Aerosol and Cloud Detection

Aerosol and Cloud Detection Software Package for High Resolution Infrared Sounders

Current version: v3.1, September 2020

The cloud and aerosol detection software identifies infrared (IR) sounder channels that are affected by contamination from cloud, aerosol, or excessive trace gas concentrations. The software also indicates channels that are sensitive to land surface emission.
The software consists of four mutually independent modules:

  1. The cloud detection algorithm works by taking the first guess (FG) departures (i.e., the departure of brightness temperature observation from its expected value, calculated from a short-range NWP forecast) and looking for the signature of cloud. As clouds are usually colder than underlying surface (or air below them), the contamination tends to make observations colder than their NWP-based counterparts, resulting in negative FG departures.
  2. The aerosol detection algorithm compares IR window channel brightness temperature observations on both sides of the 9.6 micrometer ozone absorption band. Presence of aerosol is diagnosed if pre-defined threshold values are exceeded. Depending on brightness temperature observations at strategically chosen spectral positions, the aerosol detected over sea is interpreted as either Saharan dust, volcanic ash, or other aerosol. The affected channels depend on the aerosol type classification.
  3. The trace gas detection algorithm computes mean observation and mean FG departure within a set of tracer channels and compares those against reference values computed from a set of control channels. The tracer channels are sensitive to the targeted trace gas, while control channels are not. Trace gas contamination is diagnosed where the tracer minus control differences (in both observation and FG departure) fall below pre-defined thresholds. Currently only one target trace gas is supported (Hydrogen Cyanice, HCN).
  4. The land sensitivity detection algorithm compares input land fraction and vertical channel height assignments with pre-defined thresholds. Given that the channel height assignments are calculated from a short-range NWP forecast, this method allows to identify land-sensitive IR channels in a situation-dependent manner.

The software package is NWP-dependent, such that it requires external input that can be produced from an NWP system with a radiative transfer model. The software is sufficiently modular to “plug-in” to most NWP systems.

Input data: For each channel selected in the IR field-of-view, the algorithm requires the background (NWP model -based) brightness temperature, the observed brightness temperature and a height assignment for each channel in units defined by the user (e.g., NWP model level, pressure level). The algorithm also requires input land fraction in the IR field-of-view.

Output data: Output consists of binary flags that indicate those IR channels that are affected by cloud, aerosol, trace gas, or land sensitivity. The four output flags are independent of each other.

Further information

  • McNally, A.P. and P.D. Watts, 2003. A cloud detection algorithm for high-spectral-resolution infrared sounders, Q J Roy Meteorol Soc, 129, 3411-3423.
  • Eresmaa, R., 2014. Imager-assisted cloud detection for assimilation of Infrared Atmospheric Sounding Interferometer radiances, Q J Roy Meteorol Soc, 140, 2342-2352.


Aerosol-and-Cloud-Detection-related Publications