HIGH-SPECTRAL RESOLUTION INDICES FOR FOREST LEAF AREA INDEX
F.M. DANSON, K.J. WINSTANLEY and A.M. STOCKS,
Telford Institute of Environmental Systems,
Department of Geography,
University of Salford,
Manchester M5 4WT, UK.
ABSTRACT:
The aim of the research was to test hypotheses on the relationships between high-spectral resolution reflectance
data and forest leaf area index (LAI). The results of an experiment undertaken in 1991 showed that the
normalised difference vegetation index (NDVI), measured using a helicopter-mounted spectroradiometer, was
positively related to plot LAI but that the strongest relationship was with the position of the ‘red-edge’. The
preliminary results of a second experiment, undertaken in 1993, are presented in which the feasibility of
collecting high-spectral resolution data from a helicopter in the 400-2500nm range was demonstrated.
KEY WORDS: Forest, LAI, high-spectral resolution, spectral indices, helicopter
1. INTRODUCTION
The key spatial variable required to drive forest ecosystem models over large areas is leaf area index (LAI), the
one-sided area of leaves per unit ground area (Running and Coughlan 1988). LAI is a quantitative measure of
the surface area available for the interception of radiation and for transpiration (Jarvis and McNaughton 1986)
and it is intimately related to the metabolic processes which control net primary production in vegetation
canopies. Routine estimation of forest LAI by remote sensing is an achievable goal but it will require significant
advances in our understanding of the relationships between the biophysical properties of forest canopies, such
as LAI, and their spectral response (Danson and Curran 1993). The aim of the work described in this paper was
to conduct a series of experiments to investigate these relationships in an upland coniferous forest plantation. 2
2. BACKGROUND
Remotely-sensed data have been used to estimate the LAI of a wide range of agricultural crops and grassland
canopies but work on forests canopies has progressed more slowly. This may be attributed to (i) the difficulty
in measuring forest LAI and (ii) the difficulty in obtaining spectral measurements over forest canopies (Milton
and Danson 1991). As a consequence, there have been very few studies undertaken to examine the relationships
between forest LAI and spectral response and most of these have employed simple vegetation indices such as
ratios of near infrared to red radiance or the normalised difference vegetation index (NDVI).
Work in the United States showed that the NDVI was related to forest LAI across an environmental
gradient in Oregon (Peterson et al. 1987) and later work confirmed this observation for stands with high canopy
cover in Montana and Oregon (Spanner et al. 1990). Curran et al. (1992) showed that seasonal variation in LAI
in Slash pine plantations in Florida could be monitored using Landsat Thematic Mapper-derived NDVI data and
Running et al. (1989) employed the NDVI, calculated from NOAA-AVHRR data, to estimate LAI across a forest
area in Montana and drive the FOREST-BGC model.
The NDVI, however, is known to be sensitive to change in the reflectance of the soil background or
understorey vegetation and to change in canopy structure. Spanner et al. (1990) found that the relationship
between the NDVI and forest LAI failed when canopy cover was low and there was a difference in understorey
reflectance. Similar conclusions were reached by Badwhar et al. (1986) in work on the seasonal variation of LAI
in Aspen canopies. These problems limit the usefulness of the NDVI for estimating forest LAI across large areas.
A response has been the development of complex ratios such as the Soil Adjusted Vegetation Index (Huete 1988)
and the inclusion of data in a broad middle infrared waveband to normalise the effect of variable forest canopy
cover on the NDVI (Nemani and Running 1993).
Another approach has indicated that high-spectral resolution data, where measurements of canopy
radiance are made in several hundred narrow contiguous wavebands, may offer advantages over broad waveband
239
