Remote Sensing Science
Characterizing canopy nonrandomness with a multiband vegetation imager (MVI)
Article first published online: 21 SEP 2012
Copyright 1997 by the American Geophysical Union.
Journal of Geophysical Research: Atmospheres (1984–2012)
Volume 102, Issue D24, pages 29455–29473, 26 December 1997
How to Cite
1997), Characterizing canopy nonrandomness with a multiband vegetation imager (MVI), J. Geophys. Res., 102(D24), 29455–29473, doi:10.1029/97JD01175., , , and (
- Issue published online: 21 SEP 2012
- Article first published online: 21 SEP 2012
- Manuscript Accepted: 26 MAR 1997
- Manuscript Received: 8 MAY 1996
A new method for measuring plant canopy nonrandomness and other architectural components has been developed using a 16 bit (65535 gray scale levels) charged-coupled device (CCD) camera that captures images of plant canopies in two wavelength bands. This complete system is referred to as a multiband vegetation imager (MVI). The use of two wavelength bands (visible (VIS) 400–620 nm and near infrared (NIR) 720–950 nm) permits identification of sunlit and shaded foliage, sunlit and shaded branch area, clouds, and blue sky based on the camera's resolution, and the varying spectral properties that scene components have in the two wavelength bands. This approach is different from other canopy imaging methods (such as fish-eye photography) because it emphasizes measuring the fraction of an image occupied by various scene components (branches, shaded leaves, sunlit leaves) under different sky conditions rather than simply the canopy gap fraction under uniform sky conditions. The MVI has been used during the Boreal Ecosystem-Atmosphere Study (BOREAS) in aspen (Populus tremuloides) and balsam poplar (Populus balsamifera) to estimate architectural characteristics of each canopy. The leaf area index (LAI), sunlit LAI, and degree of nonrandomness within a canopy are architectural properties that have been measured with the MVI. Using a crown-based Monte Carlo model for nonrandom canopies, nonrandomness factors are calculated from MVI data using two approaches (gap fraction and gap-size distribution theories) to correct total and sunlit LAI estimates from indirect methods that assume random foliage distributions. Canopy nonrandomness factors obtained from analyzing the gap-size distribution in a Monte Carlo model are shown to be a function of path length (angle) through the canopy (Ωe(θ)); thus we suggest that LAI-2000 indirect measurements of LAI be adjusted with the value of Ωe(θ) at θ=35° because this is the mean angle at which the canopy gap fraction is measured by the LAI-2000. In this study, values of Ωe(35)=0.69 in an aspen forest. Alternatively, corrections to indirect LAI measurements obtained with the MVI in this study are made using the value of Ωe(0) because the MVI is used to measure the canopy gap-size distribution and gap fraction within 15° of the zenith. Values of Ωe(0) obtained with the MVI in aspen are typically between 0.55 and 0.65; while in balsam poplar, average values of Ωe(0) are equal to 0.82. This study shows that the MVI provides an attractive indirect measurement technique to obtain accurate estimates of total LAI in aspen. Corrected canopy LAI and direct LAI measurements are greater than indirect estimates based on assuming the foliage to be randomly distributed: In aspen, total LAI is 45% larger (3.3 versus 2.0) and sunlit LAI (40° Sun zenith angle) 10% larger, while in balsam poplar, total LAI is 17% larger (2.3 versus 1.9) and sunlit LAI is only 1% larger. The importance of these clumping characteristics is best appreciated with estimates of canopy net CO2 assimilation derived from scaling leaf photosynthesis versus light relations. Aspen canopy assimilation accounting for clumping is 39% larger than estimates based on indirect measurements of total LAI and the assumption that foliage is randomly distributed.