Spectral Discrimination of Cannabis sativa L. Leaves and Canopies

The growing of marijuana (Cannabis sativa L.) on pubon measured leaf spectra and simulated canopy refleclic lands poses problems to the environment and the pubtance spectra, we would choose several relatively narrow lic. Remote sensing offers a potential way of monitoring (i.e., 30 nm or less) spectral bands in the green (550 nm), public lands for the production of marijuana. However, red (670 nm), “red edge” (720 nm), and the near-infrared very little information on the spectral properties of mari(800 nm) to discriminate marijuana leaves and canopies juana is available in the scientific literature. Our objecfrom other species. Much of the leaf spectral information tives were to 1) characterize the spectral properties of the is also available in the canopy reflectance data. Publeaves of marijuana and various other plants that occur lished by Elsevier Science Inc., 1998 where marijuana is grown in the eastern United States, 2) simulate canopy reflectance, and 3) identify wavebands for discriminating marijuana from other plants. In INTRODUCTION a series of replicated field experiments, the basic factors The growing of cannabis or marijuana (Cannabis sativa affecting marijuana growth and reflectance, including L.) on public lands poses problems to the environment planting date, plant density, and N-fertilization were varand the public. Not only are unauthorized disturbances ied. Leaf optical properties were measured periodically to the environment created, more seriously, growers ofduring the growing season with a spectroradiometer and ten set booby traps or post armed guards to protect their integrating sphere. As N-fertilization rate decreased, the plots. This is especially important with respect to the use marijuana plants produced leaves with lower chlorophyll of our national forests by the public. Remote sensing ofconcentrations and higher reflectance values in the visifers a potential way of monitoring public lands for the ble wavelength region, particularly at 550 nm. The reproduction of cannabis. However, very little information flectance spectra of the herbaceous dicot species examon the spectral properties of marijuana is available in the ined were very similar to the spectrum of marijuana. The scientific literature. reflectance spectra of the monocots and the trees differed The spectral properties of vegetation and soils must significantly from the spectrum of marijuana, particube understood to identify plant species and to estimate larly in the green and near-infrared wavelength regions. plant productivity from remotely sensed data. When Canopy reflectance spectra of marijuana and several repdealing with remote sensing of specific plants, as in agriresentative species were simulated for a wide range of culture and forestry, the problem is interpreting the reLAI and background reflectances. Major differences in flected signal produced by the soil–plant–atmosphere canopy reflectance of marijuana and other plants were system. The vegetation of interest, the underlying strata observed near 550 nm, 720 nm, and 800 nm. Dense can(such as soil, plant litter, other types of vegetation, or opies of marijuana were more spectrally discriminable water, etc.), and the intervening atmosphere between the from other vegetation than sparse canopies. Thus, based target and the sensor contribute to the sensor response. However, because plant leaves contribute most of the signal from vegetation, the spectral reflectance and trans* USDA/ARS, Remote Sensing and Modeling Laboratory, Beltsville, Maryland mittance of leaves are primary factors in understanding Address correspondence to C. S. T. Daughtry, Remote Sensing the reflectance of the full plant canopy. and Modeling Lab., USDA/AR, Bldg. 007, BARC-West, 10300 BaltiFor more than 3 decades, scientists have examined more Ave., Beltsville, MD 20705-2350. Received 17 October 1996; revised 9 December 1997. the biological and physical factors that affect leaf reflec-