Presented At:

Short Course On New Generation GOES Training (GOES-8/9)

AMS Annual Meeting, January, 28, 1996

THE ERA OF GOES-8 AND BEYOND

Interpretation of Images and Products

1. INTRODUCTION

The United States new generation of GOES satellites provides the capability for investigating oceanographic and meteorological phenomena that far exceed those available from previous generation GOES satellites. The reason for this is the satellite's improved performance in all five imager channels (Menzel and Purdom, 1994). The improved resolution and dynamic range of GOES-8's multispectral imager provide the opportunity to produce a number of advanced products that combine GOES and AVHRR imagery, opening a new era of advanced satellite image products. This paper will point out a number of such opportunities with respect to meteorological and oceanic phenomena. Detailed discussions are also available to short course participants by investigation of the COMET CD module "Satellite Meteorology:Remote Sensing using the New GOES Imager" that has been included as a part of the course materials. In addition there are a number of bulletin boards that users may wish to access over Internet that contain up to date examples of GOES-8 imagery, products and discussions of their use. Three of particular interest (CIRA, CIMSS, NASA/GSFC) are given in the reference section.

1.a Important system improvements

The new GOES system offers a significant advancement in geostationary observing capability because of:

a) earth oriented spacecraft allows for more efficient duty cycles of imager and sounder which yield higher spatial resolution and improved signal to noise;

b) improved multispectral imaging capability a five channel high resolution imager which allows for improvement in present services as well as the development of a number of advanced products;

c) improved sounding capability a dedicated 19 channel sounder with improved radiometric performance which allows for more frequent and accurate thermodynamic soundings over the USA;

d) separate sounding and imaging conflicts in scheduling between the imager and sounder which had previously existed are eliminated, allowing maximization of utilization for each sensor;

e) new ground data processing system allows for efficient dissemination of data and products to users from this improved satellite sensing system; and,

f) different data transmission formats to direct receive users allows direct receive users access to entire data set at reasonable costs, with growth potential.

1.b Significant imager advancements

In all five GOES-8 imager channels there is more detail in the imagery than with previous GOES. Table 1 compares GOES-7 and GOES-8 imager characteristics. With the new GOES, the visible and all infrared bands are over sampled. This over sampling was designed to take advantage of instrument response characteristics and scan rate in order to allow for maximum detail in the imagery. In the infrared channels there is higher resolution, with better signal to noise in most of the channels (GOES-9 performance exceeds that of GOES-8) and 10 versus 6 bit visible imagery. The visible imager uses silicon photodiodes instead of photomultiplier tubes: this provides a much more stable system, and visible imagery has virtually no striping, even with extreme enhancement tables. The infrared imagery is taken two lines at a time with different detectors for each line (with the exception of 6.7 µm which is still a single detector (note the difference in resolution). While using two detectors allows for improved resolution, 1/f noise and digitization uncertainty mean that there are times when adjacent infrared scan lines will exhibit striping.

1.c Results of imager and system improvements

Improvements in the new GOES system and imager channels is allowing for improved services and composite images.

* Having a separate imager and sounder and has allowed for more flexible scan modes which is allowing for better synchronization with other observations

* Enhanced severe storm forecasting and nowcasting is now possible using 5-channel Rapid Scan Imagery

* Best depiction ever of atmospheric changes with one-minute interval imaging

* 5 channels with much more detail allows for improved composite imagery and better multi-spectral assessment capability

* Best 6.7 µm (IR water vapor) imagery ever; an order of magnitude improvement enables identification of mesoscale disturbances embedded within synoptic scale features

* Better cloud drift winds 4 km resolution for better edge detection and improved target selection

* Improved water vapor winds in clear regions with 8 km spatial resolution and better Signal-to-Noise at 6.7 µm

* Fog, water and ice cloud detection both day and night using continuous 3.9 µm imagery with other channels

* Identification of super-cooled cloud

* Monitoring of snow and ice cover and the detection of cloud over snow

* Improved detection of forest fires and biomass burning

* Useful imagery well beyond the satellite's 60 degree zenith angle makes possible the detection and tracking sea ice and polar lows

* Improved low light imaging ability with 10 Bit visible

* Enhanced land and sea surface temperature monitoring capability using 30 minute interval multispectral IR capability

* Timely depiction of changes in atmospheric moisture and stability - better delineation of gradients

Table 1. Comparison of measured performance for GOES-7, 8 and 9

GOES-7 Characteristics :

Wavelength        IGFOV                 SSRes                   Noise 
   (µm)       (km [E/W x N/S])      (km [E/W x N/S])        
0.55-0.75      0.75 x 0.86           0.75 x 0.86        6 bit data + 2 counts (3sigma)
3.84-4.06      13.8 x 13.8            3.0 x 13.8         0.25 K @ 300 K, 6.0 K @ 230
6.40-7.08      13.8 x 13.8            3.0 x 13.8                  1.0 K @ 230 K
10.4-11.3       6.9 x 6.9             3.0 x 6.9          0.10 K @ 300 K, 0.2 K @ 230
12.5-12.8      13.8 x 13.8            3.0 x 13.8         0.40 K @ 300 K, 0.8 K @ 230 

GOES-8 Characteristics :

