Crop/weed discrimination using remote sensing

The ability to selectively apply herbicides through the use of weed mapping and variable rate sprayers offers the potential to reduce harmful effects on the environment as well as to optimize producer profitability. Over the past 25 years, there have been numerous studies involving remote sensing and weed/crop discrimination. With the development of greater spectral and spatial resolution sensors and spectral mixture analysis techniques re-investigation of the discrimination of weed and crop species present in cultivated systems appears timely. In a laboratory experiment, the spectral separability of five weeds and two crops of economic importance on the Canadian prairies was investigated. Reflectance from the uppermost fully expanded leaves of field grown plants was examined over the range 350-2500 nm using a field spectroradiometer and an integrating sphere. The various plant species were not separable using the broad bands present on the Landsat TM satellite. However, the weed/crop species could be discriminated using reflectance in approximately 30 10-nm wide bands across the electromagnetic spectrum. Results suggest that hyper-spectral remote sensing could be used for weed/crop mapping and merits further study under field conditions.