We are interested in real-time estimation of 3D fluid flows, using limited computational resources. This is possible through the coupling between synthetic data, physical models and sparse fluid flow measurements. Combining synthetic data and models, intrusive Reduced Order Model (ROM) drastically reduce the problem dimensionality. Unfortunately, even with corrections, the accumulated errors of these surrogate models increase rapidly over time. Stochastic closures (SC) and Data assimilation (DA) can alleviate these problems. SC provide a set of simulations covering probable futures (Uncertainty Quantification problem) and DA constraints these online simulations with measurements. Our SC is a multi-scale physics-based Markovian model with correlated additive and multiplicative noises, calibrated with new statistical estimators [Resseguier et al. (2021). SIAM-ASA J. Uncertain. Quantif., 9(3), 1152-1183]. Then -- from a particle filter assimilating measurements at a single location in space -- our method greatly exceeds the state of the art, for ROM degrees of freedom smaller than 10 and moderately turbulent 3D flows [Resseguier et al. (2022). J. Comput. Phys., 471, 111631]. Recently, we have obtained similar results with our new implementation in the finite-volume code OpenFOAM and the ROM library ITHACA-FV [Stabile & Rozza (2018). Comput. Fluids., 173:273–284]. We are now addressing turbulent flows, with Large Eddy Simulation (LES), hyperreduction and new dedicated SC.