Aerosol Detection - Total Operational Weather Readiness - Satellites (TOWR-S)
GOES-R Aerosol Detection
Frequently Asked Questions about the GOES-R Aerosol Detection Product
1) What is this product?
The GOES-R Aerosol Detection product identifies aerosols and classifies them as either dust or smoke. The resolution of this product is 2km. The GOES-R Aerosol Detection product is created only during the daytime hours. It is accurate for dust 80% of the time over land and 80% of the time over ocean. It is accurate for smoke 80% of the time over land and 70% of the time over ocean. The GOES-R Aerosol Detection product is reliable out to 60 degrees local zenith angle and can be prone to errors beyond that.1
This product is useful when trying to identify the type of aerosol present in the atmosphere. Being able to differentiate between smoke and dust can be helpful when the source of the aerosol is ambiguous. For example, a small fire may produce smoke, but not be identifiable by the GOES-R Hot Spot algorithm.
2) How often do I receive this data?
The cadence of the Aerosol Detection product is the same regardless of scan mode. That is, the product is produced every 15 for Full Disk, CONUS, and Mesoscale images.
3) How do I display this product in AWIPS-II?
To display this product in AWIPS-II, go to the "GOES-R" tab of the CAVE menu, then select "Derived Products." From there, select the region of interest (GOES-E, GOES-W, or GOES Test and Full Disk, CONUS, and Mesoscale). Then, select the Aerosol Detection product.
Alternately, use the AWIPS Product Browser. Select "Sat", then either "GOES-16" or "GOES-17". From there, choose "Full Disk", "CONUS", or "Meso", then select "AD."
4) How do I interpret the color maps associated with this product?
Color tables have not yet been created for this product.
5) What other imagery/products might I use in conjunction with this product?
This product can be used in conjuction with aerosol optical depth (AOD) product. While the Aerosol Detection product informs the user what type of aerosol is present in the atmosphere, AOD informs how much of that aerosol is present in the atmosphere. By combining the information from both products, a user can infer the intensity of the dust or smoke plumes associated with exceptional events. This product can also be used in conjunction with visible 0.64um satellite imagery to identify the location of aerosols. The 0.47um channel can also assist with identifying the location of aerosols. Although these channels are used in the algorithm itself, it can be beneficial to see these individual components in case the algorithm is incorrectly identifying aerosol that is not present or missing aerosol that is present. Differentiating between smoke and dust can be beneficial to a forecaster because simply looking at visible imagery channels can be misleading to tell which aerosol type is present. Smoke being present can indicate a small fire not visible in satellite imagery is burning. Also, lofted dust can be a precursor to a larger dust event in the future.
6) How is this product created?
The GOES-R Aerosol Detection product utilizes GOES-R bands 0.47 um, 0.64 um, 0.86 um, 1.37 um, 1.6 um, 2.2 um, 3.9 um, 11.2 um, and 12.3 um. The backbone of the algorithm is the distinctive spectral and spatial signature of aerosol (smoke/dust). Temporal variability has not been taken advantage of, in the current version of algorithm, but is planned for future versions. Similar to clouds, variability of smoke or dust plume is much larger than the surface over a course of day. Besides the threshold test, by tracking the variability over time, for example, variability over a course of 30 minutes, it is possible to define if a pixel is laden with smoke/dust. However, it must be noted that cloud, smoke and dust may have similar temporal variability. Taking advantage of temporal variability in smoke/dust detection has high requirement on separating clouds from smoke/dust. In addition, different ABI channels have different spatial resolution, ranging from 0.5 km for visible to 2 km for IR channels. In the algorithm, the output resolution is 2km. Hence, channels with higher spatial resolution than 2 km have to be aggregated to 2 km by sub-sampling before applying the algorithm. Like any other threshold-based algorithm, the algorithm requires optimal performance of the instrument. First, the ADP algorithm is designed to work when only a sub-set of the expected channels are available. Missing channels, especially the crucial ones, will impact directly the performance of the algorithm.1
The ADP algorithm is sensitive to instrument noise and calibration error. Thresholds are required to be adjusted accordingly to the status of instrument operation and performance. Calibrated observations are also critical, but since the algorithm does not compare the observed values to those from a forward radiative transfer model, uncertainties in calibration can be ameliorated by modifying thresholds post launch of the ABI.
