Impurity transport is an issue of utmost importance for tokamaks. One reason is the choice of tungsten for ITER divertor. Indeed high-Z materials are only partially ionized in the plasma core, so that they can lead to prohibitive radiative losses even at low concentrations, and impact dramatically plasma performance and stability. On-axis accumulation of tungsten has been widely observed in tokamaks. While the very core impurity peaking is generally attributed to neoclassical effects, turbulent transport could well dominate in the gradient region at ITER relevant collisionality. The transport of low and medium-Z impurities also results from both neoclassical and turbulent transport. Up to recently, first principles simulations of corresponding fluxes were performed with different dedicated codes, implicitly assuming that both transport channels are separable and therefore additive. The validity of this assumption can be questionned. Preliminary simulations obtained with the gyrokinetic codeGYSELA have shown clear evidences of a neoclassical-turbulence synergy for impurity transport. However no clear theoretical explanation was given. New simulations have been done using a new and more accurate collision operator, improved boundary conditions and more flexible sources. The new simulations confirm the neoclassical turbulence synergy and allow identification of a mechanism that underly this synergy. Turbulence can induce poloidal asymmetries. An analytical work performed during this thesis allows to compute the level and the structure of the axisymmetric part of the electric potential knowing the turbulence intensity. Two mechanisms are found for the generation of poloidal asymmetries of the electric potential: at large frequencies, flow compressibility is a key player for the generation of poloidal asymmetries. In this case a sine structure is predicted as in the case of GAMs. For lower frequencies, the ballooning of the turbulence is instrumental in the generation of the poloidal asymmetries. In this case, a cosine structure is predicted. A new prediction for the neoclassical impurity flux in presence of large poloidal asymmetries and pressure anisotropies has been derived during this thesis. It turns out that both banana/plateau and Pfirsch-Schlüter contributions are significantly impacted by the presence of large poloidal asymmetries and pressure anisotropies. A fair agreement has been found between the new theoretical prediction for neoclassical impurity flux and the results of a GYSELA simulation displaying large poloidal asymmetries and pressure anisotropies induced by the presence of turbulence.