Wavelength        IGFOV                 SSRes                   Noise 
   (µm)       (km [E/W x N/S])      (km [E/W x N/S])        
0.52-0.72       1.0 x 1.0             0.57 x 1.0       10 bit data + 8.1 counts (3sigma)
3.78-4.03       4.0 x 4.0              2.3 x 4.0       0.16 K @ 300 K, 4.00 K @ 230 
6.47-7.02       8.0 x 8.0              2.3 x 8.0               0.27 K @ 230 K
10.2-11.2       4.0 x 4.0              2.3 x 4.0       0.12 K @ 300 K, 0.40 K @ 230 
11.5-12.5       4.0 x 4.0              2.3 x 4.0       0.20 K @ 300 K, 0.40 K @ 230 

GOES-9 Characteristics :

0.52-0.72       1.0 x 1.0             0.57 x 1.0         10 bit data + 9.3 counts (3sigma)
3.78-4.03       4.0 x 4.0              2.3 x 4.0         0.08 K @ 300 K, 2.00 K @ 230 
6.47-7.02       8.0 x 8.0              2.3 x 8.0                 0.15 K @ 230 K
10.2-11.2       4.0 x 4.0              2.3 x 4.0         0.07 K @ 300 K, 0.15 K @ 230 
11.5-12.5       4.0 x 4.0              2.3 x 4.0         0.14 K @ 300 K, 0.28 K @ 230 

2. SOME SIGNIFICANT MILESTONES IN THE GEOSTATIONARY PROGRAM

ATS-1 was launched December 6, 1966: a stellar day in satellite meteorology

a) ATS-1's spin scan cloud camera (Suomi and Parent) provided full disk visible images of the earth and its cloud cover every 20 minutes. Weather systems in motion during daytime.

b) By the early 1970's the first movie loops being used at the National Severe Storm Forecast Center in the spring of 1972.

SMS-GOES (SMS-1 on 17 May 74) - (GOES-1 on 16 October 1975)

a) geosynchronous satellites dedicated to meteorology

b) WEFAX, the transmission of low resolution satellite images and conventional weather maps to users with low cost receiving stations;

b) DCS, which allows for the relay of data from remote data collection platforms through the satellite to a central processing facility; and,

c) VISSR, routine imaging of the earth and its cloud cover in the visible and infrared window channels.

d) GOES data became a critical part of National Weather Service operations by providing imagery to national centers (through direct receipt) and local weather service forecast offices (through GOES-TAP services).

GOES-4 launched on 9 September 1980

a) added Atmospheric Sounder to the VISSR

b) addition of channels represented a major improvement in satellite capabilities; however, imaging and sounding could not be done at the same time

The GOES-I/M system was introduced with the launch of GOES-8 on April 13, 1994. See section 1 for a brief list of improvements.

3. A QUICK LOOK AT IMAGERY USES

3a. Some significant uses, pre-1994

From its inception, the importance of cloud imagery to the meteorological satellite program has been recognized. Since TIROS-1, significant strides forward in synoptic scale weather interpretation with routine global cloud observations from polar orbiting satellites. From the earliest days of the geostationary program, cloud drift winds derived from geostationary satellite imagery were (and are today) used by operational forecast centers to help with analyses in data sparse areas.

Geostationary satellite imagery has helped advance our understanding of the mesoscale. Prior to the geostationary satellite the mesoscale was a "data sparse" region; meteorologists were forced to make inferences about mesoscale phenomena from macroscale observations. The clouds and cloud patterns in a satellite imager should be thought of as a visualization of mesoscale meteorological processes. For example, animation provides observations of convective behavior at temporal and spatial resolutions compatible with the scale of the mechanisms responsible for triggering deep and intense convective storms.

Geostationary satellite imagery has had a dramatic impact on mesoscale meteorology and in turn short term forecasts and warnings.

1) Using time sequences of high resolution geostationary satellite imagery to locate areas of incipient squall line development has aided in both the orientation and timing of severe weather watches.

2) Thunderstorm outflow boundaries (arc cloud lines), and their importance in the development and evolution of all types of thunderstorms was first recognized using animated satellite imagery. Doppler radar has confirmed the importance of recognizing and tracking arc cloud lines for short term convective forecasting.

3) Mesoscale convective complexes, their size, duration and high degree of organization were not recognized prior to their discovery using infrared geostationary satellite imagery.

4) The location, tracking and monitoring of hurricanes and tropical storms has been one of the most successful aspects of the satellite program; and, the ability to continuously monitor them with geostationary satellite imagery is exceptionally valuable to forecasters.

5) Polar Lows were not recognized as a distinct weather type prior to the polar orbiting satellite era; and they have recently begun to be studied using geostationary satellite imagery.

3b. Some significant uses, post-1994

The GOESI/M system, one of the major components of NOAA's Modernization Program, offers significant advancements in geostationary environmental satellite capabilities. Aside from enhanced ability to monitor and study phenomena pointed to in section 3a;

visible

a) improved cloud edge and cloud top feature detection for improved cloud drift winds and severe storm identification;

b) low light visible imagery;

c) better detection of pollution and haze;

d) highly accurate cloud height measurements (stereo and shadows); and,

e) use of imagery well beyond the satellite's 60 degree zenith angle, including the ability to monitor and track sea ice.

shortwave infrared window (3.78-4.03 µm)

a) identifying fog (water cloud) and cirrus at night; b) distinguishing between cloud cover and snow during the daytime;

c) delineating supercooled cloud from and ice cloud (with longwave IR);

d) detecting hot areas such as forest fires;

e) improvement of sea surface temperature measurements at night.

