On the cause of climate model wet biases in the lowermost stratosphere

Water vapor in the lowermost stratosphere (LMS) plays a critical role in the climate system, as even small perturbations can significantly affect stratospheric temperatures and the position of the subtropical and eddy-driven jets. Climate models such as the ECHAM MESSy Atmospheric Chemistry (EMAC) model simulate strong wet biases in the LMS, reaching up to 400% compared with satellite observations. The strongest biases are found in the summer hemisphere. We find that 19% of air parcels in the LMS in the EMAC simulation exceed 30 ppmv in water vapor, a feature absent in both observations and independent Lagrangian model simulations. To diagnose the origin of this bias, we perform backward trajectory simulations with the Chemical Lagrangian Model of the Stratosphere (CLaMS) to trace the pathways of LMS air parcels and sample their Lagrangian cold points (LCPs). EMAC-simulated large-scale dehydration near the tropical cold trap is consistent with the sampled LCPs and shows no indication of a moist bias. Hence, the excessive moistening must occur downstream during transport into the LMS rather than at entry into the stratosphere. We further analyze the processes contributing to the LMS model moist bias by interpolating the physical and chemical tendencies from the EMAC model along the trajectories, including convection, vertical diffusion, and methane oxidation, as well as ice water content. For the subset of anomalously moist air parcels (water vapor mixing ratios greater than 30 ppmv), these processes collectively explain at most 30% of the simulated water vapor mixing ratios. Among the model processes, ice sublimation provides the dominant contribution, followed by vertical diffusion and convection, while methane oxidation is negligible. The large unexplained residual strongly suggests that numerical diffusion during transport is the primary driver of the excessive climate model wet bias in the LMS.