Global Aerosol Direct Effect Retrieval from Passive Hyperspectral Measurements
de Graaf, Martin; Tilstra, Gijsbert; Stammes, Piet

Absorbing aerosols can have a significant local direct radiative effect (DRE), while the global average aerosol DRE remains highly uncertain. Modelling studies have shown that the magnitude and sign of the aerosol DRE at the top of the atmosphere (TOA) depend on the scene, especially on the albedo of the scene under the aerosol layer. It changes with cloud fraction, from large positive for overcast conditions when aerosols are present above the cloud, to large negative for clear sky ocean scenes. Observational studies, which are necessary to constrain the model studies, have been scarce. So far, the cloud fraction for which the aerosol DRE changes from positive to negative has been constrained by only one observational study, using active satellite instruments.

The results of modelling studies depend heavily on the assumed aerosol properties. Observational studies also need to assume aerosol type and geophysical properties to derive aerosol optical properties from radiation measurements. This introduces large uncertainties in the retrieved aerosol DRE. Furthermore, the retrieval of aerosols over clouds from passive instruments is difficult, due to the large optical thickness of clouds. Therefore, observational studies of aerosol direct and indirect effects from passive satellite instruments are invariably restricted to aerosol studies close to the cloud edges. We have developed a method to derive the aerosol DRE for smoke over clouds directly from passive satellite hyperspectral reflectance measurements, independent of aerosol microphysical property assumptions. This allows us to assess the local aerosol DRE from passive imagery directly on a pixel to pixel basis, even over clouds.

The solar radiative absorption by smoke layers is quantified using the TOA reflectance spectrum from the ultraviolet (UV) to the shortwave infrared (SWIR). UV-absorbing aerosols have a strong signature that can be detected using UV reflectance measurements. Since the aerosol extinction optical thickness decreases rapidly with increasing wavelength for smoke, the properties of the scene below the aerosol layer can be retrieved in the SWIR, where aerosol extinction optical thickness is sufficiently small. Then, using radiative transfer computations, the contribution of the scene to the reflected radiation can be modelled for the entire solar spectrum. In this way, aerosol effects can be separated from all other effects in a scene. Aerosol microphysical assumptions and retrievals are avoided by modelling only the aerosol-free scene spectra. All the aerosol effects are in the reflectance measurements. The method works especially well for bright scenes, e.g. scenes with clouds underlying the absorbing aerosol layer.

In an initial study, supported by ESA within Support To Science Element, an algorithm was developed to derive the aerosol DRE over marine clouds, using the spectrometer SCIAMACHY, which produced shortwave reflectance spectra from 2002 till 2012. The reflectance spectra from SCIAMACHY are ideally suited to study the effect of aerosols on the entire shortwave spectrum. The algorithm has been improved and adapted to suit data from any instrument, or a combination of instruments, that measures UV, visible and SWIR reflectances at TOA simultaneously. Examples include OMI and MODIS, flying in the A-Train constellation, and TROPOMI, on the future Sentinel 5 mission, combined with NOAA's NPP VIIRS. This would produce aerosol DRE estimates with unprecedented accuracy and spatial resolution.

We show results from SCIAMACHY and initial results from OMI/MODIS. The aerosol DRE over clouds over the South Atlantic Ocean west of Africa, averaged through August 2006 was found to be 23 ± 8 Wm-2 with a mean variation over the region in this month of 22 Wm-2. Locally the aerosol DRE over clouds in that month was as high as 132 ± 8 Wm-2, absorbing about 10% of the local incoming solar radiation.

Figure 1. Example of aerosol absorption retrieval from a SCIAMACHY spectral reflectance measurement (red) on 13 August 2006, taken over the South-Atlantic Ocean, when smoke was overlying a marine cloud deck. The various scene and cloud parameters were taken from the measurements as indicated. Note the retrieval of cloud optical thickness τcld and cloud effective radius reff where the aerosol absorption becomes negligible. From these parameters the aerosol-free cloud scene reflectance spectrum is computed (blue). The difference is caused by the absorption of radiation by smoke (yellow). This spectral reflectance difference can be converted into aerosol DRE directly.