water vapor channel (6.47-7.02 µm)

a) better winds in cloud free areas;

b) improved analysis of synoptic scale features;

c) ability to identify mesoscale features embedded within synoptic disturbances.

longwave infrared window (10.2-11.2 µm)

a) better cloud edge and cloud top feature detection for improved cloud drift wind velocities; b) enhanced thunderstorm development and evolution monitoring;

c) severe storm identification and location of storms with heavy rainfall;

d) improved monitoring of land surface heating and cooling;

e) good low level moisture identification with the 11.5-12.5 µm channel;

f) better nighttime SST determination with 11.5 µm and 3.9 µm channels.

split window (11.5-12.5 µm)

a) development of much more accurate low level moisture products;

b) recognizing areas with a potential for radiation fog development;

c) improved sea surface temperature products.

4. A DETAILED LOOK AT THE VARIOUS SPECTRAL BANDS AND SOME APPLICATIONS

In this section characteristics of the various spectral bands and how they relate to image quality and products will be discussed. Selected examples will be shown, mainly for visible and 3.9 micron imagery. Once again, you are reminded that the COMET CD contains a plethora of examples and allows comparisons to be made between channels: it is highly recommended that you spend some time working with that CD.

4a. Visible (0.52-0.72 µm)

Why was this particular spectral region chosen, and what are those consequence? This can perhaps be understood best by inspecting the spectral region sensed by the GOES-8/9 visible sensors, as well as comparing GOES-8 visible imagery with AVHRR Channels 1 and 2. Figure 1 shows the spectral response of the GOES-8 and GOES-9 imagers in the "0.52-0.72" range. As can be seen from that figure, a substantial amount of energy comes from beyond 0.72 µm.

Figure 1. GOES-8 and GOES-9 visible channel spectral responses.

Why? The instruments performance was defined by response at half power points, and in those terms the spectral response is fine. However, roll off is gradual on the longer wavelength side (this is being addressed) and as might be expected this effects how the imagery is interpreted.

Figure 2. Spectral albedo of different natural underlying surfaces at various solar elevations (E). 1 = snow with ice crust, E = 38 ; 2 = large-grained wet snow, E= 37 ; 3 = water surface of a lake, E= 54 ; 6 = tall greencorn, E= 56 ; 7 = yellow corn, E= 46 ; 8 = sudan grass, E=52 ; 9=chernozem, E= 40 ; stubble of cereals, E= 35 . Adapted from Kondratyev (1973).

Figure 2, shows the spectral albedo of various types of vegetation as a function of wavelength. Notice how most vegetation becomes much more reflective at wavelengths longer than about 0.7 um. This means that the GOES-8/9 visible imagery will be slightly brighter over vegetated surfaces than might have been expected otherwise. This can be seen by comparing a GOES-8 image with a similar image from AVHRR Channels 1 (0.55-0.68 µm) and 2 (0.75-1.10 µm), as is done in Figure 3. The two AVHRR channels are the ones used to produce the Normalized Derived Vegetation Index (NDVI), as one might surmise from looking at their spectral characteristics and Figure 2. Figure 3 compares AVHRR bands at 0.55-0.68 µm and 0.75-1.10 µm and the GOES band at 0.52-0.80 µm (all remapped to a common projection). Most apparent is the difference in land surface brightness as wavelength increases; this is due differences in reflection due to vegetation (brighter past 0.7 µm). Equally as striking is the difference in water brightness in the northern portion of the Chesapeake Bay and western Potomac River due to turbidity: apparent in the AVHRR Channel 1 band at 0.55-0.68 µm and the GOES visible band at 0.52-0.80 µm, but not seen in AVHRR Channel 2 band at 0.75-1.10 µm band because of the strong absorption by water at longer wavelengths. With previous generation GOES satellites, an increase in water brightness along the west coast of Florida was routinely observed after the passage of cold fronts and storms with strong winds, and similar increases in turbidity have been observed in delta regions during periods of strong river flow (Huh et al, 1995).

Figure 3a. GOES-8 visible image over the East Coast of the USA on January 27, 1995. The image was taken at 1815 GMT. This image should be compared with the NOAA-14 AVHRR images shown in Figures 3b and 3c.

Figure 3b. NOAA-14 AVHRR Channel 1 image taken within a few minutes of the GOES image shown in Figure 3a, with a viewing angle of the area very close to that of GOES-8.

Figure 3c. NOAA-14 AVHRR Channel 2 image, taken at the same time as Figure 3b.

The high quality of the new GOES visible imagery makes possible the investigation of phenomena beyond the satellite's 60 degree local zenith angle, a region that has previously been considered non-useful for geostationary viewing. Figure 4a shows the region outside a satellite's 60 degree zenith angle for a geostationary satellite located at 90o W. Figure 4b covers most of east and central Hudson Bay and extends northward from about 55.5N to within a few tenths of a degree south of the Arctic Circle. In terms of satellite zenith angle that means between about 63 and 75 degrees. This is a very interesting image:

Figure 4a. 60 degree zenith arc for a geostationary satellite at 90 West.

a) much of Hudson Bay is snow and ice covered (addressed in 3.9 µm section), although Coats Island has water around part of it, but with ice in much of the Fisher Strait between Coats and Southampton Islands; b) notice the structure in the ice over Hudson Bay, as well as the icebergs in the Fisher Strait; c) another island, Mansel, is apparent near the right edge of the image (notice the relatively large ice free region to its west); d) the water region appears filmy around Mansel

Figure 4b [not available]. GOES-8 one km resolution visible image taken on 31 May 1994 at 1300 GMT when GOES-8 was at 90o West. The image extends from about 55o N to near the Arctic Circle.

