A fuel-type map of a predominantly shrub-land area in central Portugal was generated for a fire research experimental site, by combining airborne light detection and ranging (LiDAR), and simultaneous color infrared ortho imaging. Since the vegetation canopy and the ground are too close together to be easily discerned by LiDAR pulses, standard methods of processing LiDAR data did not provide an accurate estimate of shrub height. It was demonstrated that the standard process to generate the digital ground model (DGM) sometimes contained height values for the top of the shrub canopy rather than from the ground. Improvement of the DGM was based on separating canopy from ground hits using color infrared ortho imaging to detect shrub cover, which was measured simultaneously with the LiDAR data. Potentially erroneous data in the DGM was identified using two criteria: low vegetation height and high Normalized Difference Vegetation Index (NDVI), a commonly used spectral index to identify vegetated areas. Based on the height of surrounding pixels, a second interpolation of the DGM was performed to extract those erroneously identified as ground in the standard method. The estimation of the shrub height improved significantly after this correction, and increased determination coefficients from R2 = 0.48 to 0.65. However, the estimated shrub heights were still less than those observed in the field.
Spectral mapping methods applied to LiDAR data: Application to fuel type mapping
