Cloud and clear-sky surface parameters are obtained from the International Satellite Cloud Climatology Project (ISCCP). Release 4-IP uses the newly released H series pixel level product, while the previous SRB version 3 used the D series. Details of the ISCCP-H Series can be found in Young, et al. (2018) and Rossow (2017).
Atmosphere and surface meteorology (temperature and water vapor) for this release mainly originates from the ISCCP NNHIRS dataset (Rossow 2017). The dataset adapts the HIRS neural net retrieval by Shi et al. (2016) by adding levels to the upper atmosphere, clearing cloud-contaminated values, and filling in missing data.
For the longwave flux calculations, there are several modifications to the input meteorology. Near-surface over land, the LandFlux temperature and humidity values are used, while over ocean the Seaflux V2 values are used. Because of temporal filling inherent with NNHIRS and dataset assessments of NNHIRS, MERRA-2 reanalysis is substituted in the lower levels over oceans.
Total column ozone abundance is an ancillary product available with the ISCCP-H data series. Total Ozone Mapping Spectrometer (TOMS), TIROS Operational Vertical Sounder (TOVS), Stratospheric Monitoring-group Ozone Blended Analysis (SMOBA) and OMI instrument aboard Aura satellite data are combined through regression fits then spatially and temporally filled (Rossow 2017).
Ozone profiles for GLW are based on the GOZCARDS data files which give zonal merged ozone profiles for 25 pressure levels (Froidevaux et al. 2015). This replaces the Flemming profiles used in Rel. 3.1. Within the GLW algorithm, GOZCARDS data is interpolated to the pressure levels of the grid box meteorology and cloud data.
Aerosol includes a fully variable composition aerosol, using the Max-Planck Aerosol Climatology (MAC, Kinne et al., 2013). This data, in addition to the aerosol optical depth, includes asymmetry parameter and single scattering albedo. In order to avoid dramatic steps from one month to the next, an implied daily value of all optical properties is used (Zhang et al. 2017). The daily values are chosen to vary smoothly while preserving the correct monthly average.