We introduce Spatial Pattern Regression (SPR), a method to generate gridded historical meteorological data for climate adaptation. SPR operates in two steps: first extracting spatial structure from high-resolution regional climate model (RCM) simulations as eigenvectors, then using them in linear regression to reconstruct complete gridded fields from station observations at each time step. We compare SPR with standard interpolation methods using data from RCM simulations, where virtual stations are a subset of grid cells and interpolation is done on the rest. Thirty graded case studies are created by varying three factors: region location, size, and network density. Daily precipitation, maximum temperature, and minimum temperature are considered. Results show SPR outperforms standard methods across all three variables for most of the graded case studies. A stress-test with very low network density confirms SPR's robustness. Finally, we systematically assessed how each graded factor affects SPR’s performance.