## Algorithm for extracting vegetation height from LiDAR

The procedure developed during the project to analyse LIDAR data aims at providing estimates of low vegetation, namely lower than the smallest observable difference between the first and the last detected return from the same location on ground. The procedure relies on the principle that the waveform of the backscatter signal depends on the vertical distribution of leaves within the observed vegetation canopy. The estimated travel time of the laser pulse will then be slightly different according to vegetation height. Local statistics of a sample of LIDAR measurements provide information on canopy architecture and vegetation height.

The procedure as implemented can be applied to samples of 50 to 100 individual laser measurements (raw data) constructed by using a polygon of arbitrary shape or a short segment of a linear sequence of laser measurements. In either case the reference points need to be geo-located. The frequency distribution of laser measurements in each sample provides information in different ways: difference between maximum observed elevation and ground elevation determined with GPS observations;-difference between maximum and minimum observed elevation;-standard deviation of elevation values.

The procedure has been validated in two ways:

A.For a set of reference polygons, GPS observations provided the elevation of points along the contour line of the polygon. The mean and maximum GPS elevation was in good agreement with minimum LIDAR elevation in the sample.

B. A simple simulation model was applied to estimate the expected laser above ground elevation of a vegetation canopy, given the vertical distribution of leaves within the canopy.

The mean values of observed elevation for all reference polygons were in good agreement with simulated elevation, assuming a uniform leaf distribution between the top of plants and the ground surface. The procedure can also be applied to refine maps of ground elevation by re-sampling a map of uncorrected elevation data, with the corrected elevation set equal to the minimum elevation in a sample of uncorrected data. Samples are constructed by running a box filter through the uncorrected data.

The procedure as implemented can be applied to samples of 50 to 100 individual laser measurements (raw data) constructed by using a polygon of arbitrary shape or a short segment of a linear sequence of laser measurements. In either case the reference points need to be geo-located. The frequency distribution of laser measurements in each sample provides information in different ways: difference between maximum observed elevation and ground elevation determined with GPS observations;-difference between maximum and minimum observed elevation;-standard deviation of elevation values.

The procedure has been validated in two ways:

A.For a set of reference polygons, GPS observations provided the elevation of points along the contour line of the polygon. The mean and maximum GPS elevation was in good agreement with minimum LIDAR elevation in the sample.

B. A simple simulation model was applied to estimate the expected laser above ground elevation of a vegetation canopy, given the vertical distribution of leaves within the canopy.

The mean values of observed elevation for all reference polygons were in good agreement with simulated elevation, assuming a uniform leaf distribution between the top of plants and the ground surface. The procedure can also be applied to refine maps of ground elevation by re-sampling a map of uncorrected elevation data, with the corrected elevation set equal to the minimum elevation in a sample of uncorrected data. Samples are constructed by running a box filter through the uncorrected data.