Regional and Mesoscale Meteorology Branch

At temperatures below -20 degs. C, clouds consist mostly of ice particles, while below -40 degs. C they are composed entirely of ice particles. However, between 0 degs. C and -20 degs. C, a significant number of clouds may be primarily composed of water droplets. Water droplets below 0 degs. C are supercooled, and clouds containing large, supercooled droplets can pose an extreme hazard to aviation. During day-time, 3.9 um imagery can be used to infer droplet phase at cloud top because of differences in reflection between ice particles and water droplets, as discussed earlier in the section on emission and reflection. By using the 3.9 um imagery to identify phase, and the 10.7 um imagery to determine cloud top temperature, cloud tops consisting of supercooled water droplets may be located.

During the night-time hours, water clouds can also be distinguished from ice clouds by using the “fog product” (see discussion on “night-time fog”). Similar to the day-time application described above, the “fog product” and the 10.7 um imagery can be used together to locate cloud tops consisting of supercooled water at night.

The user should keep in mind that when multilayered clouds are present, locating supercooled clouds may not be possible because of other cloud decks obscuring them from the satellite’s view.

Night-time and day-time examples of supercooled water identifying imagery are presented on the following display pages.

Because it distinguishes water clouds from ice clouds, regardless of temperature, the “fog product” can aid in the detection of supercooled water clouds at night, when used in combination with other data. This is done by locating the water cloud area within the “fog product”, and then using the 10.7 um data, along with a representative rawindsonde or model sounding, to determine if cloud top temperatures and atmospheric vertical structure are within appropriate boundaries for supercooled water clouds.

An example is the fog product image (at top left) from 0345 UTC, on 9 Jan 1996. The bright regions are water clouds, while the noisy and darker regions are ice clouds. The area around Buffalo, NY (BUF) is covered by water clouds. The corresponding 10.7 um image (top right) shows the brightness temperature of those water clouds to be near -12 C. Similar cloud top and ground temperatures make it difficult to locate the cloudy area using the 10.7 um imagery alone. The 00 UTC sounding, from Buffalo on the 9th, shows a near-isothermal cloud deck just below 700 mb (hPa), with a temperature near -10 C. The entire water cloud region shown in this image must almost certainly be supercooled.

As with the night-time “fog product”, during the day-time the 3.9 um imagery can be used, in association with the 10.7 um data, to find areas of supercooled cloud. In this example, the 3.9 um image (top left) from 1445 UTC on 10 Jan 1996 is displayed as energy (see the earlier discussion on imagery presentation). The areas covered by water cloud appear bright, due to the addition of the reflected solar component. Water cloud can be seen in the Buffalo vicinity, to the south of Lake Michigan and other areas.

The corresponding visible image (top right) also shows these water cloud areas. An estimate of cloud top temperature from the corresponding 10.7um image (right) determined that for the clouds around Buffalo, typical cloud top temperatures were about -24 C, which, along with the 3.9 um data, indicates a supercooled cloud. The 1200 UTC sounding from BUF on the 10th shows cloud top temperatures to be around -20 C, an important verification. (This is especially true if additional cloud layers exist above the suspected supercooled water cloud, since they would affect the 10.7 um brightness temperatures.)

Snow cover can be easily located in GOES imagery by taking advantage of its appearance differences in the visible, the 3.9 um and the 10.7 um channels. Snow will appear fairly bright in the VIS imagery, depending upon its age. Additionally, landmarks such as rivers and lakes are often very well defined as dark areas, surrounded by smooth, white snow fields as seen here, over northeast OK.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

In the 10.7 um band such landmarks appear as small, warmer segments within the cooler areas since snow cover is usually colder than nearby snow-free areas. This same snow-covered region will appear black in the day-time 3.9 um imagery due to its poor reflectivity, with the warmer lake and river areas having distinct shapes.

Mid- or upper-level cloud can be detected over snowfields when landmarks become obscured as high clouds move across them, and/or by the cloud-to-underlying-snow temperature differences. Imagery at 6.7 um will also show cirrus, but not the ground. Observe the 10.7 um imagery over northeast OK and southeast MO for an example of upper-level cloud versus snow discrimination.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

Prior to GOES-8, it was difficult to identify low clouds over snow. VIS imagery typically shows little difference between snow and low cloud or fog, while at 10.7 um they often are nearly the same temperature. Day-time 3.9 um imagery, distinguishes low cloudiness and fog from snow, due to reflectivity differences.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

In this example, most of IA and WI are covered by snow. As in the earlier case, a good portion of this snow cover can be readily identified in the VIS imagery by looking for dark, geographical features against the brighter background of snow.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