Island, indeed, the filmy appearance extends over the ice and snow on Hudson Bay, with its western edge appearing as a darker curving region that just intersects the southern end of Coats Island - use of visible, 3.9 (see Figure 17) and 10.7 µm channel imagery revealed that the "hazy" region was thin super cooled cloud. The realization that useful information is obtainable as far from subpoint as the 74 degree satellite zenith angle is very exciting. The ability to observe detail to near the arctic circle provides the capability for observing polar lows, icebergs, cloud drift winds and a plethora of phenomena that far exceed those available from earlier generation geostationary satellites. Indeed, with GOES-West, much of Alaska, the Aleutian Islands and Bering Sea which lie beyond the satellite 60 degree zenith angle will now be well covered with geostationary satellite imagery.

In Figure 4b, it is also instructive to examine the detail that can be seen in the snow and ice fields in the Hudson Bay area. For many years the location and extent of snow cover has been located using visible imagery by identification of landmarks such as rivers, lakes, cities and coniferous forest regions within the snow field. With the high quality imagery available with the new GOES visible channel, this task becomes easier because of a users ability to enhance that imagery and have the resulting image remain near noise free. However, as nice as the visible imagery is for this task (and it should be used because of its high resolution), users should employ the multi-spectral capability of the new GOES and combine information from the visible channel with 3.9 µm imagery. This is done in section 4.c.1d.

Figure 5 is an example of how well sea ice areas can be detected and followed with the new GOES satellite's visible imagery. Normally monitoring ocean ice with satellite imagery is done using AVHRR data from sequential polar orbiting satellite passes. While this remains an excellent option because of the high resolution nature of the AVHRR data, problems have been noted when the ice area is covered with cloud. Animation dramatically illustrate the capability of the new generation of GOES satellites to follow the movement and changes in sea ice. This has been observed near Nova Scotia and in the Gulf of St. Lawrence, Figure 5. Especially interesting (in animated half-hourly imagery) is the movement of a large ice area in the Atlantic near Cape Breton Island, and the distortion and breaking away of another large ice area on the second and third day along Cape Breton Island's N.W. coast. This ability, to monitor changes in ice fields and icebergs should prove exceptionally exciting to oceanographers.

Figure 5a. GOES-8 visible image showing ice area in Gulf of St. Lawrence and south of Cape Breton Island. 1245 GMT on March 30, 1995 .

Figure 5b. Close-up of ice south and north of Cape Breton Island in Figure 5a.

Figure 5c. Change in ice around Cape Breton Island, March 31, 1995, at 1715 GMT.

The United States new GOES satellites provide the capability for investigating meteorological phenomena that far exceed those available from the previous generation. What continues to be a fascinating area of investigation is what can be seen in, and derived from, animated imagery at very frequent intervals (one to three minute interval imaging). Both GOES-8 and GOES-9 have provided research scan imagery during their checkout phase and covered a tremendous variety of phenomena, including hurricanes and severe thunderstorms. While long periods of research scan imagery from GOES-8 was fairly limited, due to its being the first in a series, one minute interval research scans for GOES-9 have approached 24 continuous hours; and, during the month of November, 1995, GOES-9 provided near continuous 3 minute interval imaging over the continental United States. Note, that there are never 60 one minute scans (or 20 three minute scans) in a one hour period due to house-keeping requirements, such as blackbody observation for calibration and star searches for navigation.

In papers such as this, it is only possible to show still imagery or results of various measurements made from that imagery. However, the COMET CD has examples of one minute interval imagery, and electronic means now exist that allow people to access Internet and the World Wide Web to view one minute imagery in animation. Two home pages on the world wide web, known to the author, where one minute imagery may be viewed are: a) RAMM (1994) INTRODUCTION TO GOES-8; and, b) on the GOES homepage at NASA/GSFC (Chesters, 1994).

For the first time ever, on July 20, 1994, one minute interval imagery was taken by a geostationary satellite: from 2201-2219 GMT and 2031-2249 GMT. The one minute interval sector covered from eastern New Mexico to eastern Oklahoma, and from southern Texas to northern Kansas. During the one minute interval imaging period a variety of interesting phenomena were observed. A thunderstorm just north of Tulsa, Oklahoma, under went explosive growth at cloud top, Figure 6. That storm produced large hail and winds in excess of 70 mph a short time after the explosive growth at storm top. One minute interval imagery for the period 2237-2249 GMT, which shows how rapidly changes may occur at storm top and the generation of waves within the anvil cirrus, as well as how easily cumulus and cirrus may be followed at one minute intervals, may be viewed on the INTRODUCTION TO GOES-8 tutorial mentioned above. In that same historic scan period (not shown) in southern Colorado, low level wave clouds were observed in cloudiness to the north of a frontal boundary. Inspection of imagery at 3.9 µm's confirmed that those clouds were composed of water (sect 4c). Towering cumulus and thunderstorms were observed in west Texas and eastern New Mexico. As the towering cumulus grew, some were observed in the 3.9 µm imagery to change from brightly reflective to dark, an indication of phase change at cloud top from water to ice.

