Horizontal aerosol transport, not local shear, drives aerosol–boundary layer decoupling: lidar evidence from contrasting regimes in Thailand

Abstract. Aerosol vertical structure is often inferred from lidar observations under the assumption that the dominant aerosol gradient approximates the planetary boundary layer (PBL) height. This assumption may not hold when aerosol layers decouple from boundary-layer turbulence. We investigated regime dependence of aerosol-boundary layer coupling during June 2021 to May 2022 at two contrasting sites in Thailand: an inland valley (Chiang Mai) and a coastal site (Songkhla). Results are interpreted within a regime-based framework rather than as climatology. The aerosol layer top (ALT) is compared with a thermodynamically constrained PBL height, with decoupling defined as ALT minus PBL exceeding 0.5 km and quantified using frequency, diurnal and wind-conditioned occurrence, event duration, and autocorrelation timescale. The inland site exhibits frequent (around 53 %) and persistent decoupling with longer autocorrelation timescales (around 5.6 h), indicating thermodynamic control and residual-layer influence. Richardson number diagnostics show 83 % of decoupled strong-wind events occur under stable stratification, and back-trajectory analysis confirms elevated aerosol layers originate from distinct source regions (median separation 502 km), attributing decoupling primarily to aerosol advection rather than in-situ shear lofting. The coastal site shows lower decoupling (around 20 %), weaker diurnal contrast, and shorter persistence (around 3.1 h), reflecting mechanically regulated mixing. Decoupling frequency uncertainty is 0.9 % at Chiang Mai and 9.0 % at Songkhla, leaving the regime contrast robust. The stable stratification is consistent with temperature inversions widely invoked in Thailand's haze discourse, suggesting aerosol horizontal transport may co-equally drive elevated pollution layers alongside thermodynamic trapping, with implications for air quality communication and source attribution.