Daily global cloud droplet number concentrations (Nd) have been calculated at 1x1 degree resolution from pixel-level MODIS (MODerate Imaging Spectroradiometer) Collection 5.1 Joint Level-2 (Aqua satellite) optical depth (tau) and the 3.7 micron effective radius (reff) data (and other supporting data) using the adiabatic cloud assumption (liquid water content increases linearly with height, Nd is constant throughout the cloud depth and the ratio of the volumne mean radius to the effective radius is assumed constant).

The Nd data is contained in separate NetCDF files for each year for the period 2003-2015. Nd is contained in the "Nd" variable and has units of cm^{-3}. This is a 360x180xNdays (lon x lat x Ndays) sized array, where Ndays is the number of days in the year. The lon x lat grid is a regular 1x1 degree grid. The time is provided as both a 1D array of size Ndays ("time") with units of days since 1st Jan, 1970 and an array of size Ndays x 3 ("time_vec") that contains numbers for the year month and day for each of the Ndays entries.

A number of filters have been applied to the data in order to remove retrievals that are likely to be problematic, or to violate the adiabatic cloud assumptions. Data is only included if:

1) Pixels are determined to be liquid pixels by MODIS.
2) The 1x1 degree mean cloud top height (calculated using the MODIS cloud top temperature and the sea surface temperature) is below 3.2km.
3) The 1x1 degree liquid cloud fraction was larger than 80%.
4) The 1x1 degree mean solar zenith angle was 65 degrees or less to avoid biases at high angles (Grosvenor and Wood, 2014).

Note, that the filtering is different to that described in Grosvenor, AMTD, 2018 in the following ways :-

1) 1km resolution tau and reff are used to calculate Nd, which is then aggregated to 1x1 degree resolution (rather than using 1x1 degree tau and reff).
2) Only Nd based on the 3.7 micron reff retrieval is provided here.
3) No filtering for the presence of sea-ice is done here - it is recommended that this is done if using for high latitudes.
4) The data here is not restricted to tau>5.

Also note that the vertical penetration bias correction described in Grosvenor, AMTD, 2018 is NOT applied here. In addition, as described in the latter paper, further pixel-level screening is performed in order to select high quality data.

Details on the reasons for restricting to low solar zenith angles can be found in Grosvenor and Wood, ACP, 2014. Information on the pixel level filtering applied can be found in Grosvenor et al., AMTD, 2018 (noting the differences explained above). A comparison of this dataset with others can be found in Grosvenor et al., Reviews of Geophysics, 2018.

This dataset calculates a product that is not provided as standard by MODIS. It uses improved optical depth and effective radius data compared to the standard MODIS Level-3 data since situations (e.g., high solar zenith angles, broken clouds) that have been shown to cause retrieval issues have been filtered out at the Level-2 stage before being averaged into Level-3 droplet concentration data.