Forest structure without ground data: Adaptive Full-Blind Multiple Forward-Mode reflectance model inversion in a mountain pine beetle damaged forest

Abstract

A new approach for using canopy reflectance models (CRMs) is presented that requires no field data or knowledge about the study area or imagery. Multiple Forward-Mode Adaptive Full-Blind (MFM-AFB) modelling provides forest biophysical structural information (BSI), and can also be used for classification and spectral mixture analysis at sub-pixel scales without user-specified model inputs, training data or endmember spectra, as these are instead automatically derived. In an example application using 2007 Landsat imagery of forest damaged by a mountain pine beetle (MPB) epidemic in British Columbia, Canada, overall BSI accuracy was within ±1000 stems ha^–1 for stand density, ±0.5 m for crown radius and ±1 m tree height for healthy and MPB stands. MFM-AFB software is suitable for regional, multi-temporal and unknown imagery and areas. By not requiring user-specified a priori model inputs to infer BSI, the MFM-AFB approach may help enable mainstream use of diverse and advanced CRMs for image analysis.

Acknowledgements

We gratefully acknowledge NSERC, AICWR, PARC, NRCan, WestGrid, Boston University, CFS-MPBI, S. Soenen (University of Lethbridge), P. Teti and D. Lewis. The Editor and reviewers are thanked for their comments that helped to clarify the presentation.