I assume your question is how to downscale the GCM climate scenarios to a regional or local-scale atmospheric variables. There are two main approaches; dynamical downscaling method and statistical downscaling. Stochastic weather generator may also be used produces synthetic time series of weather data. Dynamical downscaling is similar to the use of an RCM driven by a GCM to simulate regional climate. There are various projects which are established to provide high-resolution climate change scenarios for specific regions for example CORDEX (Coordinated Regional Climate Downscaling Experiment) for Africa region or more... RCMs are developed by research institutions that have sufficient computational capacity and technical expertise. Various RCMs differ in their numerical, physical, and technical aspects (http://www.ciesin.org/documents/Downscaling_CLEARED_000.pdf).
All climate datasets have to be validated using station/observed data before applying for climate impact assessments if you can possibly obtain series of station data.
If you are going to us RCM outputs there are a bunch 'bias correction' methods such as Delta change method, linear scaling, power transformation, quantile mapping or distribution mapping methods...
GCMs don't provide an accurate description of the local climate. To overcome this discrepancy, downscaling is applied to produce local-scale climate predictions based on corresponding GCM scenarios.
Downscaling can be done either dynamically or statistically.
Dynamic downscaling is computationally intensive as it makes use of the lateral boundary conditions alongwith with regional-scale forcings to produce Regional Climate Models (RCMs) from a GCM.
Statistical Downscaling is a 2-step process consisting of: 1) the development of statistical relationships between the local climate variables and large-scale predictors, and 2) the application of such relationships to large-scale output.