Geoscience Reference
In-Depth Information
included
Spartina alterniflora
and
Juncus roemerianus
(saline marshes),
Spartina patens
(brackish
marsh), and
(fresh marsh). Our primary purpose was to devise simple opera-
tional field collection methods and postanalysis procedures that could detect and monitor canopy
structure differences between and within wetland types that might influence variability in the remote
sensing image data. Our goal was to improve reliability and accuracy of current classifications
based on remote sensor data, and as a consequence to extend the biophysical detail extractable
with remotely sensed data.
Our first objective was to test the light penetration measurements for reliability, accuracy and
comparability over time and space and within and between marsh types. A ceptometer device that
measures PAR along an 80-cm probe was chosen for the light penetration measurements. In addition
to the choice of measuring device, four variables related to field data collection and postdata analysis
design were examined with respect to fulfilling data reliability and accuracy but also to maximizing
the potential for operational use for remote sensing calibration and assessment of classification
accuracy. These variables were the (1) horizontal (planar) and (2) vertical (canopy profile) spatial
sampling frequencies, (3) description and possible exclusion of atypical canopy structures, and (4)
normalization of measurements at different sun elevations.
At each site, we used 30-m transects in the north and south and east and west directions. Initially,
sample locations were at the site center, the transect extremes, and midpoints; however, sample
variability indicated higher sample frequency was needed. Within the early testing, sampling protocol
along transects was changed to collecting light penetration measurements every 3 m. This 30-
Panicum hemitomon
¥
30-m site area and higher 3-m transect sampling frequency helped ensure more accurate depiction
of the local variability and more reliable mean and variance measures and matched or encompassed
the spatial resolution of most common resource remote sensing sensors. Similarly, after testing a
relative canopy profile sampling of top, middle, and bottom, the vertical sampling frequency was
standardized to light penetration measurements every 20 cm from the ground surface to above the
canopy. At each profile location, the above-canopy light recordings were used to normalize lower
canopy light recordings, transforming light absolute magnitudes to percentage penetrations. The
relative profile sampling (top, middle, bottom) did not adequately replicate the canopy light atten-
uation profile compared to the 20-cm sampling frequency, especially in fully formed canopies. The
20-cm sampling interval was selected to ensure our goals to increase the extractable canopy detail
and improve the predictability of canopy structure with remote sensing data.
To improve the reliability, accuracy, and detail of the canopy light attenuation data, at each profile
location the state of the sky condition (sunny or cloudy) and canopy structure (undisturbed, partial
gap, or completely lodged) was recorded. In this analysis, only sunny sky conditions were processed.
Observed differences relative to including or excluding disturbed canopy profiles were mostly attrib-
utable to the level of canopy disturbance. In most cases, excluding the disturbed profiles at each site
from the aggregate site light attenuation profile increased our ability to compare aggregate profiles
taken at the same site during multiple occupations. If the aggregate profiles were more comparable
within a site after exclusion, comparison between sites and marsh types also would improve with
exclusion. The inclusion of all profiles, each designated with a flag as to the sky and canopy condition,
allows us to view and analyze selected aspects (include or exclude) of the canopy variability and thus
greatly enhances our ability to understand and relate remote sensing reflectance to canopy structure.
Although more noticeable in
canopies, in all
marshes as sun zenith increased the rate of light fall-off with canopy depth increased. To ensure
comparability of aggregate site light attenuation profiles, canopy light penetration measurements
were normalized to a nadir sun zenith. The successfulness of the normalization seemed to be
associated with the preferred orientation of the marsh canopy. In the most vertical canopies such
as
Juncus roemerianus
and less in
Spartina patens
Juncus roemerianus
, the correction worked well up to a sun zenith around 60
∞
. In highly lodged
or horizontal canopies such as
Spartina patens
, the light penetration seemed less effective and
. Nontheless, normalized light attenuation profiles were more
consistent with the expected changes in canopy structure. Even in this normally highly lodged
restricted to sun zenith angles < 49
∞
