We have processed up to 15 years of continuous GPS data including about 30 episodic GPS sites observed in Tanzania from 2006-2011, testing various noise models, in order to derive a new, continent-wide, velocity field for Africa. We also combined our GPS solution with a global DORIS solution to derive a new position/velocity solution for Africa as a contribution to the upcoming Africa Reference Frame (AFREF). We examined the noise characteristics in both GPS and DORIS time-series and determined realistic velocity uncertainties. The resulting velocity field describes horizontal plate motions at more than 120 GPS sites and 9 DORIS sites operating in Africa. We compared the agreement of velocity fields obtained from different noise models in E-W and N-S components. We then used the derived velocity fields to update Nubia-ITRF2008 angular velocity and compared the plate angular velocities with those from previous studies using different noise models. This is because the reference frame for Africa is likely to be fixed to the Nubian plate with time-invariant coordinates at a given epoch, since the primary purpose of AFREF is geo-referencing for surveying applications. We also used the velocity field to quantify the level of rigidity of the Nubian plate and updated the angular velocity estimate for the Somalian plate and smaller sub-plates Victoria and Rovuma. We found an Euler pole for Rovuma located about 16 degrees south of previous estimates. The Victoria microplate Euler pole, however, shifted only a few degrees to the SSW. Disagreements in angular velocities are likely due to improved spatial distribution and observation time span of GPS instrumentation. The spatial density of the currently available GPS and DORIS networks is not sufficient to estimate robust angular velocities for a remaining smaller sub-plate (Lwandle) without using additional external geophysical data.