Remote sensing techniques are now widely used in agriculture, for cultivar screening as well as for decision making tools. Empirical methods relate directly the remote sensing measured values to crop characteristics. These methods are limited by the important amount of ground data necessary for their calibration. Their validity domain is generally not very well defined as well as the associated uncertainties. Conversely, radiative transfer models allow simulating a wide range of conditions, and thus developing retrieval methods for canopy structure characteristics and biochemical composition of organs with wide and better known domain of validity. The SAIL model, one of the most widely used, relyies on strong assumptions on leaf size and spatial distribution that is acceptable for some crops like well developed cereals, but unrealistic for sunflower or vineyards. To better account for the actual architecture of canopies, 3D models offer a convenient way to evaluate and develop retrieval methods from remote sensing observations. They allow realistic simulations of the signal recorded by the sensor based on physically sound techniques such as projection and ray tracing.

This presentation demonstrates how 3D models are used to evaluate current retrieval techniques and optimize the measurement configuration including spatial and directional sampling. Estimation of the LAI are investigated based on canopy transmittance and reflectance measurements. This are applied to two contrasted crops: wheat and vineyard. Sensitivity analysis of the gap fraction measured with a commercial digital camera oriented in a specific direction provides accurate and precise LAI values under a well defined validity domain. Another example shows the effects of row orientation, sun geometry on vineyard reflectance and on the corresponding NDVI. It allows developing an interpretation model that relates NDVI, row orientation and sun position to the LAI value.

Conclusions are finally drawn on other possible applications of these techniques.