Global Earth system model simulations of ocean carbon export flux are commonly interpreted only at a fixed depth horizon of 100-m, despite the fact that the maximum annual mixed layer depth (MLD_max) is a more appropriate depth horizon to evaluate export-driven carbon sequestration. We compare particulate organic carbon (POC) flux and export efficiency (e-ratio) evaluated at both the MLD_max and 100-m depth horizons, simulated for the 21^st century (2005-2100) under the RCP8.5 climate change scenario with the Biogeochemical Elemental Cycle model embedded in the Community Earth System Model (CESM1-BEC). These two depth horizon choices produce differing baseline global rates and spatial patterns of POC flux and e-ratio, with the greatest discrepancies found in regions with deep winter mixing. Over the 21^st century, enhanced stratification reduces the depth of MLD_max, with the most pronounced reductions in regions that currently experience the deepest winter mixing. Simulated global mean decreases in POC flux and in e-ratio over the 21^st century are similar for both depth horizons (8-9% for POC flux and 4-6% for e-ratio), yet the spatial patterns of change are quite different. The model simulates less pronounced decreases and even increases in POC flux and e-ratio in deep winter mixing regions when evaluated at MLD_max, since enhanced stratification over the 21^st century shoals the depth of this horizon. The differing spatial patterns of change across these two depth horizons demonstrate the importance of including multiple export depth horizons in observational and modeling efforts to monitor and predict potential future changes to export.