Adequate design of control treatments in long term agroforestry experiments with multiple objectives

Agroforestry experiments usually include control plots of either pure crop or pure tree stands. A clear distinction should be made between intensively managed biophysical controls and farming system controls with realistic labour input and management regimes. Trying to draw biophysical conclusions from farming system controls (or the reverse) is often not justifiable. The design and management of these elusive control plots is a complicated issue which is often overlooked. Many factors beyond the control of the experiment manager can disturb long term field agroforestry experiments. Some examples from French agroforestry experiments illustrate how uncontrolled factors may bias the results, including the proportion of harvested to planted trees, the weeding regimes, and the use of tree-shelters. The analysis of agroforestry data could be more efficient when considering a continuum of tree - crop mixture management options between the agroforestry plot and the non agroforestry plot. The concept of biophysical control plots becomes then less essential. A relevant modelling approach of interactions between trees and crops should 1) perform correctly for any tree/crop proportion and even for pure stands, when setting the parameters of the other component to zero, 2) provide for the inclusion of new, uncontrolled factors that could emerge through time. The biological efficiency of agroforestry systems may however be a subordinate criterion for agroforestry adoption, as observed at the moment in France. Agroforestry systems with poor biological outcomes can even be very attractive in some ecological or sociological conditions, and only farming system controls may bring this aspect to light.