4, 9-71 Nov. 1999 
t are contiguous or even 
thus fully illuminate the 
'rage canopy structure that 
small footprint altimeters. 
ts typically contains some 
ius consistently yielding a 
ire of vegetation height for 
oh flight altitude minimizes 
eceived backscatter energy 
tance due to topographic 
etation. Accurate pointing 
jated geolocation software, 
> location of each footprint 
tly correlated with ground 
nages. SLICER's control 
> designed to be flexible so 
otprint size, spacing and 
1 of canopy structure could 
ument characteristics are 
[CER data. First, the laser 
| uniform and thus canopy 
| unequally. The pattern of 
footprints that each have a 
ser energy yields a swath 
an inverted egg carton. 
corded in the waveform is 
laser energy. The transfer 
1ved by the instrument (i.e. 
resulting digital count 
nown due to uncalibrated 
function varies spatially, as 
e swath, and temporally on 
| of operating conditions. 
can not be compared in an 
only be used as a relative 
1 of backscattered energy 
detection scheme to define 
arget within a footprint. 
canopy top requires that 
received exceeding the 
tter intensity depends on 
'd (NIR) reflectance of the 
le. Thus SLICER's ability 
ulting derivation of canopy 
f the outer canopy surface 
ents making up the outer 
conifer tips with NIR-dark 
n a concentration of NIR- 
well defined, umbrella-like 
canopy characteristics, the 
ight can be biased low to 
'uter-most canopy surface. 
International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999 
3. WAVEFORM LIDAR PROCESSING ALGORITHMS 
Waveform Height Algorithm 
To process the waveform into an estimate of the total height of 
the vegetation sampled, a sequence of processing steps is 
applied to the raw waveform data (Lefsky, 1997, Harding et al., 
Submitted). First, to improve the signal-to-noise ratio of the 
distribution, the raw amplitudes (Fig. la) are summed by 
accumulating the signal in adjacent waveform bins. Generally, 
either 6 or 9 adjacent bins have been summed, yielding either a 
66 or 99 cm vertical sampling. The 66cm vertical sampling is 
approximately equal to the vertical resolution defined by the 
laser pulse width and detector response, while the 99 cm 
sampling has been used for compatibility with non-lidar field 
measurements. Next, the mean and variance of the background 
noise is established using the final portion of the waveform, 
beyond any potential last ground return. The mean background 
noise is subtracted from the summed distribution yielding 
signal above the noise level (Fig. 1b). 
  
  
1 
20 
  
Distance Below Canopy Top (meters) 
  
  
  
  
  
  
  
| | | | | 
ü 50 0 200 a 1 7 fn 0.04 
Eneitzer Summed Cumulative Fraction of 
Counts Counts — Distributions Plant Area 
  
Fig. 1. Steps in converting a raw SLICER waveform to various 
processing levels (Panels a-d left to right) Harding et al., 
(In Prep.) 
We then distinguish the ground reflection in the signal by 
assuming that it is the last return above noise. The end of the 
last return is defined as the last signal above a threshold that is 
a multiple of the background noise variance (Fig. 1b). The 
peak of the last return is defined to be the first inflection in 
signal strength prior to the end of the last return, identified 
using its first derivative. The start of the last return can not be 
uniquely identified from the raw distribution because 
backscatter return from low vegetation could be convolved in 
time with the ground return. Therefore, the start of the last 
return is identified based on the width characteristics of the 
system impulse response. The impulse response is the 
theoretical signal recorded from a smooth and flat surface and 
depends on the convolved effects of pulse width and detector 
response. The SLICER impulse response is established from 
water surface returns. A ratio is determined for the impulse 
response between the width from the signal end to peak as 
compared to the width from peak to start. The observed end-to- 
peak width of the last return is scaled by this ratio in order to 
define the start position of the last return. This method 
accounts for any pulse broadening of the last return due to 
slope or roughness of the ground within the footprint. After 
automated identification of the last returns, the results are 
interactively evaluated, and modified where necessary, by 
examining profile plots of last return start, peak, and end 
elevations. Anomalous variations in elevation or last return 
width, either along or across the SLICER swath, reveal 
improperly identified ground returns that are then manually 
corrected. Recently, morphological filtering operations have 
been applied to the problem of correcting anomalous ground 
return estimates, eliminating the need for manual correction 
The amplitude of the ground reflection is then scaled to account 
for the difference between average canopy and ground NIR 
reflectance at 0° phase angle. In the existing work with 
SLICER, the ground return amplitude was increased by a factor 
of two based on the assumption that the reflectance of the 
ground, dominantly comprised of leaf-litter with some bare soil 
and rare live foliage, was half that of the canopy. The results of 
our work have been relatively insensitive to potential errors in 
this reflectance scaling factor, as described in Harding et al., 
(Submitted). 
Intercepted Surfaces / Transmittance Algorithm 
A cumulative height distribution for the canopy return can then 
be calculated, normalized by the adjusted total return (canopy + 
scaled ground), yielding a height distribution of canopy closure 
(Figure 1c). This Normalized Cumulative Power Distribution 
(NCPD) can be further transformed to estimate the vertical 
distribution of transmittance at the sensor orientation. The 
NCPD is used to estimate transmittance as follows: 
TsLicer(h)=1-NCDP, #1 
where Tsricer(h) is the SLICER estimate of transmittance at 
height h and NCPD(h+1) is the normalized cumulative power 
distribution at h+1. Since SLICER measures the reflectance of 
the laser at each height (assuming the contribution of multiple 
scattering to signal delay is small), Tsricer at a height is equal 
to one minus the cumulative reflectance from the height above 
it. In reality, transmittance is equal to one minus the sum of 
cumulative reflectance and absorbence. The NCPD can be 
used as a proxy for the sum of cumulative reflectance and 
absorbence when the ratio of absorbence to reflectance does not 
vary with height. Where the assumption of a constant ratio is 
violated, the error in Tsy;cgg compared to actual transmittance 
should be small because absorbence by foliage and needles at