Shape from spectra

Nimrod Carmon*, Alexander Berk, Niklas Bohn, Phillip G. Brodrick, Jeff Dozier, Margaret Johnson, Charles E. Miller, David R. Thompson, Michael Turmon, Charles M. Bachmann, Robert O. Green, Regina Eckert, Elliott Liggett, Hai Nguyen, Francisco Ochoa, Gregory S. Okin, Rory Samuels, David Schimel, Joon Jin Song, Jouni Susiluoto

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


We introduce a new unified atmospheric–topographic correction approach that estimates surface geometry directly from the radiance measurement. Surface topography influences the at-sensor radiance measurement, making precise topography modeling critical in applications like vegetation or snow studies in mountainous terrain. Currently, elevation maps are used to derive topographic variables such as the slope and sky-view factor. This process is error-prone since static global digital elevation models do not generally achieve the accuracy required, and even minor mismatches in spatial resolution can introduce significant artifacts in downstream processing. Here we demonstrate that it is possible to estimate topographic parameters directly from spectral data, ensuring perfect physical consistency, temporal coincidence, and spatial alignment. We present experiments estimating topographic slope in two scenes in Southern California, with data from NASA's Next Generation Airborne Visible/Near Infrared Imaging Spectrometer (AVIRIS-NG). We compared our radiance-based estimates against high-resolution lidar datasets. Our initial validation result showed a correlation of R2=0.864 (n=160) over the homogeneous surface of Beckman Auditorium's cone-shaped roof on the Caltech campus in Pasadena, California. We then validate the model over a larger study site near Santa Clarita, California, finding R2=0.923 (n=40,000) in a 350×350 m area. The accuracy of our model estimates, combined with its systematic advantages over the alternative, show the potential of the approach for use in both airborne campaigns and orbital missions.

Original languageEnglish
Article number113497
JournalRemote Sensing of Environment
StatePublished - 1 Apr 2023
Externally publishedYes


  • Atmospheric correction
  • Imaging spectroscopy
  • Orbital imaging spectroscopy
  • Reflectance retrievals
  • Surface biology and geology
  • Topographic correction
  • Topography


Dive into the research topics of 'Shape from spectra'. Together they form a unique fingerprint.

Cite this