In a context of climate change, adaptation of perennial plantations to water constraints becomes a major concern for wood production. Projections point towards more extreme precipitation regimes coupled with more intense, more frequent and longer lasting heat waves inducing frequent and severe droughts. During two years, phenotypic plasticity of Pinus pinaster Ait. to water availability was quantified in the field by planting 1-year-old seedlings under a greenhouse, open at its borders, for rainwater exclusion. Water was provided to re-supply evapo-transpiration losses to half the plants by aerial irrigation. One well growing and one slow growing half-sib families, both of the local improved provenance were compared. The soil water content, water table level, air temperature and humidity were monitored. Shallow soil water content decreased to 6% in the dry treatment in late summer. Predawn leaf water potential was regularly measured and reached -1.37 MPa in the dry treatment/slow-growth family (versus -0.2 in the irrigated treatment) at the end of experiment. After measuring height and diameter (3 times per year in March, July and September 2008 and 2009), 40 to 70 saplings were uprooted for aerial and root biomass and architecture assessment. At the two last sampling times, architecture was measured by 3D digitizing which allowed precise numeric representation of the geometry and topology. Each plant was digitized in two multiscale tree graphs (MTG) of the aerial and root architecture. Several characteristics of architecture were extracted by queries including root/branch number, length, diameter, volume, spatial position, ramification order, branching angle and length of interlaterals. Surprisingly, after the first year of the experiment, water shortage did not affect total biomass of the saplings. However, the water stressed trees had slimmer stems, more biomass allocation to needles (+18%), and distinctly less allocation to roots (-30%), especially distal roots. After two years, the improved saplings in the dry treatment were less tapered. An increase of diameter growth was observed only at the end of the second year of the experiment. We hypothesize that Pinus pinaster saplings stop root growth when the soil is too dry, but maintain their productivity by setting more needles.< Réduire