Leaf area index (LAI) is a key variable for the understanding of several eco-physiological processes within a vegetation canopy. The LAI could thus provide vital information for the management of the environment and agricultural practices when estimated continuously over time and space thanks to remote sensing sensors. This study proposed a method to estimate LAI spatial and temporal variation based on multi-temporal remote sensing observations processed using a simple semi-mechanistic canopy structure dynamic model (CSDM) coupled with a radiative transfer model (RTM). The CSDM described the temporal evolution of the LAI as function of the accumulated daily air temperature as measured from classical ground meteorological stations. The retrieval performances were evaluated for two different data sets: first, a data set simulated by the RTM but taking into account realistic measurement conditions and uncertainties resulting from different error sources; second, an experimental data set acquired over maize crops the Blue Earth City area (USA) in 1998. Results showed that the proposed approach improved significantly the retrieval performances for LAI mainly by smoothing the residual errors associated to each individual observation. In addition it provides a way to describe in a continuous manner the LAI time course from a limited number of observations during the growth cycle.