Geoscience Reference
In-Depth Information
4.3.5
Hemispherical Photography Quality Assurance
The height at which each hemispherical photograph was taken represented a potential source
of positional errors (~ 5 to 10 cm). At relatively level sampling points, the tripod legs and center
shaft were fully extended to attain a height that approximated breast height. However, at sites with
steep and/or uneven slopes, the camera height may have varied between repetitive measurement
dates due to variations in the extension of the tripod legs, possibly resulting in inclusion or exclusion
of near-lens vegetation.
Several comparisons of hemispherical photographic estimates of LAI with direct estimates in
broad leaf and conifer forest stands have been reported (Neumann et al., 1989; Chason, 1991; Chen
and Black, 1991; Deblonde et al., 1994; Fassnacht et al., 1994; Runyon et al., 1994). These
comparisons all showed that there was a high correlation between the indirect and direct methods,
but the indirect methods were biased low. This was because the clumping factor was not accounted
for using a random foliar distribution model (Chen et al., 1991).
To assess analyst repeatability, a set of 31 hemispherical photographic images collected in
eastern Oregon were analyzed and threshold values charted using SAS QC software (SAS, 1987).
Two analysts in the APB study repeatedly analyzed the 31 images to develop an ongoing quality
control assessment of precision compared to the Oregon assessment.
4.4 DISCUSSION
4.4.1
LAI Accuracy Assessment
Chen (1996) provided an estimate of errors in optical measurements of forest LAI using
combined TRAC and LiCOR 2000 PCA instruments. We assumed that the PCA was equivalent to
digital hemispherical photography for this discussion. Chen states that, based on error analysis,
carefully executed optical measurements can provide LAI accuracies of close to or better than 80%
compared to destructive sampling. The approximate errors accumulated as follows: PCA measure-
ments (3 to 5%); estimate of needle-to-shoot area ratio (
) (5 to 10%); estimate of foliage element
clumping index (3 to 10%); estimate of woody-to-total area ratio (5 to 12%). These factors sum
to an approximate total error of 15 to 40% in ground-based optical instrument estimates of LAI.
Chen (1996) also reports that the highest accuracy (~ 85%) (relative to destructive sampling)
“can be achieved by carefully operating the PCA and TRAC, improving the shoot sampling strategy
and the measurement of woody-to-total area ratio.” A crucial issue for this analysis was to better
understand the robustness of published values of needle-to-shoot area ratio (
g
g
) and woody-to-total
area ratio (
), because direct sampling of these quantities was logistically infeasible in this research
effort. Published values have been used in this analysis (Leblanc et al., 2002).
a
4.4.2
Hemispherical Photography
Figure 4.10 presents a chronosequence of hemispherical photographic images taken at the
midpoint (50 m) of the C transect at the Hertford site at five different dates in 2002. The images
were the registered black-and-white bitmap images produced by GLA. The date and LAI Ring 5
values were displayed to the right of each image. LAI Ring 5 represented a 0˚ to 75˚ field of view.
In the March 5, 2002, image, near-lens understory foliage was observed in the lower-left portion.
However, in subsequent images, the large-leafed obstruction was absent. The reason for the disap-
pearance of this understory image component was unclear. The tripod height may have been adjusted
to place the camera above the near-lens foliar obstruction, or perhaps field-crew effects may have
resulted in the disappearance of the obstruction. The presence of the near-lens foliage in the March
5 image may account for the somewhat elevated LAI value before leaf-out.
