The adjustment of cloud liquid water path (LWP) to droplet number concentration (N<sub>d</sub>) remains a leading uncertainty in the aerosol-induced cloud radiative effect. We combine nine years (2006–2014) of collocated, vertically resolved CALIOP aerosol extinction and MODIS cloud retrievals over the southeastern Atlantic around Ascension Island, and partition the meteorology into three weather regimes (weakened-high, background, and strengthened-high) with a deep-learning clustering of ERA5 circulation, to quantify the LWP adjustment to droplet concentration (L<sub>0</sub>) by regime and by aerosol vertical placement relative to the cloud top. In the strengthened-high regime, which encompasses the smoky dry season, increasing separated free-tropospheric aerosol steepens L<sub>0</sub> from −0.07 to −0.31, as the shortwave absorption of smoke aerosol enhances the inversion that caps the cloud. This steepening persists despite the moisture co-transported with the plume, which partly offsets the entrainment-driven loss of cloud water. Contacting absorbing aerosol thins the cloud in both the strengthened-high and background regimes, but the LWP–N<sub>d</sub> coupling weakens toward zero in the humid background regime. Combining all days without regime distinction weakens both signals. In contrast, the nudged Energy Exascale Earth System Model version 2 (E3SMv2) produces a uniformly positive LWP adjustment to N<sub>d</sub> across the regimes and aerosol placements, consistent with a precipitation-suppression response. We suggest that cloud-water adjustments in smoke-topped stratocumulus be quantified by weather regime and with vertically resolved aerosol predictors, and that the rain-mass state and sub-cloud updraft in E3SMv2 be calibrated before the simulated adjustments are used to constrain aerosol forcing.