GEOS-Chem High Performance (GCHP) is a widely used MPI-based distributed atmospheric chemistry model for Earth-scale simulations. It partitions the atmosphere into vertical columns that can be simulated individually, with blocks of adjacent columns assigned to MPI ranks. It suffers from significant performance loss due to load imbalance, however, since the per-column workload varies strongly across space and time. To address this issue, we develop load distribution solutions for GCHP that account for implementation constraints, including those imposed by its MAPL/ESMF framework, and differences in communication costs across platforms. We propose multiple GCHP-compatible options, evaluate them on a private HPC cluster and a public cloud, identify the dominant cost trade-offs governing performance, and select a strategy that is effective across configurations. Compared to the GCHP baseline, our best approach improves overall runtime by more than 20 % in most configurations and by over 40 % in high-resolution configurations.