<p>In this paper, we are concerned with a new stochastic benthic-drift model (SBDM) with memory-based self-diffusion. Firstly, we prove the unique stationary distribution of the stochastic benthic-drift model without memory-based diffusion term. Secondly, with the memory-based self-diffusion coefficient regarded as parameter, we prove the stochastic bifurcation of the reduced system derived by stochastic parameterizing manifolds near the critical point, which shows the impact of noise and the averaged memory period on the stochastic bifurcation of SBDM. More precisely, pitchfork bifurcation is observed from the reduced system without memory-based diffusion term, but Hopf bifurcation is observed from the reduced system with memory-based diffusion term. Finally, we derive rigorous error estimates between the reduced systems and those of the original SBDM emanating from numerical analysis.</p>