This thesis focuses on the valuation of real options when there is flexibility given by the choice between two risky outputs. We develop models to value these rainbow options and to determine optimal operating and investment policies. These models are studied in the context of commodity applications because output flexibility is particularly relevant in volatile commodity markets. We provide insights into the behaviour and sensitivities of option values and operating policies and discuss implications for decision-making.In the early stages of real options theory, research centred on basic options with closed-form solutions, modelling single uncertainty in most cases. The challenge is now to incorporate more complexities in the models in order to further bridge the gap between theoretical models and reality, thereby promoting the widespread application of real options theory in corporate finance.The new option models developed in this thesis are organised in three self-contained research papers to address specific research problems. The first research paper studies an asset with flexibility to continuously choose the best of two risky commodity outputs by switching between them. We develop quasi-analytical and numerical lattice solutions for this real option model, taking into account operating and switching costs. An empirical application to a flexible fertilizer plant shows that the value of flexibility between the two outputs, ammonia and urea, exceeds the required additional investment cost (given the parameter values) despite the high correlation between the commodities. Implications are derived for investors and policy makers. The real asset value is mainly driven by non-stationary commodity prices in combination with constant operating costs. In the second research paper, we study an asset with flexibility to continuously choose the best of two co-integrated commodities. The uncertainty in two commodity prices is reduced to only one source of uncertainty by modelling the spread, which is mean-reverting in the case of co-integration. Our quasi-analytical solution distinguishes between different risk and discount factors which are shown to be particularly relevant in the context of mean-reversion. In an empirical application, a polyethylene plant is valued and it is found that the value of flexibility is reduced by strong mean-reversion in the spread between the commodities. Hence, operating flexibility is higher when the commodities are less co-integrated. In the third research paper, we develop real option models to value European sequential rainbow options, first on the best of two correlated stochastic assets and then on the spread between two stochastic co-integrated assets. We present finite difference and Monte Carlo simulation results for both, and additionally a closed-form solution for the latter. Interestingly, the sequential option value is negatively correlated with the volatility of one of the two assets in the special case when the volatility of that asset is low and the option is in-the-money. Also, the sequential option on the mean-reverting spread does not necessarily increase in value with a longer time to maturity.