The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
245
1991; Dawson and Curran, 1998). Derivative analysis of the
near contiguous bands of field spectra enable the identification
of narrow waveband ratios which are tightly coupled to changes
in photosynthetic function resulting from vegetation stress
(Zarco-Tejada et al., 2003). First derivatives were computed for
all spectra, using the finite divided difference approximation
method (Chapra and Canale, 1988; Tsai and Philpot, 1998). The
derivative peaks were used to select two vegetation band ratios:
the vegetation stress ratio of Smith et al. (2004) (exploiting a
double peak in the red-edge) and a modified version of this
ratio.
The vegetation stress ratio of Smith et al. (2004) exploits the
magnitude of the first derivative at 725 and 702 nm within the
red-edge, which form the mid-point of the red-edge peak
maximum and its shoulder. The ratio has enabled identification
of stress due to short-term sub-surface gas leaks in grass, and
long-term sub-surface gas leaks in winter wheat and bean crops
at the canopy scale under controlled field test site conditions.
The vegetation stress ratio is obtained by Smith et al. (2004):
125nm
702«w
Where 725 nm is the value of the first derivative of the
reflectance spectrum at 725 nm and 702 nm is the value of the
first derivative of the reflectance spectrum at 702 nm. The
modified vegetation stress ratio takes the form:
723nm
700nm
Where 723 nm is the value of the first derivative of the
reflectance spectrum at 723 nm and 700 nm is the value of the
first derivative of the reflectance spectrum at 700 nm.
2.4 Soils acquisition and analysis
Soil acquisition was conducted in May 2006 to coincide with
field and airborne hyperspectral acquisitions, and to minimise
crop disturbance. Sub-soil sampling depth was kept as close as
possible to 0.1 - 0.2 m. Soil was extracted at random locations
within each of the disturbance zones producing a composite
sample for each zone of approximately 500 g (Figure 2). The
samples were analysed for organic carbon, Potassium,
Phosphorus and pH, to establish if differences in soil fertility
were evident between disturbed and undisturbed soil. Indirect
measurements of soil strength, structure and compaction were
obtained by penetrometer measurements and bulk density core
sample extraction in order to validate the penetrometer readings
(Bradford, 1986).
3. RESULTS AND DISCUSSION
Field spectroradiometry data were acquired during four field
visits in June 2005 and May 2006. A pseudo RGB CASI-2
image overlaid with a red-edge position (REP) blue shift
classification (Guyot and Baret, 1988) and pipeline route was
used to identify locations where potential vegetation stress
coincided with the 508 mm pipeline for the 2005 campaign
(Figure 1). Two of the sites of spring barley fields selected as
test sites in 2005 are labelled 05B and 05C in Figure 1.
Different cropping regimes prevented the use of the same test
sites in 2006. New test sites were identified for 2006 based on
expert field knowledge from local farmers (Pers. Comm. Donal
Cullen, Bruce Mackie and Gerald Banks). One of the sites
selected as a test site in 2006, labelled 06A, a winter barley
field, (Figure 1) provided the most conclusive results and is the
focus of discussion in this paper.
Notable differences in reflectance for crop stress transects are
exhibited for winter barley at test site 06A. An increase in
reflectance is evident with proximity to the pipeline of up to
-2% at the green peak between 10 and 60 m West (Figure 3). A
blue shift in the red-edge of ~5 nm is evident with proximity to
the pipeline at 25% reflectance. Similar blue shifts in the red-
edge for stressed vegetation have also been reported by Jago et
al. (1999) and Lelong et al. (1998). Pronounced steps in
reflectance occur in the NIR with proximity to the pipeline
range between -19% at 0 m at peak C and -48% at 70 m West
at peak B. The steps in the NIR are most likely to be a direct
result of the proportion of soil background to barley leaf
coverage, lower NIR reflectance values corresponding with
increased soil background proportions, particularly at 0 m and
60 m (Figure 3). The same reflectance responses in the VIS and
NIR were also observed by Smith et al. (2004) for soil gassed
grass, bean and winter wheat.
Exceptions to this trend occur at 10 m and 20 m West, which
display high reflectance throughout the wavelengths sampled,
with maximum reflectances of -54% and -48% at peak B
(Figure 3). The higher reflectance values could be attributed to
poor atmospheric conditions during acquisition, when cumulus
cloud intermittently obscured the Sun. Another exception is at
60 m West, which has comparatively low reflectance of-18%
at peak C (Figure 3). This is likely due to a combination of
increased soil background within the FOV of the ASD
spectroradiometer and crop stress (Figure 3).