Detecting the layer height of smoke and dust aerosols over land and ocean using ultraviolet dual-wavelength measurements

Vertical distribution of atmospheric aerosols has a significant impact on climate change, air quality, cloud-aerosol interactions, atmospheric remote sensing, and global transport. However, current aerosol layer height (ALH) retrieval algorithms, based on passive radiometry, are largely affected by surface reflectance, limiting their applicability to surface with low reflectance, for example, ocean and dark-target land surfaces. The Aerosol Single-scattering albedo and layer Height Estimation (ASHE) algorithm can be well adapted to both land and ocean, but it is strongly influenced by aerosol optical depth (AOD) uncertainty and requires estimation of expected aerosol-free radiation. In order to reduce the effects of AOD uncertainty and to simplify the method, a UV band difference quotient coefficient (UVD) method is developed in this study. First, in the retrieval process, UV aerosol index (UVAI), AOD, and Ångström exponent (AE) are incorporated to perform cloud screening and aerosol type classification (distinguishing between smoke and dust). Next, utilizing the high sensitivity of Rayleigh scattering and the wavelength dependence of aerosol absorption, the logarithmic difference of UV dual-band normalized radiance $\ln \left(r_{340}\right)-\ln \left(r_{378}\right)$ is used to capture and amplify the impact of ALH on radiance differences. By incorporating the normalization effect of the two-band difference $r_{340}-r_{378}$, the effects from AOD uncertainty are partially offset, resulting in a monotonic, stable, and sensitive response of the UVD to the ALH. Finally, lookup tables (LUTs) for smoke and dust aerosols are built to retrieve ALH. The UVD method was validated using CALIOP observations from a series of well-documented North American biomass burning events and Atlantic dust storms. The validation results indicate that the RMSE ranges from 0.67 to 1.42 km (mean: 0.99 km), while the bias varies between −0.62 and 0.57 km, with a correlation coefficient of 0.71. Additionally, the error distribution of UVD-CALIOP is concentrated within the range of −1 km to 1 km, with 75 % of the data points having an error within 1.13 km.