A semi-empirical, kernel-driven Ambrals BRDF (bidirectional reflectance distribution function) model was developed for correcting and studying view and illumination angle effects of a wide variety of land covers in remote sensing applications. This model was validated by demonstrating its ability to model 27 different multiangular data sets well, representing major types of land cover, including crops (maize, soyabeans, wheat, sorghum, vineyards, vegetables and sunflowers), grasslands (Konza prairie), Jack pine [Pinus banksiana], black spruce [Picea mariana] and aspen [Populus spp.] forests, and bare and ploughed soil. The selection of the kernels used in the model was related to land cover type, and the inversion accuracy was good in nearly all cases: the correlation coefficient between modelled and observed reflectances was larger than 0.9 for about half of the data sets and larger than 0.70 in all but two cases where the observations were irregular. The average root mean squared error of the inversions was 0.034. A new kernel modelling the sun zenith angle dependence of multiple scattering improved fits for dense vegetation. Operation of the Ambrals model was demonstrated by applying it to an ASAS (Advanced Solid-State Array Spectroradiometer) image on a per-pixel basis