The use of a constant
LAD
does not fit the study objective of
producing both orientation and density structure indicators,
however, constant extinction coefficients are commonly em
ployed to simplify calculation of
LAI
so the comparison is use
ful. Equation 4 was used as before, however, precalculation
of an optimized
LAD
based on the measured
PAR
profile was
bypassed and a constant spherical distribution (
LAD
= 0.5) was
used. The predicted and observed
PAR
transmittance exhibited
high correspondence with a slope and bias of 1.00 and 0.0273,
respectively, and R
2
of 0.99 (
n
= 105). The resulting correspon
dence was similar to that based on optimized
LAD
values.
LAI
calculated by using a constant
LAD
value and
LAI
calcu
lated with optimized
LAD
values were directly compared. The
LAI
based on a constant (
LAD
= 0.5) explained 92 percent of
the
LAI
variance based on the optimized
LAD
(Figure 5a). The
relationship had a 0.70 slope and 0.067 intercept. Differences
in the two
LAI
calculations, however, increased with increasing
LAI
. The strength of covariance with measured total biomass
was also used as an indirect measure of which of the two
LAI
better represented marsh canopy structure. A simple regression
found
LAI
calculated with a constant
LAD
was a poor predictor
(slope = 111.2 ±80.4 gr/m
2
,
p
= 0.19, bias = 524.8 ±349.3 gr/m
2
,
p
= 0.15, and R
2
= 0.11) of total biomass (
RMSE
= ±506 gr/m
2
).
Although still low,
LAI
calculated with an optimized
LAD
was a
better predictor (slope = 240.5 ±124.0 gr/m
2
,
p
= 0.071, bias =
239.9 ±398.9 gr/m
2
,
p
= 0.56, and R
2
= 0.20) with a somewhat
lower total biomass prediction error (
RMSE
= ±480 gr/m
2
). The
results of these regressions are depicted in plots of the observed
versus predicted total biomass (Figures 5b and 5c). As shown in
Figure5a,
LAI
based on a constant
LAD
dominantly fell in a nar
row range between 3 and 5 while
LAI
based on the optimized
LAD
had a more uniform distribution extending from 0 to 5.
Description of Calculated KM and LAD and Associated LAI
Whether based on
KM
or
LAD
measures, canopy
LAI
varied
widely over time and from site to site (Table 1, Figure A1).
LAD
tended to be slightly lower than
KM
magnitudes due to its
adjustment by the sun zenith. Other than a single peak near
0.4,
LAD
values were nearly evenly distributed from just over
0.3 to a little over 1. The
KM
distribution exhibited a more
normal distribution ranging from 0.4 to 1.1. In contrast to
KM
and
LAD
relative magnitudes,
LAD

LAI
tended to have slightly
higher magnitudes than
KM

LAI
. Both exhibited fairly uniform
distributions. The bulk of
KM

LAI
values ranged from just
under 2 to a little below 5 with a single low value at 1.1.
LAD

LAI
ranged a bit higher from just over 2 to 5 with a low value
at 1.2. As in the measured biomass quantity and composition
distributions, the canopy structure parameterized as
LAI
and
KM
and
LAD
exhibited wide ranges and fairly uniform distri
butions well suited for accomplishing our objectives.
Discussion
We developed a methodology for producing threedimension
al marsh canopy vertical density and orientation structure di
rectly from field measurements of PAR transmittance without
reliance on operator inputs. The data driven parameterization
of the PAR profiles as standard canopy structural measures,
LAI and LAD, provides both nonbias estimates and a stan
dard calculation method based on accepted relationships
and published field protocols. The direct calculation of LAI
profiles and a LAD average in these complex wetland sys
tems offers a more meaningful representation of the marsh
that should improve the understanding and representation of
biophysical function, and importantly, provide variables more
amenable to remote sensing mapping (Zheng and Moskal,
2009), particularly polarimetric radar (Pairman
et al.
, 1999).
The final assessment of the methodology and produced den
sity and orientation variable, LAI and KM (or LAD), from the
PAR measurements is whether they are of reasonable magni
tudes and distribution and better represent the marsh canopy
structure as compared to simpler approximations.
LAI and LAD Compared to Published Values
Combining
LAI
ranges from seven grassland, prairie, and pasture
references produces an aggregate
LAI
range of 1.14 ±1.40 to 5.22
Figure 3. Total biomass dry weights listed in Table 1 plotted
against the siteaveraged bottom PAR measured.
Figure 4. PAR transmittance predicted with the implemented
methodology and the siteaveraged PAR transmittance profile
measured at all sites and times listed in Table 1 (note: only bot
tom PAR is listed in Table 1). Slight clustering near 1 is a result of
the standard abovecanopy reading at all sites.
812
October 2015
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING