The application of the reverse time migration (RTM) in land seismic data is still a great challenge due to its low quality, low signal-to-noise ratio, irregular spatial sampling, acquisition gaps, missing traces, etc. Therefore, prior to the application of this kind of depth migration, the input pre-stack data must be conveniently preconditioned, that is, it must be interpolated, regularized, and enhanced. There are several methods for seismic data preconditioning, but for 2D real land data, the regularization of pre-stack data based on common reflection surface (CRS) stack method provides high quality enhanced preconditioned data, which is suitable for pre-stack depth migration and velocity model building. This work demonstrates the potential of RTM combined with CRS-based pre-stack data regularization, applied to real land seismic data with low quality and irregularly sparse spatial sampled, from geologically complex areas with the presence of diabase sills and steep dip reflections. Usually, determining the wavelet of the seismic source from land data is a challenge, because of this, RTM migration is often applied using artificial sources (e.g. Ricker). In this work, from the power spectrum of the pre-stacked data, we determine the wavelet of the seismic source to apply the RTM to real land data. We present applications of the pre-stack data preconditioning based on CRS stack and of the RTM in 2D land data of Tacutu and Parnaiba Basins, Brazil. Comparisons with the standard Kirchhoff depth migration reveals that the RTM improves the quality and resolution of the migrated images.