We develop a distributed algorithm to compute feature-aligned poly-square maps for large-scale 2D geometric regions and use it to construct low-distortion semi-structured quad meshes. Our proposed algorithm has two main compo- nents. The first is a feature-aware graph partitioning that considers workload balancing, minimal communication, geometric regularity, and feature-preserving. The second is a feature-preserved poly-square parameterization. We demonstrate that our algorithm is effective on meshing huge complex coastal/terrain data and can consequently benefit scientific simulations that run on such meshes using high-performance computer clusters.