The ancillary data used in the Aerosol Detection product include:
- Land/Water Mask: The only static input data required by the algorithm is a global 1km land/water mask. The global land cover classification collection created by The University of Maryland Department of Geography with Imageries from the AVHRR satellites acquired between 1981 and 1994 is the source (http://glcf.umiacs.umd.edu/data/landcover/).
The derived data used to calculate the cloud top height includes:
- Cloud Mask: The purpose of using cloud mask in the algorithm is to eliminate pixels with obvious clouds, such as high and ice cloud, before performing smoke/dust detection. Hence, the requirement of the algorithm for cloud mask is more specific than just cloud or clear mask. Stringent cloud mask has the potential to classify smoke as cloud, while loose cloud mask increases the chance of misidentifying clouds as smoke. The algorithm intends to use only individual tests in ABI cloud mask product which indicate the existence of high cloud, ice cloud and thin cirrus cloud. However, this dependency was not tested because ABI cloud algorithm was not ready to run on MODIS. Efforts will be put once common proxy data become available. Currently, the algorithm is using MODIS data as proxy, including MODIS cloud mask. Based on the definition of individual test from both ABI cloud mask and MODIS cloud mask, the individual test used in algorithm is mapped to ABI cloud mask and they are given in Error! Reference source not found.
- Snow/Ice Mask: Primary source of snow/ice is ABI Level-2 Snow/Ice Product. However, under the situation that the primary source is missing, Interactive Multisensor Snow and Ice Mapping System (IMS) (http://nsidc.org/data/g02156.html) snow/ice mask will be the secondary source. In addition, the ADP algorithm has internal snow/ice test over land, whose function is to eliminate the residuals from external snow/ice mask over land. It is applied after the primary /secondary snow/ice mask.
- Sun Glint Mask: The algorithm is designed to generate internal sun glint mask based on ABI viewing and illuminating angles as second source. An area with calculated sun glint angle greater than zero and less than 40 degrees is defined as sun glint area.
- Day/Night Mask: A day/night flag is determined internally based upon the solar zenith angle. Day is defined as solar zenith angle of less than 87 degrees, while night is defined as solar zenith angle greater than 87 degrees.
The algorithm attempts to separate cloudy and clear pixels from those with smoke or dust. The detection of smoke or dust relies on the distinctive signature of smoke or dust which is often expressed in terms of spectral variations of the observed brightness temperature or solar reflected energy. The spectral variation of the refractive index plays an important role in the success of these methods. In addition, the scattering and absorption properties of an aerosol also depend on the particle size distribution and the particle shape. Several aerosol remote sensing techniques have been developed using observations from the Advanced Very High Resolution Radiometer (AVHRR). Similar to the dust plumes, the volcanic ash plumes often generate negative brightness temperature differences between 11μm (BT11) and 12 μm (BT12). Prata  has demonstrated the detection of volcanic aerosols using two infrared channels, while Ackerman and Strabala  applied observations at 8.6, 11 and 12μm from the Hyper Spectral Infrared Sound (HIRS) instrument to study the Mt. Pinatubo stratospheric aerosol.
Image based aerosol detection always involves assumptions of the radiometric characteristics of aerosol, clear and cloudy scenes. The surface conditions also influence the separation of aerosol pixels from those with clear-sky or cloud. The algorithm currently uses spectral and spatial tests to identify pixels with smoke or dust in the daytime. Temporal tests are planned for future versions of the algorithm. The algorithm also treats the detection differently for water and land.1
1NOAA NESDIS Center for Satellite Applications and Research Algorithm Theoretical Basis Document: ABI Aerosol Detection Product v.2.0. 30 September 2010. http://www.goes-r.gov/products/ATBDs/baseline/AAA_AIP_v2.0_no_color.pdf