RCPs are meant to represent uncertainty in future. I recommend starting with ensemble mean for your region to calculate mean climatic conditions. You can use the other RCP scenarios to capture the uncertainty in the yields resulting from variation away from the mean RCP.
Thanks for your response, Timothy, but I meant that getting the future climate data I should choose one appropriate Representative Concentration Pathways for my region based on that one RCPs climate data I will analyze the impact on crop production future. Suppose, I choose RCP 8.5 that means 8.5 w/m2 of Radiative forcing will be impacting the greenhouse gas concentration and corresponding that crop production will be affected, but if that RCP would not be the case during 2050, suppose, then analysis will not make any sense, right? That's why I want to know the reasons to choose one RCPs that could represent the climatic condition for my desired analysis period.
A scenario is a description of what could possibly happen. The assumption for the different RCP scenarios is that they are equally likely for the future, depending on different possible developments (technology development, population growth, etc.). Therefore, it would be good if you can consider a range of possible RCP scenarios for the estimation of future crop productivity.
RCM simulations are available for different regions (see e.g. CORDEX). Depending on your needs and applied crop yield models (regional or plot scale models) you should consider to do a bias correction of the data, which is unfortunately also not straightforward.
What region and which crops are you interested in?