The land equivalent ratio (LER) of an agroforestry system (AFS) is a measure of the advantage of mixing species for production. LERs of annual crop mixtures are easy to measure, as each growing season provides an estimate. LERs of perennial AFS are more difficult to measure or predict. The long-term growth of the tree component must be measured, and the productivity of the intercrops must be monitored for a long time, until crop production is no longer profitable. Several pitfalls of LER calculation must also be avoided, and require data for sole crop and sole tree plots that are very often missing in many experimental designs. This explains why almost no measured LERs of AFS have been published to date. We monitored a poplar-cereal AFS from tree plantation in 1996 until tree harvest in 2008 in southern France. This allowed us to measure the actual value of the final LER of the system. This is probably the first field measurement of an agroforestry LER in Europe and in the temperate zone. Four systems were monitored, differing by tree row orientation and poplar clones. The average LER is very high compared to previously published estimates of LERs in temperate AFS. Using simulation models of the system, we tried to separate the impacts of various factors in explaining such high productivities of tree-crop mixtures. Three factors appear to be essential: phenology lags between tree and crop components, plasticity of the root systems of the tree component in response to the competition by the crop component, and the availability of a deep resource of water that is within reach of the trees’ rooting system. Such productivity data are at odds with many predictions of the value of temperate AFS, and could be a stimulus for the adoption of AFS throughout Europe