The nature of turbulence in molecular clouds is one of the key parameters that control star formation efficiency: compressive motions, as opposed to solenoidal motions, can trigger the collapse of cores, or mark the expansion of Hii regions. We try to observationally derive the fractions of momentum density ($\rho v$) contained in the solenoidal and compressive modes of turbulence in the Orion B molecular cloud and relate these fractions to the star formation efficiency in the cloud. The implementation of a statistical method developed by Brunt & Federrath (2014), applied to a $^{13}$CO(J=1-0) datacube obtained with the IRAM-30m telescope, allows us to retrieve 3-dimensional quantities from the projected quantities provided by the observations, yielding an estimate of the compressive versus solenoidal ratio in various regions of the cloud. Despite the Orion B molecular cloud being highly supersonic (mean Mach number $\sim$ 6), the fractions of motion in each mode diverge significantly from equipartition. The cloud's motions are on average mostly solenoidal (excess > 8 % with respect to equipartition), which is consistent with its low star formation rate. On the other hand, the motions around the main star-forming regions (NGC 2023 and NGC 2024) prove to be strongly compressive. We have successfully applied to observational data a method that was so far only tested on simulations, and have shown that there can be a strong intra-cloud variability of the compressive and solenoidal fractions, these fractions being in turn related to the star formation efficiency. This opens a new possibility for star-formation diagnostics in galactic molecular clouds.< Réduire