We report detailed depth profiles of the particle-reactive radionuclides 234Th (24.1 days) and 7Be (53.3 days) in sediment cores collected at different seasons in the Thau Lagoon, located in the south of France. Two sites with contrasting characteristics were selected: C4 in the middle of the lagoon and C5 nearby oyster farming. Both sites were visited six times between December 2001 and May 2003. 234Th in excess (234Th$_{\mathrm{xs}}$; i.e. supplied to sediment by settling particles) and 7Be both show seasonal variations in activities and in penetration within sediment. With their very short half-lives and moderate sedimentation rates at both sites (around 0.1 - 0.3 cm per year), 234Th$_{\mathrm{xs}}$ and 7Be should be present only at the water-sediment interface. However all the profiles show penetration of both short-lived radionuclides to variable depths, from 1 up to 8 cm, that indicates efficient biological mixing of upper sediments. Such a mixing is usually described as a biodiffusive transport defined by bioturbation coefficient (D$_{\mathrm{b}}$). The simplest way to calculate D$_{\mathrm{b}}$ from radionuclide profiles is to assume steady-state within the mixed layer. For site C4, steady-state bioturbation rates (D$_{\mathrm{b}}$-SS) range between 1 and 12 cm2 yr-1, with a weak seasonal signal. The mixing of surface sediments at site C5 presents a greater range (1 - 31 cm2 yr-1) with the highest values observed in summer. Radionuclide fluxes at the water-sediment interface show too seasonal variations, particularly marked at site C5 in relation with its position. Such a time-series of short radionuclides at the water-sediment is most unusual. It provides the opportunity to develop a non-steady state modelling in order to test the significance of bioturbation rates based on steady state assumption. These tests allow to figure out how variations (seasonal, episodic) of radionuclide fluxes could induce errors on D$_{\mathrm{b}}$-SS determinations.