Cloud Phase Determined by GOES-16 Brightness Temperature Differences

March 16th, 2017 |

Channel Difference Field (8.4 µm – 11.2 µm), 2027 UTC on 16 March 2017 (Click to enlarge)

The GOES-16 data posted on this page are preliminary, non-operational data and are undergoing testing.

One of the GOES-16 Band Difference Products available in AWIPS is shown above (click here for the same image with a default enhancement), the 8.4 µm – 11.2 µm ‘Cloud Phase’ Channel Difference. There has been considerable work showing a good correlation between the 8.4 µm – 11 µm brightness temperature difference and cloud-top phase. Click here for example. The key take-aways:

  1. “Radiative transfer simulations indicate that the brightness temperature difference between the 8.5- and 11-micron bands (hereafter denoted as BTD[8.5-11]) tends to be positive in sign for ice clouds that have an infrared optical thickness greater than approximately 0.5. Water clouds of relatively high optical thickness tend to exhibit highly negative BTD[8.5-11] values of less than -2K.”
  2. “Clear-sky BTD[8.5-11] values tend to be negative because the surface emittance at 8.5 microns tends to be much lower than at 11 microns, especially over non-vegetated surfaces.”
  3. “Small particles tend to increase the BTD[8.5-11] values relative to large particles because of increased scattering.”

This Algorithm Theoretical Basis Document (ATBD) contains further information and references on the topic.

In the color enhancement above (generated using just 3 colors — blue, red and yellow, and easily changeable for those whose eyesight is color challenged), negative brightness temperature differences — ice clouds — are denoted by red to yellow values. Positive values are blue. Water-based clouds are white or light blue, ice clouds are red. Large Brightness Temperature Differences occur over the Desert southwest and strong negative values (blue in the enhancement) are present because of emissivity differences from the soil at the two wavelengths. The toggle below cycles through the Band Difference field, the 8.4 µm and 11.2 µm imagery, the upper level water vapor (6.19 µm), the cirrus channel (1.378 µm) and the Veggie channel (0.86 µm).  Click here for a toggle between the Cloud Phase Product and Water Vapor and Cirrus channels only (to highlight ice clouds), or here for a toggle between the Cloud Phase and Veggie Band (a band in which clouds composed of water droplets are visible).

One of the GOES-16 Baseline Products will be Cloud Phase. A beta version of this product should be flowing before the end of May. The ATBD for that product shows that both 8.4 µm and 11.2 µm data are used in its creation, and the channel difference shown here shows why.

Cloud Phase Brightness Temperature Difference (8.4 µm – 11.2 µm), Window Channel (11.2 µm), Infrared Cloud Phase (8.4 µm), Upper Level Water Vapor (6.19 µm), Cirrus Channel (1.38 µm) and Veggie Band (0.86 µm), all at 2027 UTC on 16 March 2017 (Click to enlarge)

Fact Sheets are available for the 8.4 µm and 11.2 µm Channels on GOES-16 from the GOES-R website.

AWIPS Notes: AWIPS sometimes mislabels the 8.4 µm channel as 8.5 µm. The Channel Difference field shows missing data where the brightness temperature difference field is exactly zero. Such points are apparent in the image above over Minnesota and North Dakota as black speckles.

GOES-16 RGB Imagery in AWIPS

March 15th, 2017 |

0.64 µm, 0.86 µm and 1.61 µm imagery and the computed RGB from GOES-16. 1524 UTC on 15 March 2017 (Click to enlarge)

The GOES-16 data posted on this page are preliminary, non-operational data and are undergoing testing.

The ABI on GOES-16 contains 16 Channels, and those channels can be combined into RGB Imagery to highlight features that the individual channels can identify (Click here for general information on RGBs). For example, the ‘Icing RGB’ in AWIPS (also called the ‘Day Land Cloud’ RGB) uses 1.61 µm imagery for the Red component of the RGB, the 0.86 µm for the Green component and the 0.64 µm for the Blue. (This is similar to the oddly-named EUMETSAT ‘Natural Color’ RGB). The toggle above shows the three individual channels, and then the combination in the RGB. A version of the RGB was sent in this Tweet from NWS Lincoln IL.

Cyan regions are those with high values from the green component (0.86 µm) and the blue component (0.64 µm) but little from the red (1.61 µm); such regions include snow on the ground, and/or glaciated clouds. Consider, for example, the toggle below between the 0.86 µm and 1.61 µm imagery.  Lake Effect clouds are distinct over Lake Michigan in both channels, where they show up against the dark background.  Snow on the ground and Water Clouds look very similar at 0.86 µm (or at 0.64 µm, part of the toggle at top of this blog post) and it’s difficult to distinguish clouds from snow over land in a still image.  However, the 1.61 µm imagery is much darker in regions of snow (most of the Midwest United States had snow cover on 15 March 2017).  Water-based clouds show up distinctly against the darker background in the 1.61 µm imagery, and the Lake Effect clouds can be seen easily over Indiana and Michigan. There is apparently some glaciation in the lake effect clouds over land, however, because they do have a cyan tint to them.