Figure 6 (not available). GOES-8 visible imagery at 0.57x1.0 km resolution taken between 2219 GMT and 2233 GMT of a storm near Tulsa, Oklahoma

On October 13, 1994, Hurricane Rosa was observed at one minute intervals while it was south of Baja. The one minute interval loop of hurricane Rosa dramatically illustrated the ability to analyze detailed cloud motions using one minute interval GOES 8 imagery. Knowledge of detailed wind fields in and around hurricanes is important in determining their intensity and the distribution of gale force and higher winds. Rosa's maximum low level winds were estimated to be 75 knots, and tracking of low level cumulus near the storm's center using one-minute interval GOES-8 images found cloud motions of 70 knots. The long lifetime and abundance of cirrus allowed for detailed mapping of the upper level winds associated with hurricane Rosa, Figure 7. Obvious is a strong outflow jet over Baja, to the northwest of the storm. The Rosa winds were derived using full resolution visible imagery. Notice how the outflow winds vary greatly depending on which quadrant of the storm is inspected.

Figure 7 (not available). Manually derived cirrus winds for hurricane Rosa, 13 October 1994.

A plethora of interesting phenomena are currently under investigation using one and three minute interval imagery. During the 1995 tornado season GOES-8 one minute interval imagery was taken in support of a scientific field program known as VORTEX: those data are under active investigation in conjunction with WSR-88D data (NEXRAD Doppler Radar). Numerous cases of seabreeze evolution and general convective development have been observed by both GOES-8 and GOES-9; thunderstorm outflow boundaries are easily detected and their role in the generation of new convection reaffirms earlier satellite observations of that phenomena. One particular case, May 31, 1995, points to the role of thunderstorm outflow in the development of a tornadic storm in southwest Texas. The 1995 tropical season was much more active than 1994, and a number of hurricanes have been observed at one minute intervals by GOES-8, including landfalls of Allison, Erin, and Opal, as well as Felix as it sat near stationary off the Atlantic sea-board. From 3-17 September, GOES-9 was in its special atmospheric science testing mode, and a variety of phenomena were observed at one minute intervals, including hurricanes Luis and Marilyn for most of their life times. As a part of that experiment, NESDIS obtained one minute interval stereo observations from GOES-8 and GOES-9, and although the two satellites were only separated by 15 degrees, the stereo shift was dramatic - giving hopes for highly accurate clouds height assignment once GOES-8 and GOES-9 are on permanent station and their imager scans are synchronized.

Past research using ground based stereo cameras, 5 minute interval visible imagery, and manual cloud tracking techniques showed that cirrus could be tracked using satellite imagery with an error of a few tenths of a meter per second, and heights could be determined within a few hundred meters using cloud shadows, Figure 8.

Figure 8. Cloud height accuracy by shadows as a function of time of day over Chicago, for a satellite located at 112 W.

Other research comparing rapid scan cirrus winds with ground based profilers, and using shadows for height, showed that the winds from the two systems matched to within a meter per second. The previous studies were with GOES-7 imagery.

Winds science improvement activities with one to five minute interval GOES-8/9 imagery have shown (for manual tracking):

a) Target selection problem virtually disappears

b) With visible imagery can distinguish shear between multiple layers

c) Storm relative motion makes tracking much easier, especially in multi-layered complex situations where automated methods fail

d) Shadows are crisp and distinct for height assignment

There is certainly great promise for highly accurate cloud drift winds, using GOES-8 and GOES-9 imagery - especially for cirrus where 15 minute interval viewing appears adequate for accurate target identification. There remain some questions to be addressed in the short term:

a) How do we combine rapid scan GOES imagery with WSR-88D for wind determination?

b) What imaging frequencies are needed to track different types of clouds without ambiguity - over land and over water?

c) How accurately do cumulus motions reflect low level wind?

d) What is the Northern extent for winds with GOES-8/9?

e) Can measure of anvil motion (possible blocking) and gravity wave generation at storm top, be used to predict deviations in storm motion and severity?

Two areas mentioned earlier deserve a short discussion. They are: a) low light visible imagery; and, b) better detection of pollution and haze. This discussion will be very short since both topics are covered well in the COMET CD. However, it is worthy of note that with the 10 bit visible imagery having such good detector performance, and since the new detectors can be well matched (inter-calibrated) that the imagery can be greatly enhanced without striping. This allows for detection of subtle features that would go unnoticed in unenhanced imagery. This means we need to learn how to operate in a new realm - enhanced visible imagery. The opportunity to explore sea surface roughness in regions of sun glint by enhancing visible and 3.9 µm imagery is discussed in section 4.c.

4b. The infrared bands

As was mentioned earlier, the new GOES satellite imagers take data in four spectral bands that lie beyond visible wavelengths. Each band has special attributes that will be discussed in individual sections that follow. However, one important consideration will be discussed in this section: just what is does each band detect? This is not as straight forward as it may first seem, and relies on a number of factors. They include such items as diffraction and detector noise, both instrument dependent, as well as basic physics. Noise will be discussed with each channel, as will atmospheric characteristics that influence what is detected by the sensor. Discussed next are diffraction and response to sub-pixel temperature variations; it is extremely important that these influences be recognized before proceeding further. Table 2 below provides a comparison of GOES-7 and GOES-8 instruments parameters pertinent to this question. The mirror diameter of GOES-7 was considerably larger than for GOES-8. As can be seen by the size of the airy disk compared to the detector size, diffraction should have a significant effect on GOES-8 10.7 um and 12 um images.

