This paper presents results of validating the Collection 4 Moderate Resolution Imaging Spectroradiometer (MODIS) leaf area index (LAI) product using LAI data collected in a 3 ¿ 3 km agricultural (grasses and cereal crops) area near Avignon, France, and 30 m resolution Enhanced Thematic Mapper (ETM+) image. Estimates of the accuracy, precision, and uncertainty with which the ETM+ data convey information about LAI underlie the derivation of a 30 m resolution reference LAI map by accounting for both field measurement and satellite observation errors. The 30 m reference LAI was then extrapolated from sampling points to a 58 km2 area without loss in the quality and was degraded to a 1 km resolution LAI map. The latter was taken as a reference to assess the quality of the MODIS LAI product. Comparison of the reference and corresponding MODIS retrievals suggests that Collection 4 MODIS LAI is accurate to within an accuracy of 0.3 with a precision and uncertainty of 0.23 and 0.38, respectively. It was found that the Collection 3 MODIS land cover product, input to the Collection 4 operational LAI algorithm, misclassified the 58 km2 area as broadleaf crops. The use of correct biome type in the operational processing improves the accuracy in LAI by a factor of 2 with an almost unchanged precision and uncertainty. Our results also indicate that the retrieval of LAI from satellite data is an ill-posed problem; that is, small variations in input due to observation errors result in a very low precision of the desired parameter. Any retrieval technique based on a simple model inversion or empirical relationships is unable to generate stable retrievals. The use of information on input errors in the retrieval technique is necessary to generate solutions to the ill-posed problem. The MODIS operational LAI algorithm meets this requirement.