Note how the easternmost lake effect band over Lake Michigan shows evidence of glaciation in the clouds.  There is a noticeable change in reflectance between 0.86 µm and 1.61 µm in the toggle below — and that region also shows cyan in the RGB.

0.86 µm and 1.61 µm imagery from GOES-16. 1524 UTC on 15 March 2017 (Click to enlarge)

Over the East Coast, this RGB helps better discriminate between low clouds and high. The example below, also from 1524 UTC on 15 March, cycles through the three channels and then shows the RGB.  The gradual glaciation of the ‘ocean effect’ clouds over the Atlantic is apparent east of New Jersey, as is the glaciation of some of the clouds in the north-south frontal band offshore.  Low clouds are bright in all three channels (0.64 µm, 0.86 µm and 1.61 µm) and therefore appear white-ish in the RGB. Snow on the ground in clear skies is dark in the 1.61 µm imagery and cyan in the RGB.

0.64 µm, 0.86 µm and 1.61 µm imagery and the computed RGB from GOES-16. 1524 UTC on 15 March 2017 (Click to enlarge)

Long-time readers of this blog are familiar with a MODIS-based product that also uses the 1.61 µm channel (in the green and blue) and the visible channel in the red to produce a Snow RGB that has Red snow and cirrus clouds, as shown in this figure from this recent blog post. The key channel for snow-detecting or cirrus-detecting RGBs is the 1.61 µm Channel because ice crystals strongly absorb radiation at that wavelength, reducing the solar reflectance.

Fact sheets are available on the 0.64 µm, 0.86 µm and 1.61 µm Channels on ABI.

====================================================================
Added, 17 March 2017

Icing RGB at 2002 UTC on 17 March 2017 (Click to enlarge)

The red visible (0.64 µm), veggie band (0.86 µm), snow/ice channel (1.61 µm) and RGB, above, gave information about snowcover in the Northeast in the wake of the strong winter storm on 13-15 March. The demarcation between snow and no snow is particularly apparent in central New Jersey. Note snow/land discrimination in the Veggie Band is reduced compared to the visible (click here for a toggle between the two channels) — because of very strong surface reflectance over bare ground. There are northwest-to-southeast streaks in the RGB imagery from southwestern Ontario into northeastern Pennsylvania. These are present because of cirrus clouds as highlighted by the Cirrus Channel at 1.38 µm.  The RGB is also able to distinguish between low clouds over western Pennsylvania, West Virginia and eastern Ohio (that are mostly white in the RGB) and higher ice-laden clouds that are cyan.

AWIPS Note:  Visible (0.47 µm and 0.64 µm) and Veggie Band (0.86 µm) imagery can show missing data in regions of high reflectance near solar Noon, because albedo values then can exceed 1.  When those bands are then used in RGBs, the missing data points are apparent. A fix on this to allow an albedo >1 is in progress.

GOES-16 Multispectral views of the eastern United States

March 14th, 2017 |

GOES-16 Snow/Ice (1.61 µm) animation, from 1100 UTC on 12 March through 1800 UTC on 14 March [click to play mp4 animation]

The GOES-16 data posted on this page are preliminary, non-operational data and are undergoing testing.

The ABI on GOES-16 includes a Snow/Ice Channel at 1.61 µm and a Cirrus Channel at 1.38 µm. These bands offer different perspectives on the evolution of the atmosphere before and during the strong 13-14 March 2017 winter storm on the East Coast of the United States. The Snow/Ice channel, above, is dark in regions where ice clouds or snow on the ground are present because ice is a strong absorber of radiation with a wavelength of 1.61 µm. Water clouds, in contrast, readily reflect such radiation and appear brighter white. Consider the mp4 animation above (available here as a 150-megabyte animated gif). At the start of the animation, on 12 March, snow is indicated over Tennessee, a dark stripe that erodes on that day under the strong March sun. Cirrus retreating southward over the southeastern United States (cloud signals that are grey in the 1.61 µm imagery) reveal much brighter low-level stratus clouds (made of water droplets). Cirrus contrails are apparent above those low clouds. The Mesovortex over the Great Lakes is bright white — water clouds — and is gradually obscured by high-level cirrus clouds from the west and northwest. Terrain-induced wave clouds are also present over Pennsylvania. Their bright color suggests they are composed of water droplets.

On 13 March, low clouds are moving northward over the Piedmont of North Carolina and Virginia as Cirrus clouds spread northeastward from the Deep South. By 14 March, a well-developed wave cyclone is apparent with a large cirrus canopy outlining a warm conveyor belt.

What does the Cirrus Channel, below show? (Click here for a large animated gif.) Strong absorption by water vapor molecules occurs at 1.38 µm. Note, for example, that on 12 March the mesovortex is not apparent in the Cirrus Channel — but the wave clouds over Pennsylvania are. The conclusion is that the atmosphere over the northeast is much dryer than that over the western Great Lakes. On 13-14 March, lake-effect clouds are apparent downwind of Lake Superior in the Upper Peninsula of Michigan and over Wisconsin. The airmass has dried over the Upper Midwest in two days, allowing the Cirrus Channel to view features closer to the surface. In general, the cirrus channel provides outstanding delineation of cloud-top structures over the developing and mature extratropical cyclone.

