The present study proposes a new method for estimating nonlocal mixing parameters from microstructure data processing. The method could be particularly useful in studies of biological-physical interactions at small scale because it overcomes some of the limitations of the eddy diffusivity concept when dealing with the complex vertical pattern of biological and chemical tracers. The proposed method obtains empirically the coefficients of the transilient matrix, this being the discrete descriptor used in nonlocal mixing closure. The estimation is based on microstructure data analysis, in particular Thorpe displacement profiles, and includes three main steps: turbulent patch identification within each profile, turbulent patch characterization and multi-profile data averaging. A field case study is included to show how the estimated transilient coefficients, and the mixing parameters derived from them, are in accordance with what can be expected from the background external forcing and the observed thermal structure.