Notice, in the 3.9 um imagery, the bright cloud mass moving to the southeast across MN. These are water clouds, as determined by their high reflectivity.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

AVHRR data has been used for many years to provide estimates of sea surface temperature (SST). The potential exists to supplement AVHRR SST by using GOES data. This is because GOES can observe an area as frequently as once every 15-to-30 minutes, providing a greater probability of cloud free observations than the AVHRR, which may view that same area only twice a day. During daylight, the GOES 10.7 and 12 um channels can be used to correct surface radiance temperatures that have been corrupted by low-level water vapor absorption. At night-time, information from the 3.9 um channel can be added to that from 10.7 and 12 um to improve the SST product. At 3.9 um, moisture contamination is less since it is a cleaner window; diffraction is less, which effectively means higher resolution; and cloud sensitivity is less.

This night-time image example is an average of five consecutive half-hourly images from the 3.9 um channel. Cold land areas are purple and white, lakes and water along the coast are warmer than the land (light green and blue) but cooler than the warmer waters in the Gulf Stream, which are red and yellow. By averaging the images, noise is greatly reduced and the sharp temperature gradients over the water are easier to see.

Radiation from below passes through thin cirrus clouds, making the satellite-measured IR temperatures warmer than the actual cloud top temperatures. This effect is more evident at 3.9 than at 10.7 um because of the stronger response at 3.9 um to the warm radiation from below. In addition, thin cirrus is often patchy and only partially fills an FOV, further enhancing response at 3.9 um. As a result, in regions of thin cirrus, 3.9 um images often reveal lower cloud layers. At night, the underlying clouds may have different motions, leading to their detection with animated imagery. During daytime, water clouds, with their higher reflectivity, can be detected at 3.9 um, while they are obscured, or very difficult to observe, in the VIS and 10.7 um imagery.

In this daytime example, a small water cloud, just south of Georgian Bay (encircled in light-blue), is moving southward and appears as a bright spot in the 3.9 um image, while it is obscured by thin cirrus at 10.7 um.

Use caution when interpreting 3.9 um images in the presence of thin cirrus. Energy from below will increase the detected radiance and may result in image shades very close to other cloud scenes, and even clear regions. Corresponding 6.7 um imagery is useful in isolating this effect; see the section on Winter Storms for another example.

Imagery from the 3.9 um band of the GOES Imager makes it possible to locate urban heat “islands” under clear sky conditions, especially during the night-time hours. As discussed earlier, there is inherent variability in land-surface temperatures over distances of fractions of a kilometer, even more so around cities and their suburbs. The GOES FOV over cities may be thought of in the same sense as partially filled fields-of-view over fires. This results in higher radiance temperatures being measured at night over cities at 3.9 um than at 10.7 um.

In this example, the signatures of several Midwestern cities can be seen in an image from the early morning of 14 July 1995. The cities are darker (warmer) than their surroundings.

Its strong sensitivity to sub-pixel “hot-areas” makes the 3.9 um channel very useful in fire detection. Fig. 5a may be used to compare the temperatures at 3.9 and 10.7 um with the percent of a pixel covered by fire, where the hot-area is at 500 K and the remainder of the pixel is at 300 K. Note that if only 5% of the pixel is at 500 K, the 3.9 um measured brightness temperature of that pixel is 360 K, while the corresponding 10.7 um brightness temperature is less than 320 K.

Fig. 5b shows the DIFFERENCE in the brightness temperature for the 3.9 and 10.7 um pixel, hypothesized in this example. Note how much more responsive the 3.9 um channel is over smaller fire areas.

NOTE: Refer to the text with Fig. 2c, in the “Energy Sources” section, for an important caveat regarding 3.9 um channel saturation.

This map shows the area covered by a fire which began on the morning of 1 July 1994 in northern Colorado, near Colorado State University’s Pingree Park mountain campus. By the time of this map, nearly 800 acres (one square mile is equivalent to 640 acres) had been scorched and two major spot fires were burning apart from the primary fire. The fire had travelled across the northwestern edge of the campus, totally destroying two buildings and heavily damaging seven others.

The information on this map was provided by the U.S. Forest Service and local firefighters who were on the scene during the fire. Temperature estimates were derived from a series of about two dozen interviews with those personnel.

