Four one-dimensional radiative transfer models are compared in direct and inverse modes. These models are combinations of the PROSPECT leaf optical properties model and the SAIL (Scattering by Arbitrarily Inclined Leaves), IAPI, KUUSK, and NADI (New Advanced Discrete Model) canopy reflectance models. To evaluate their ability to estimate canopy biophysical parameters, inversions were first performed on synthetic reflectance spectra (10 wavelengths in the visible and near-infrared). The simulated spectral and directional reflectances showed good agreement among the four models. A 1997 airborne experiment in the United States was used to test their performance on real data. This experiment gathered a unique data set composed primarily of 200 reflectance spectra acquired over corn ( Zea mays L.) and soybean ( Glycine max) fields, and the corresponding ground truth (chlorophyll a+b content and leaf area index). Only the first three models, which ran fast enough to allow the processing of a large data set, were actually inverted by iterative optimization techniques. Inversions were conducted in successive stages where the number of retrieved parameters was reduced. No significant difference can be observed between the three models. Globally, the leaf mesophyll structure parameter and leaf dry matter content couldn't be estimated. The chlorophyll content, the leaf area index, and the mean leaf inclination angle yielded better results, although the latter wasn't validated due to missing ground data. Assuming that model inversion by iterative optimization techniques is a promising method to extract information on plant canopies, the SAIL and KUUSK models, which perform well in terms of accuracy and running time, proved to be good candidates for remote sensing application in ecology or agriculture (precision farming).