It is extremely challenging for CFD modeling to accurately predict the combustion system and its evolution, due to uncertainties of the numerical algorithms, the physical models, and the model parameters used in the equations governing combustion. In addition, initial and boundary conditions are also a source of significant uncertainties that can accumulate over time. This talk will discuss an emerging method by applying data assimilation to the combustion science and engineering for model parameter estimation, and thus the predictions can be improved. Data assimilation, originated from meteorology, oceanography, and climatology, has had dramatic impacts on the improvement of weather forecasts. For CFD modeling of combustion, we believe, by melding observation with model prediction, our understanding of combustion physics and relevant model parameters will be improved. However, numerical challenges arise when applying data assimilation techniques to engineering fluids with combustion, and issues on conservation and preservation of positivity become critical. This talk will illustrate the application by using the Ensemble Kalman Filter (EnKF) algorithm to perform parameter estimation for a flame propagation model. Through this investigation, we have established the proper workflow, identified numerical challenges, and developed techniques to overcome the numerical difficulties. The modified EnKF algorithm will be presented for maintaining conservation and preserving positivity.