The three pixels pointed to by the arrow in this 3.9 um image are those being directly affected by the Pingree fire at 9:00 PM LST (time of map, previous page). The GOES is viewing an unobscured fire scene. At this time, the only clouds in the region were a scattered, thin cirrus deck not over the fire scene (fires can be observed through thin cirrus at 3.9 um). VIS imagery showed smoke from the fire was advected eastward, and with the satellite viewing from nearly due south, smoke obscuration was minimal. Finally, a significant part of the hottest burn was in the crowns of the trees, eliminating the possibility of tree canopy obscuration of a brush fire below, allowing the GOES a nearly unobscured view of the fire scene.

The 3 rows of numbers show the temperatures (K) of the pixels in the immediate proximity of the fire, including the three fire-affected pixels (asterisked).

During day-time, visible imagery can be used to infer fire locations from smoke plumes. Smoke from larger fires is particularly easy to see in visible image loops, appearing as light-colored, hazy plumes emanating from a small area. Observe the smoke plumes produced by very large, controlled burns near the cities of Pensicola and Apalachicola, FL (both, near the Gulf coast) in the loop below. Arrows are added to identify the positions of the two larger Florida smoke plumes, as well as two smaller plumes, one in MS, and the other in AL.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

Smoke from smaller fires, including the two smaller plumes just mentioned, is not always evident in VIS imagery. This will become apparent when the 3.9 um imagery for this case is presented on the next “display page” (click on “Continue” below).

The 3.9 um image shown above corresponds to the VIS imagery on the previous page. These day-time images are displayed as reflectivity (see earlier discussion on emission and reflection in the “Basic Radiation Science” section), so that lighter shades are hot and darker shades are cool. Animation reveals small, bright pulses at several locations; each one showing controlled-burn fires. Notice that there are considerably more fires than were apparent in the VIS images. At 3.9 um, smoke plumes are difficult to detect because of poorer spatial resolution (when compared with VIS imagery) and poorer reflective properties. In this example, smoke can be observed as very subtle, dark areas emanating from the two larger fires.

Notice that the fire pixels seem to be “smeared” in the east-west direction. This is due to: 1) sensor lag, a part of the signal from hot spots is retained by the sensor, affecting the values of at least one other pixel down-scan from the hot spot; and 2) oversampling, the 3.9 um IR sensor observes an area of 4 x 4 km and is oversampled by a factor of 1.75 in the east-west direction.

Twelve controlled burns can be found in this animation sequence; the arrows are placed to help locate seven of them.

An image product that approximates reflected energy at 3.9 um shows exceptional promise for a number of applications. This so-called “reflected product” is derived by equating the 10.7 um channel brightness temperature to a corresponding radiance value at 3.9 um, for each pixel within an image. That radiance value is then subtracted from the actual 3.9 um radiance, resulting in an approximate reflected radiance at 3.9 um. The 3.9 um “reflected product” shows water clouds as bright white, and poorly reflective ice clouds and snow as dark grey shades. These images show that clouds are not easily distinguished from snow cover when only the VIS channel is viewed; however, the addition of information from the 3.9 and the 10.7 um channels, along with the 3.9 um “reflected product”, aid greatly in discriminating between snow cover, clear ground, ice cloud and water cloud.

NOTE: This product’s utility for night-time cloud discrimination is also under investigation.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

Biomass burning can be monitored using the 3.9 um imagery. In this loop, covering one afternoon and evening, hundreds of fires can be seen as bright spots, covering large regions of northern Brazil. As described in earlier discussions on reflection at 3.9 um, water clouds are white and cirrus clouds are dark in this product; while thin cirrus around the edge of a thunderstorm anvil appear as filmy, slightly brighter regions. Also, in recalling the discussion regarding the difference in response at 3.9 versus 10.7 um for partially filled FOVs (the first “Fire Detection” display page), it can be seen that this product portrays fire areas as regions of enhanced “reflected energy” at 3.9 um, where the increased radiance is solely due to the fire’s contribution.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

The image above, from 10 September 1995, 1845-2115 UTC, reveals the magnitude of biomass burning. Note that the 3.9 um “reflected product” continues to show the fire areas at night, as would the normal 3.9 um imagery, after the smoke plumes are no longer trackable with VIS imagery.

The “reflected product” is also useful at night for detecting other sub-pixel related phenomena such as cities or a hurricane’s eye that is covered by thin cirrus.

In some cases atmospheric contaminates are able to be detected using GOES imagery and derived image products. At the time of the visible image, a cold front was moving across west TX, with the strong winds behind it raising an area of dust. In this VIS image, the blowing dust appears as a slightly more reflective (lighter) area due to the forward scattering of visible radiation; it is almost undetectable in the corresponding 3.9 um image. However, a derived product image (DPI), created at CIRA, clearly shows the area of blowing dust. This DPI is a Principal Component Image (PCI), which is basically the 3.9 um channel output minus that from the other longwave IR channels. The detectability of other visibility contaminates is also under investigation, including smoke, which appears to be dependent upon solar radiation and scattering angle, and volcanic ash, which may be detectable both day and night.