The values in Table 2 were supplied courtesy of Dr. Dan Cousins of MIT Lincoln Labs. He also provided the information on MTF, concluding with his estimates of encircled energy within one IGFOV for GOES-8.

Table 2. Comparison of selected optical components for GOES-7 versus GOES-8.

GOES-7 VAS :

0.406m diameter f/7.2 telescope with 16% obscuration

f/1 IR relay optics, effective focal length = 0.406m

HCT detector size = 78 µm

IFOV = 6.9 km = 192 µrad

Airy disk at 12 µm, B= 30 µm

GOES-8 Imager :

0.311m diameter f/12.2 telescope with 6% obscuration

f/1.7 IR relay optics, effective focal length = 0.518m

HCT detector size = 55 µm

IFOV = 4.0 km = 112 µrad

Airy disk at 12 µm, B= 48 µm

MTF is calculated for GOES-8 to keep track of image quality and is determined from the combined effect of (detector area) x (diffraction) x (optical aberrations) x (scan mirror flatness) x (electronic sampling) x (thermal distortion). For GOES-8 12 µm channels at the Nyquist frequency of 4490 cycle/radian, predicted MTF = (0.63)(0.70)(0.99)(0.99)(0.8)(0.95) = 0.32. This is dominated by detector size effects

To be consistent with GOES-8 Imager MTF performance, Dr. Cousin's estimated the encircled energy fractions within 1 IGFOV:

ch2, 3.9 µm 4km IGFOV: 85%

ch3, 6.7 µm 8km IGFOV: 85%

ch4, 10.7 µm 4km IGFOV: 68%

ch5, 12.0 µm 4km IGFOV: 65%

This means that when looking at the same scene at 3.9 µm or 6.7 µm, versus 10.7 µm or 12 µm that they are seeing different amounts of energy from the same IGFOVs.

It is interesting to note that for the GOES-VAS system that encircled energy amounts were on the order of 80% for an IGFOV of 6.9 km at 10.7 µm, and about 90% for 2 IGFOVs.

The Planck function is given by,

B( L,T) = c1/( (L**5)exp(c2/ L T) - 1),

where B is the Planck radiance for a given wavelength (L) and temperature (T), and c1, c2 are constants, is not linear. When one takes the derivative of B with respect to T (this is most easily done by substitution), the result is

dB/dT =(c1 B)/ (L T**2) , or dB/B = (c1/ L T) (dT/T).

To get a sense of variation of B with respect to T for different wavelengths, we can make the assumption that the temperature region of interest is near constant, or c1/ L T = a/L, where a is a constant. This approximation gives dB/B = (a /L ) (dT/T), which has as a solution

B = T**(a/L) .

What this means is that for various wavelengths the increase of radiance with respect to increase in scene temperature varies basically as T raised to some power, which is a constant divided by . Thus response to increases in scene temperature are much larger at shorter wavelengths. Now of what practical importance could this possible be? Enormous! While the various infrared channels measure radiance from the "same" scene, which is eventually converted to temperature, they respond differently to a scene temperature unless that temperature is absolutely uniform across the scene (we're ignoring emissivity differences as a function of wavelength for the moment). Should we expect the same temperatures for a scene on a clear and perfectly dry night over land areas at 3.9, 10.7 and 12.5 µm? Absolutely not, unless the temperature across that region were the same, not for just one IGFOV, but for several. Furthermore, since there is inherent variability in land surface temperatures, especially in and around cities and their suburbs, we should expect to measure hotter surface temperatures (at night, we'll see why later) at 3.9 µm vs 10.7 µm because of stronger response at 3.9 µm to the warm areas as well as more energy coming from within one IGFOV than at 10.7 µm. This is the basic reason why it is often said that "3.9 likes it hot, 10.7 likes it cold." This is shown schematically in the graph in the section below where it is shown how the 3.9 micron channel may be used to detect fires, which while filling only part of an IGFOV are very hot in that fractional IGFOV area.

4c. 3.9 microns

Radiance received by the GOES imager channel at 3.9 µm's is from reflected solar and emitted earth/cloud during daytime, and from emitted earth -cloud at night since there is no solar component. This is shown graphically in Figure 9. This fact, coupled with the information provided in the previous section, allows for the determination of a variety of interesting and important features. However, before going into discussions of those features several important areas will be addressed: first, the effect of instrument measurement uncertainty (noise) on interpretation of 3.9 µm imagery and the development of multispectral products; second, reflection from ice and water clouds; third, response to hot areas within an IGFOV; and, fourth, cloud and surface emissivity; and, finally, how should imagery from this channel be viewed? Figure 10 shows GOES-8's response at 3.9 µm's for various scene temperatures and the inherent accuracy of those measurements.

Figure 9. A zoomed in view of the region where the Planck curves cross for a 6000 K source (sun) and 300 K source (earth).

Figure 10. Plot of radiance versus temperature at 3.9 µm's. Note the effect of instrument accuracy (0.0086) on temperature measurement accuracy.

