK), 3.35 m for the Southern Old Aspen Site, Saskatchewan, Canada, and 5.13 m for a boreal coniferous site in Norunda, Sweden. Fractional cover showed MAE of 0.10 for Forest of Dean and 0.23 for Norunda. Coefficient of determination between ALS and GLAS estimates over the combined dataset gave R2 values of 0.71 for height and 0.48 for fractional cover, with biases of −3.4 m and 0.02 respectively. Smallest errors were found where overpass dates for ALS data collection closely matched GLAS overpasses. Explicit instrument parameterisation means the method is readily adapted to future planned spaceborne LiDAR instruments such as GEDI.

Topic

  • Earth and Related Environmental Sciences

Keywords

  • GLAS/ICESat
  • Model inversion
  • Forest canopy parameters
  • FLIGHT
  • Monte Carlo radiative transfer model
  • Waveform
  • LiDAR

Status

Published

ISBN/ISSN/Other

  • ISSN: 0034-4257