240
data for estimating LAI (Demetriades-Shah et al. 1990). High-spectral resolution data provide spectral
information that is not available in broad waveband data. Data analysis techniques can therefore exploit features
that give rise to a change in the shape of the spectrum. High-spectral resolution data, processed as first and
higher order derivatives, may suppress the effects of spectrally flat variation in background reflectance and the
inflexion point in the spectrum near 720nm, termed the ‘red-edge’ may provide information on canopy LAI.
The availability of high-spectral resolution data has been limited, but some progress has been made in
understanding the relationships between LAI and spectral response of agricultural crops (e.g., Malthus et al.
1993) however, only two studies have related high-spectral resolution data to the LAI of forest canopies. Gong
et al. (1992) used imaging spectrometer data to investigate the relationships between single wavebands, waveband
combinations and derivatives and the LAI of Ponderosa pine stands in Oregon. They found that spectral
derivatives were superior to the reflectance data for LAI estimation although no comparisons were made with
the NDVI. Curran et al. (1991) examined the relationship between the red-edge position and LAI in Slash pine
plantations in Florida but found no statistically significant correlation.
In this research, the aim was to conduct experiments to test following working hypotheses:
• The NDVI is an inappropriate tool for estimating forest LAI because it is sensitive to variation
in canopy structure and understorey reflectance.
• High-spectral resolution data can provide new tools for deriving accurate and robust estimates
of forest LAI which can be constructed through theory development and by experimentation.
3. EXPERIMENTAL WORK AND FIELD SITE
This paper provides a summary of the results of an experiment conducted in 1991 as part of the Multiple
Airborne Campaign (MAC) Europe and presents the preliminary findings of a second experiment which began
in 1993. The work was carried out in an upland coniferous forest in the Tywi valley, 35 km north of Swansea,
Wales and was centred on a 600km 2 forest plantation on the east side of the Llyn Brianne reservoir. The forest
is dominated by stands of Sitka spruce (Picea sitchensis ) and Lodgepole pine (Pinus contorta) planted between
1957 and 1964 at an average altitude of 450m. Variation in drainage gives rise to large between-stand differences
in leaf area index.
4. 1991 EXPERIMENT
MAC Europe was a NASA-funded remote sensing programme in which the Airborne Visible/InfraRed Imaging
Spectrometer (AVIRIS) and an airborne SAR (AIRSAR) were deployed, along with a number of other sensors,
over European test sites (Danson et al. 1992) to provide a unique opportunity to collect, at one site, the array
of data that will be available once the Polar Platform sensors are in orbit. Fifty one Sitka spruce plots were
surveyed and at each plot a range of biophysical variables were measured (Danson et al. 1992). LAI was
estimated using well-established techniques which involved the destructive sampling of 13 trees to determine a
relationship between tree girth and leaf area. This equation was used to determine plot LAI which was found
to vary from 3 to 12.
This paper is concerned with the use of data from a helicopter-mounted spectroradiometer, a Spectron
Engineering SE590, which records radiation in 252 narrow wavebands over the range 400-1100 nm with a
spectral resolution of 11 nm and a sampling interval of 2.8 nm (Milton and Danson 1991). A downward pointing
sensor head with a nominal 1 ° field-of-view was mounted on a short aluminium bar along with a camcorder and
camera. The helicopter used in the experiment was a four seater Bell JetRanger. A rear door was removed and
the instruments were positioned to view between the aircraft body and landing skids. Irradiance was measured
with a cosine-corrected sensor head fitted to a bracket and mounted above the helicopter cabin. The target plots
were marked using small helium-filled balloons which were inflated on the morning of the flight and floated
through the canopy.
The data were collected between 11:40 and 12:40 hrs on 17 July 1991 from an altitude of around 300m
above ground level. At this altitude the marker balloons were visible and the nominal ground area sampled was
5x5m. One irradiance spectrum followed by four radiance spectra were recorded for each target and spectra were
obtained for 16 of the plots.
The spectra for each plot were averaged to reduce the effect of noise and a non-optimised Fourier
transform algorithm was then used to low pass filter both the averaged radiance and irradiance data before