In Figure 10, notice how Planck radiance increases rapidly with increasing temperature. Also notice how "flat" the curve becomes at cold temperatures. Since the inherent measurement accuracy of the GOES-8 instrument at 3.9 µm's is constant, the result is a much less accurate measurement at cold temperatures versus warm temperatures. This graph shows that the 3.9 µm imagery is not useful for analyses where information that relies on cold temperatures is necessary, such as thunderstorm top analyses; however, for measurements of sea surface temperature the 3.9 µm channel should do a fine job. When certain products, such as the "fog" product are being made, and the cloud top temperature of the fog or stratus cloud is very cold (say -10 C), the result will be a noisier product than when the temperatures are warmer.

Reflection at 3.9 µm is sensitive to phase, and very sensitive to particle size. This is shown graphically in Figure 11. Notice how water droplets are always more reflective for ice of the same size. Furthermore, the droplet spectra in water cloud normally ranges between droplet diameter of 5 to 15-20 µm, depending on cloud type, see tables 3, while ice is normally an order of magnitude larger, see table 4. Next, the reflection that is seen from a satellite is multiple scattering (the graph in Figure 11 is single scatter), thus as multiple reflections occur the numbers keep multiplying. The result is that clouds with small water droplets, such as cumulus, fog and stratus over land are much brighter when viewed at 3.9 µm than ice clouds, which are very poorly reflective and dark.

Figure 11. Single scattering albedo for ice and water droplets at various sizes.

TABLE 3. Characteristics of cloud-droplet populations

Cloud type                               rmin(µ) r50  rd  rm  rmax(µ) n(cm-3) w(g/m3)
Small continental cumulus (Australia)*     2.5    6    -   -    10      420     0·40
Small continental cumulus (England)*        -     -    4   6    30      210     0·45
Small trade-wind cumulus (Hawaii)*         2.5    10  11  15    25       75     0·50
Cumulus congestus                           3     -    6  10    50      100     1·0
Cumulonimbus                                2     -    6  20   100      100     2·0
Orographic cloud (Hawaii)                   5    13    -   -    35       45     0·30
Stratus (Hawaii)                           2.5   13    -   -    45       24     0·35
Stratocumulus (Germany)                     1     -   3·5  4    12      350      -
* Not more than 7000 ft deep.

TABLE 4. Weickmann's observations of predominant crystal forms in different cloud types

Level of	Temperature       Cloud types	   Crystal forms        Crystal sizes 
observation	   range                                                  (approx.)
Lower            0 to -15C        Nimbostratus,    Thin hexagonal       50µ to ½ mm dia.,
troposphere                       stratocumulus,   plates               10 to 20µm thickness
                                  stratus						
   Star-shaped crystals  0.5-5 mm diameter						
   showing dendritic
                                                   structure

Middle          -15 to -30C      Altostratus,      Thick hexagonal       200µm diameter
troposphere                      altocumulus       plates

                                                   Prismatic columns,    200µm length
                                                   single prisms and
                                                   twins

Upper              < -30C        Isolated cirrus   Clusters of prismatic  1 mm. diameter
troposphere                                        columns containing
                                                   funnel-shaped
                                                   cavities

                                                   Single hollow          0.5 mm diameter
                                                   prisms

                                 Cirrostratus      Single complete        100µm length,
                                                   prisms                 length to diameter 
                                                                           = 1-5

Certain products and interpretations of 3.9 µm imagery rely on the difference in response to cold and warm partially filled fields of view with respect to the 10.7 µm channel. Consider a field of view that is filled with a substance (this may be fire or cloud for example) of quantity N. Since the field of view is 100% of an area, the non-substance part takes up an area equivalent to 1-N. Ignoring the effects of differences in emissivity for the moment,

the radiance measured at any wavelength can be expressed as:

R = R(1-N)_T1 + R(N)_T2,

where the radiance of the non-substance and substance areas are denoted by _T1 and _T2, respectively, corresponding to their temperatures (T1, T2).

As discussed earlier, at short wavelengths increase in radiance with temperature is much stronger than at long wavelengths. At 3.9 µm radiance increases much more strongly than at 10.7 µm, approximately at T**(a/4) versus T**(a /11). Thus the effect of partially filled fields of view where N is hotter than the surrounding background is to read a much warmer temperature at 3.9 µm than at 10.7 µm, as is shown in Figure 12. The opposite is true for partially cloudy scenes, as is shown in Figure 13. This type of information, difference in measured scene temperature at 3.9 µm and 10.7 µm, makes derivation of information such as fractional area covered by cloud or fire possible - if a number of assumptions are met.

Figure 12. Fractional coverage of a hot area at 500 K within a 300 K background scene versus satellite measured temperature for 3.9 µm and 10.7 µm.

The fourth consideration is emissivity. The satellite sensors measure radiance, and through inversion of the Planck equation for a given wavelength a blackbody temperature is determined. However, what is measured is not actually Planck radiance for that wavelength, but rather a radiance that has been reduced by some amount due to the emissivity of the radiating surface (no surfaces radiate as a blackbody, although in some wavelength regions they approach blackbody emission). Thus, the radiance measured is actually the Planck function for that wavelength that has been reduced some amount because of the emissivity of the radiating surface. It is this radiance that is inverted to give scene temperature, and if an adjustment for emissivity is not made the radiance derived temperature will be too low (that is, it will not be adjusted upward by some amount due to the emissivity correction).

Figure 13. As for Figure 12, but with a cloud temperature of 260 K.

