Evaluation of the Congo Air Boundary in reanalyses during the 2022 DRY-CAB field campaign

Abstract The Congo Air Boundary (CAB) is a synoptic-scale dryline and convergence zone that forms across southern Africa from August-November and plays a central role in both rainfall onset and projections of regional drying. Until recently our understanding of the CAB has relied almost entirely on reanalysis and model data, but the fidelity with which the CAB is reproduced in these datasets is not known. Using a new observational benchmark derived from the DRY-CAB field campaign, this study evaluates the ability of three reanalyses – ERA5, MERRA-2, and JRA-3Q – to reproduce the CAB. All three reanalyses exhibit substantial errors in the CAB’s mesoscale structure and circulation: no product captures dryline strength (biases >±25%), and only ERA5 successfully reproduces its diurnal evolution. A process-based evaluation reveals that these errors stem from mis-represented moisture transport to the north of the CAB, with moisture fluxes of the wrong sign in MERRA-2 (−120% bias), exhibiting an incorrect diurnal cycle in JRA-3Q (maximum at 20:00 rather than 02:00), and arising from compensating biases in ERA5. ERA5 and MERRA-2 additionally exhibit excessive 11:00 LT instability and daytime near-surface drying (biases 71-221%). We identify cross-dryline soil moisture as a key driver of differences in representations of the CAB and show that these errors shape reanalysis rainfall onset meteorology, with ERA5 and MERRA-2 failing to capture the sharp observed convective regime shifts and JRA-3Q reproducing observations through a spurious circulation (“right for the wrong reasons”). Our findings underscore the need for caution when using reanalyses across the data-sparse subtropics.