Those interested in learning more about the use of PCIs in highlighting various meteorological features are invited to read the following paper: Hillger, D.W., 1996; Meteorological features from principal component transformation of GOES-8/9 Imager and Sounder data; Eighth Conference on Satellite Meteorology and Oceanography, AMS, 28 Jan – 2 Feb, Atlanta, GA, 4-p.

For many years, the orientation and the extent of sun glint, as seen with visible imagery, has been used to locate regions of smooth seas and weak surface winds. For instance, if the surface winds are calm and the water surface is smooth, reflection is strong when the sun-satellite geometry is optimized. The high quality VIS imagery from the new GOES gives users the opportunity to enhance the imagery to reveal greater detail in the sun glint region. There is very strong reflection of solar radiation at 3.9 um because of a difference in the refractive index of water at that wavelength. This causes sun glint to be very bright in the 3.9 um imagery and, at low solar angles, the sensor (and thus the image) becomes saturated.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

These four-frame sequences of coincident VIS, above, and 3.9 um (displayed as reflectivity), below, imagery show sun glint detected by GOES-9 on 11 Sept 1995, when the satellite was located at 90 W. The glint region is at the right-center portion of the images and progresses to the left (west) until it is barely detectable in the 4th frame of the sequence. Note the extensive glint area in the 3.9 um images as compared with that seen in the visible images.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

At night, the 3.9 and the 10.7 um channels behave similarly, since only emitted radiation is present. At times, this allows for the detection of cumulus and cumulus bands; such as when differences between surface and cloud temperatures allow for the easy identification of the cloudy area. However, when land surface and cloud temperatures are nearly the same, discriminating between cloud and ground is extremely difficult with information from only one IR channel.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

This is illustrated in the 3.9 um imagery, above, from over the Great Lakes, during the night of 7 Dec 1995. As the loop starts, notice the distinct difference between cloud (cumulus bands), ground and surface water. As the evening progresses, and the ground cools, the distinction between cloud and ground becomes less evident. This is not the case when the fog/stratus product is used, and the cumulus bands (water cloud) are easily detected over, and downwind from, the Great Lakes. In the fog/stratus product loop below, cumulus bands are white and cirrus cloud is black.

Please download the latest version of Java Virtual Machine from http://www.java.com if you’re having trouble viewing this loop.

During the daylight hours, the potential for identifying cumulus clouds which have entered the ice phase is greatly improved when using 3.9 um imagery. This is because of the difference in reflection between ice and water cloud at 3.9 um. This image shows a large thunderstorm complex along the Louisiana Gulf coast from the VIS and 3.9 um channel imagers. In the VIS image, well-organized bands of cumulus over the Gulf of Mexico can be seen feeding into the storm area. When the same storm complex is viewed with the 3.9 um channel, many of the cumulus appear black while others nearby do not. In this particular lookup table, black is set to represent clouds with low reflectivity. It is likely, then, that the black cloud areas in this 3.9 um image have already entered the ice phase.

The detection and monitoring of volcanic ash clouds is particularly important for aviation safety since volcanic particulates can cause serious damage to aircraft engines. The volcanic ash cloud in this example is from an eruption on Mt. Ruapehu, in New Zealand, on June 17th of 1996. This ash cloud and subsequent eruptions caused the Wellington airport to cease operations for several days.

Three products from GOES are useful in monitoring the evolution of volcanic ash clouds. They are daytime VIS imagery (seen in the upper left at 0300 UTC) during daylight hours; the 3.9 um reflectivity product (upper right); and a new ash cloud product (lower right), a scaled temperature difference between the 10.7 and the 12.0 micrometer channels, made useful because of the absorption by sulphur dioxide in the 12 um band. Notice how well the volcanic ash plume can be seen in all three images at 0300 UTC. At night, 10.7 micrometer imagery (replacing the VIS imagery in the upper left at 0900 and 1200 UTC in this example) does not reveal the ash cloud, however it is detectable in both the reflectivity and the ash cloud products at 0900 UTC and, three hours later, at 1200 UTC. Note how the volcano’s hot lava region is detectable in the reflectivity product, since it is not obscured by ash cloud.

Images at 0300 UTC

Images at 0900 UTC

Images at 1200 UTC