GOES-16 Cirrus Channel (1.378 µm) animation, from 1100 UTC on 12 March through 1800 UTC on 14 March [click to play mp4 animation]

The Cirrus Channel and the Snow/Ice Channel rely on reflected Solar energy to provide a signal. As such they are useful primarily during the day.

Added: Animations showing the evolution of the three GOES-16 Water Vapor bands during part of the East Coast storm’s lifecycle are available here. A water vapor animation from a CONUS perspective is available here. A better-quality animation centered over the northeast US is available here. Here is an animated gif/mp4 with 6.9 µm brightness temperatures and surface observations. A blog post on this storm is here.

Multi-spectral views of smoke and fire with GOES-16 Data

March 10th, 2017 |

GOES-16 Infrared 3.9 µm images on 7 March 2017 [click to enlarge]

GOES-16 Infrared 3.9 µm images on 7 March 2017 [click to enlarge]

The GOES-16 data posted on this page are preliminary, non-operational data and are undergoing testing.

Tweets on Tuesday 7 March 2017 highlighted the fast-moving fires over the High Plains (that began burning on 6 March), and they also highlighted different bands available from the GOES-16 ABI. For example, this tweet references the loop above, showing an animation of 3.9 µm temperatures; that shortwave infrared channel is used because it is more sensitive to hot temperatures than longer wavelength infrared channels. The Norman WFO also tweeted out imagery, shown below, that included the 0.86 µm ‘Veggie’ band and the 0.47 µm visible band. Why use those two channels?

GOES-16 0.86 µm (near infrared) and 0.47 µm (visible) imagery from 07 March 2017 [click to enlarge]

GOES-16 0.86 µm (near infrared) and 0.47 µm (visible) imagery from 07 March 2017 [click to enlarge]

The 0.47 µm imagery is observing a part of the visible electromagnetic spectrum where scattering is largest, so smoke plumes are more apparent at that wavelength than at 0.64 µm. For a very obvious event such as this one, this might not be as important, but for a modest fire event over Florida, shown next, it can be. The 0.86 µm imagery is useful because it very distinctly shows fire burn scars; that is, the contrast at 0.86 µm between vegetated soil and adjacent burned regions is greater than occurs at other visible wavelengths. That is shown in the toggle below that steps through 0.47 µm, 0.64 µm, 0.86 µm, 1.61 µm and 3.9 µm imagery for one time on 7 March. The smoke plume is most distinct at the shortest wavelength 0.47 µm; it is very difficult to discern at 0.86 µm and especially at 1.61 µm because these near-infrared channels sense radiation at longer wavelengths that is unaffected by scattering of light by the small smoke particles. Note, however, that the small lakes do jump out at both wavelengths because of the very different reflectance properties of land and water at both 0.86 µm and 1.61 µm.

Finally, compare the 0.64 µm and 0.86 µm with special focus on the burn scars (here is a toggle between the two). Although the spatial resolution is greatest in the 0.64 µm visible imagery (0.5 km at the sub-satellite point, vs. 1 km at the sub-satellite point for the 0.86 µm imagery), the burn scars nevertheless are more distinct at 0.86 µm, in part because vegetated ground is more reflective at 0.86 µm than at 0.64 µm (See the figure in ‘Tim’s Topics’ on page 2 of the 0.86 µm fact sheet).

GOES-16 imagery from 2227 UTC on 07 March 2017. Wavelengths indicated in the image [click to animate]

GOES-16 imagery from 2227 UTC on 07 March 2017. Wavelengths indicated in the image [click to animate]


=========================================================================
A less extensive fire event occurred on 10 March 2017 in Florida. Focus on the largest hot spot (black pixels) in the 3.9 µm imagery in the center of the top third of the image below; this point is in southeastern Polk County. For this event, the smoke plume is more easily visualized in the 0.47 µm imagery than in the 0.64 µm or the 0.86 µm imagery. A burn scar does not appear in this case.

GOES-16 imagery from 1931 UTC on 10 March 2017. [click to animate]

GOES-16 imagery from 1931 UTC on 10 March 2017. [click to animate]

The GOES-R Website includes Fact Sheets for Band 1 (0.47 µm), Band 2 (0.64 µm), Band 3 (0.86 µm), Band 5 (1.61 µm) and Band 7 (3.9 µm).

AWIPS Note: The default enhancement (“IR_COLOR_CLOUDS_WINTER”) for 3.9 µm results in imagery that shows too little gradation over Florida during the daytime; for fire detection, either modify the colormap (this changed the temperature range from the default [-109 to 55] to -70 to 75, and is shown above) or switch to the”IR_COLOR_CLOUDS_SUMMER” enhancement.