This dissertation covers the theme of stochastic uncertainties modeling with mean reversion, and its applications in real options valuation. The use of alternate stochastic processes other than random walk, or Geometric Brownian Motion (GBM), usually does not have analytical closed solutions for valuing contingent claims and therefore numerical approaches must be used. The dissertation is divided in three main chapters that cover this theme. In the first of these, after a discussion on the validity of stochastic process, the most widely known Mean Reversion Models (MRM) are presented: four single factor processes, one arithmetic and three geometric, and a two factor process. For all of these we show or develop the discrete process, the expression for the expected value and estimation of parameters from historical data. This last point is fundamental since generally only historical data is available for uncertainties involved in most real options applications. The chapter also estimates parameters of the main models presented for data from prices of ethanol and sugar paid to producers in the State of São Paulo. The second of these chapters deals with lattice modeling as a discrete method for approaching mean reversion processes. This approach allows the valuation of contingent claims written on a variable whose value follows this stochastic behavior. Lattice modeling is already a classic approach used in countless papers and originates from the methodology developed by Cox, Ross and Rubinstein (1979). But this latter method only fits variables showing an MGB dynamic, excluding the whole range of assets for which an autoregressive process is a better description of their price behavior. Two mean reversion lattice models are then explained. Also shown is an approach allowing the composition of a bivariate lattice with two independent yet correlated stochastic processes, of which one is a MRM and the other either an MGB or another MRM. These approaches are then used to value an expansion option available to a sugar producing plant to also produce ethanol. The third chapter uses the methodologies developed in the first two to value the switch option embedded in Brazilian sugarcane flexible plants. These can switch production from one output to another (ethanol and sugar) and this flexibility has significant value. The chapter values this real option using the bi-variate lattice approach, combining two correlated MRMs for both prices of the possible output products. This modeling is then compared to the results of a simulation method, which confirms the convergence of both results. We also analyze the sensibility of the option switch value to the correlation of both stochastic processes. Finally the dissertation concludes on the importance of the correct choice of the stochastic process when modeling the uncertainties involved in real options valuation, and suggests further research in the same line, such as parameterization of jump processes.