A lot of data is available in the literature on root length density profiles (i.e., root length per unit of soil volume versus soil depth), because they are a traditional way of representing root distribution in the field. In a complementary approach comprehensive models of the root system architecture are being developed, but the parameters required for these models are often difficult to assess in field conditions. In this paper, we bridge both approaches, empirical and comprehensive, by evaluating the capability of architectural models to simulate the observed diversity in root length distribution on the one hand and the possibility of estimating developmental parameters from root length density profiles on the other. For this purpose, we constructed a simple model with only six parameters, to represent the root system architecture. It reproduced a large diversity of root profiles comparable to observations reported in the literature and encompassing many different crops. The impact of each model parameter, as well as their interactions, on the shape of the profiles was quantified using a global sensitivity analysis. Finally, a statistical meta‐model was designed and estimated to simulate the same collection of profiles without intermediate simulating the whole architecture. The meta‐model allows for estimation of architectural parameters from profile shapes (inversion). Some architectural parameters could be estimated from the profiles with good accuracy, especially those quantifying the growth potential and gravitropism of individual roots, because of their specific impact on root length at a specific depth. But others (like inter‐branch distance, life duration), which modify root density in a more diffuse way throughout the profile, could not be identified correctly using this method. Additional data involving specific measurements are necessary to identify these last parameters.