Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #264661


Location: Hydrology and Remote Sensing Laboratory

Title: Can hyperspectral remote sensing detect species specific biochemicals?

item Vanderbilt, Vern - National Aeronautics Space Administration (NASA) - Jet Propulsion Laboratory
item Daughtry, Craig

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 3/23/2011
Publication Date: 8/1/2011
Citation: Vanderbilt, V.C., Daughtry, C.S. 2011. Can hyperspectral remote sensing detect species specific biochemicals? [abstract]. Geoscience & Remote Sensing Symposium. 2011 CDROM.

Interpretive Summary:

Technical Abstract: Discrimination of a few plants scattered among many plants is a goal common to detection of agricultural weeds and invasive species. Detection of clandestinely grown Cannabis sativa L. is in many ways a special case of weed detection. Remote sensing technology provides an automated, computer based, land cover classification capability that holds promise for improving upon the existing approaches to Cannabis detection. Yet little research on the application of remote sensing technology to the Cannabis detection problem has been reported, possibly because of the difficulty of detecting what essentially is a weed grown in ways to foil detection. In this research, we investigated whether hyperspectral reflectance of recently harvested, fully turgid Cannabis leaves and floral clusters depends upon the concentration of cannabinoids, including Tetrahydrocannabinol (THC) that, if present at sufficient concentration, presumably would allow species-specific identification of Cannabis. Pistillate (female) plants of several drug- and fiber-types of Cannabis were grown in a secure field. In mid-September and early October as the plants matured, apical stem segments with leaves and floral clusters were harvested, arranged in optically thick layers in black painted sample trays. Reflectance spectra were acquired with a spectroradiometers (FieldSpec Pro, ASD, Inc.) over the 350 to 2500 nm wavelength region. The samples were illuminated by six 100-W quartz-halogen lamps that were current regulated. The 8º fore-optic of the ASD and a digital camera were aligned and positioned 60 cm from the sample surface at a 0º view angle. Sample weight and spectral reflectance were acquired intermittently over 2-4 weeks as the samples in the trays air-dried. All spectral data were converted to reflectance factors with reference to measurements of a Spectralon standard (Labsphere, North Sutton, New Hampshire, USA). Cannabinoid profiles of the three randomly selected sub-samples from each tray were determined. Spectral correlation coefficients were calculated for specific pairs of spectral reflectance. The reflectance spectra from the drug- and fiber-type Cannabis plants appeared closely similar, displaying no gross differences or unique absorption features suitable for use as a key for Cannabis detection. Analysis of the reflectance factor data of the ensembles of leaves and floral clusters at 0 and 1 day after harvest showed no evidence of information related to THC concentration within any of the narrow spectral bands. There is a small possibility of THC related information in the reflectance spectra of air-dried samples (>13 days after harvest). We propose that species-specific plant biochemicals tend not to be present in healthy leaves at concentrations detectable in remotely sensed imagery even with application of hyperspectral remote sensing technology. Conversely, concentrations of various remotely sensible biochemicals do not appear in general to be uniquely characteristic of specific plant species. Accurate, scientifically repeatable plant species identification based solely upon application of hyperspectral remote sensing analysis techniques applied to detect species-specific, spectral absorption features remains to be demonstrated. Distinctive foliage color differences generally do allow application of remote sensing technology to discriminate, for example, green weeds growing among ripe yellow wheat plants or leafy spurge (Euphorbia esula), an invasive species displaying yellow flowers in contrast with the all green foliage of nearby native species. But when the reflectance spectra of the foliage of both the target species and the background species are similar, discrimination often proves difficult even with application of hyperspectral remote sensing technology