In this article, the variability of the relationship between the normalized difference vegetation index (NDVI) and the daily interception efficiency (epsilon(i)) of a discontinuous canopy is investigated using a simple radiative transfer model developed for regular-clumped crops. Sensitivity to the canopy and irradiance parameters is first studied for 1) the relationship between epsilon(i) and LAI and 2) the relationship between NDVI and LAI. The analysis is conducted varying one parameter at a time to provide an understanding of the influence of that variable on the behavior of the relationships. Next, with the same procedure, the relationship between epsilon(i) and NDVI is investigated. This investigation focuses on the sensitivity of the relationship to the structural parameters and its linearity. It shows that for most of the semiarid vegetation types, characterized by low LAI and bright background, the epsilon(i)-NDVI relationship could be simplified to a linear relation with an abscissa offset equal to the normalized difference vegetation index of the background (NDVI(s)): epsilon(i) = a(NDVI - NDVI(s)). Further studies show that 1) the more the canopy is discontinuous, the more the slope a is sensitive to the solar zenith angle, 2) the slope a tends to a unique value 1.3 as the canopy tends to a homogeneous medium, and 3) the relative standard error of epsilon(i) is time-dependent (minimum at theta(s) = 30-degrees) and all the higher as the structure of the canopy is unknown. The relevance of these investigations to satellite data is then discussed, resulting in long-term recommendations as satellite orbital characteristics providing the optimal geometric irradiance for epsilon(i) estimation, or additional new sensors providing complementary information about the canopy. At short term, further epsilon(i)-NDVI sensitivity analyses are encouraged, especially for other types of discontinuous canopies and off-nadir viewing.