Geoscience Reference
In-Depth Information
was interpolated between measured values. Under partially cloudy conditions, the operator stopped
recording photon flux during cloud eclipse of the solar beam.
Operators performed a check on the data in the field immediately after download to a portable
computer. Typically, this involved plotting the PPFD in graphical form and comparing the number
of segments collected to the number of 10-m intervals traversed. An important quality assurance
measure was the use of paper and computer forms for data entry. To ensure that all relevant ancillary
data (i.e., weather conditions, transect orientation, operator names, data file names) were captured
in the field, operators filled out paper forms on-site for TRAC, hemispherical photography, and
biometric measurements. These data forms were then entered into a computer database via pre-
scribed forms, preferably immediately after data collection. This was a simple but valuable step to
ensure that critical data acquisition and processing parameters were not inadvertently omitted from
field notes. The computer forms provided a user interface to the relational database containing all
the metadata for the APB project.
4.3.4
Hemispherical Photography
Two Nikon Coolpix 995 digital cameras with Nikon FC-E8 fish-eye converters were used in
conjunction with TRAC at all six APB research sites. Exposures were set to automatic with normal
file compression (approximately 1/8) selected at an image size of 1600
1200 pixels. Hemispherical
images were not collected while the sun was above the horizon, unless the sky was uniformly
overcast. Images were primarily captured at dawn or dusk to avoid the issue of nonuniform
brightness, resulting in the foliage being “washed out” in the black-and-white binary image.
The camera was mounted on a tripod and leveled over each wooden stake along each A through
E photo transect. The height of the camera was adjusted to approximately breast height (1.4 m)
and leveled to ensure that the “true” horizon occurred at a 90˚ zenith angle in the digital photographic
image. The combination of two bubble levelers, one mounted on the tripod and the other on the
lens cap, ensured the capture of the “true” horizon in each photograph. Using a hand-held compass,
the camera was oriented to true north so that the azimuth values in the photograph corresponded
to the true orientation of the canopy architecture in the forest stand. Orientation did not affect any
of the whole-image canopy metrics (i.e., LAI, canopy openness, or site openness) calculated by
GLA. However, comparison of metrics derived by hemispherical photography, TRAC, densiometer,
or forest mensuration measurements required accurate image orientation.
After the images were captured in the field they were downloaded from the camera disk, placed
in a descriptive file directory structure, and renamed to reflect the site and transect point. A GLA
configuration file (image orientation, projection distortion and lens calibration, site location coor-
dinates, growing-season length, sky-region brightness, and atmospheric conditions) was created for
each site. Next, images were registered in a procedure that defined an image's circular area and
location of north in the image. Image registration entailed entering pixel coordinates (image size-
and camera-dependent) for the initial and final X and Y points. The FC-E8 fish-eye lens used in
this study had an actual field of view greater than 180˚ (~185˚). The radius of the image was
reduced accordingly so that the 90˚ zenith angle represented the true horizon. Frazer et al. (2001)
described the procedure for calibrating a fish-eye lens. Calibration results were entered into the
GLA configuration file (Canham et al., 1994).
The analyst-determined threshold setting in GLA adjusted the number of black (“obscured”
sky) and white (“unobscured” sky) pixels in the working image. This was perhaps the most
subjective setting in the entire measurement process and potentially the largest source of error in
the calculation of LAI and other canopy metrics from hemispherical photographs. As a rule of
thumb, the threshold value was increased so that black pixels appeared that were not represented
by canopy elements in the registered color image. The threshold was then decreased from this point
until the black dots or blotches disappeared and the black-and-white working image was a reason-
able representation of the registered color image (Frazer et al., 1999).
¥
