The canopy radiation model EXTRAD was used to quantify the accuracy of leaf area index (LAI) estimations from vegetation indices (VIs), derived from green and infra-red crop reflectance. The VIs were the infra-red/green (IR/GR) ratio, the normalized difference vegetation index (NDVI), the perpendicular vegetation index (PVI) and the weighted difference vegetation index (WDVI). The accuracy of LAI estimation was calculated in relation to variation in leaf green and infra-red colour, leaf angle distribution, soil background and illumination conditions. The theoretical calculations were supported with a field experiment on sugar-beet. Variation in illumination conditions and soil background gave relatively small estimation errors with all four VIs. The largest estimation errors resulted from variation in leaf colour and leaf angle distribution. With variation in green leaf colour, the estimation errors were lowest with the WDVI. With variation in leaf angle distribution, the errors were lowest with the IR/GR ratio and the NDVI. In practice, the magnitude of the error in LAI estimation will depend mostly on the magnitude and combination of occurring variation in leaf colour and leaf angle distribution. In an average of 100 random combinations of disturbing conditions, and in a field experiment with sugar-beet, the absolute estimation errors ranged between 0.1 (and less) for 0 < LAI < 1 and 0.35 for 3 < LAI < 5.
