A central question in theoretical population biology is whether spatial ecological heterogeneity helps or hinders the evolutionary success of mutants. Here I present a unified framework for selection on lattice graphs with arbitrary environmental heterogeneity, spanning two regimes: constant selection and frequency-dependent evolutionary games. In the first part, I calculate the fixation probability of advantageous mutants. I observe that the genotype specificity of the environment determines whether heterogeneity amplifies or suppresses selection, while spatial correlation in environmental states acts as a secondary determinant of its magnitude. In the second part, I extend the framework to the evolution of cooperation under spatially varying payoffs, where spatial correlation becomes the primary determinant, predicting both the fixation probability of cooperators and the timescales of selection. I discuss applications to the evolution of drug resistance, cooperation in microbial communities, and somatic evolution in epithelial tissues.