While this is something to be aware of, and at first glance may cause great alarm, in many cases it actually works in our favor since the emissivity for clouds and certain surface types is different depending on the spectral band under investigation, this allows for the derivation of some very valuable products. Figure 14 gives emissivity for some different surfaces.

Figure 14. Infrared emissivity versus wavelength for certain surface types.

Finally, the question of how this channel should be viewed comes into play: "should it be displayed as visible or infrared?" A short step back into history is appropriate. With the early TIROS, visible imagery from vidicons showed clouds as bright white and ground as dark, a direct relationship between scene energy and picture gray scale. As the first infrared imagery began being received a problem for interpretation was immediately apparent: the direct relationship between scene energy and display gray shade was reversed! When visible lookup tables were used to display infrared energy, high energy areas (ground and ocean) were white and low energy areas (cirrus and thunderstorm tops) were dark. Backwards from the way people were used to looking at ground and cloud. A simple fix was made - invert the IR display table so that low energy was displayed as white and high energy as dark. This has served us well for years, but now, since the 3.9 µm channel has both reflected and emitted radiation during the daytime, a choice must be made concerning how imagery from that channel will be displayed (perhaps in time we'll determine that this imagery is best viewed as a product in combination with other channels, perhaps). At CIRA, we choose to display this channel in terms of energy versus gray scale (visible). Thus, day or night: a) cold clouds, ice, ice clouds and snow appear as dark; b) warm surfaces, water clouds and sun glint appear as light to bright (sun glint at 3.9 µm is much more intense than at visible wavelengths); c) land surfaces can be both hot and reflective and may appear as very bright. This is purely a matter of choice, and others prefer displaying the channel as infrared - regardless, a user needs to think in terms of energy and cloud/surface type or confusion may rapidly occur.

With the above thoughts in mind, let's look at some 3.9 µm imagery and derived products. Note that certain of these are only in the test mode at the time of this writing.

4.c.1 Ice versus water cloud

While emissivity differences in ice cloud are close to the same for 3.9 and 10.7 µm, for water cloud the emissivity at 3.9 µm is less than at 10.7 µm. This difference in emissivity allows for the location of water cloud at night, by simply subtracting the 10.7 µm image temperature from the 3.9 µm image and scaling the temperature differences. In regions of cirrus, where the cloud is thick, the resulting difference image is very noisy (refer to Figure 10). In areas of thin cirrus, since it is often patchy and since 3.9 µm is sensitive to warmer temperatures, that area will yield a difference of the opposite sign as for the water cloud area, helping distinguish between the two (also, thin cirrus often exists around the edges of thicker cirrus, an additional factor making both thick and thin cirrus detection easier). During daytime, the difference in reflection between ice and water cloud makes it fairly simple to distinguish between thin cirrus and water cloud: cirrus is both cold and non-reflective, making it dark in a reflectivity image, while water cloud will be both warmer and reflective making it brighter. However, this may not be true for thin cirrus, see section 4.c.6. During daytime ice and snow covered surfaces are obviously dark.

4.c.1a Fog and stratus at night

This is one of the new and exciting products available from the new GOES. It is made by subtracting the 10.7 and 3.9 micron image temperatures at night. This multi-channel product is made possible partly because of the improved resolution and cleaner signal detected by the new GOES. The technique has been in use for over a decade with AVHRR (Rao, 1985). AVHRR's high resolution infrequent views when coupled with GOES-8&9 frequent updates point to a potential strong partnership. With the new GOES satellite's imagery available every 15 to 30 minutes, the evolution of nighttime fog and low level stratus clouds is observable with this multichannel image product (this is nicely shown in the COMET CD). Users should be aware of two caveats when using what is now popularly called the "fog product." First, this product detects water cloud, so bands of cumulus downwind from regions such as the Great Lakes at night will have the same appearance as fog and stratus; obviously this is true for other water cloud. Second, differences in surface emissivity will yield a weak "fog" return over certain areas, such as the Cap Rock Escarpment in West Texas. Figure 15a, from the night of March 14 and early morning of the 15th, shows coastal fog and stratus spread southward over the Atlantic Ocean, and inland to cover parts of Delaware, Maryland, Virginia and North Carolina, as well as the Delaware and Chesapeake Bays. In this image, the fog and stratus areas are white, while any areas with cirrus are very dark, ground shows up as a medium gray tone. In this case, visibility as low as one tenth of a mile had a pronounced effects on ground, air and sea transportation. The visible image in Figure 15b confirms the location of the extensive area of fog and stratus.

Figure 15a. GOES-8 nighttime fog product for March 14 and 15, 1995, where areas of fog and stratus are shown as white. From upper left to lower right, times are 18:45, 22:15, 02:32 and 0:615 EST.

Figure 15b. Visible image from the early morning of March 15, 1995, confirming the location of fog and stratus in the lower right panel of Figure 15a.

4.c.1b Other water cloud at night

As mentioned previously, the "fog" product actually is useful for locating any water cloud at night, this is shown in Figure 16, where cumulus cloud bands over the Great Lakes are evident. Notice that while the cumulus bands are white, cirrus cloud (lower left) is black. When this nighttime product is used in conjunction with the 10.7 µm channel, cloud top temperatures may be used to identify regions suspected of being super cooled cloud. This is done by locating the water cloud area and then using the 10.7 µm channel (along with an appropriate representative rawinsonde or model sounding) to see if cloud top temperatures and atmospheric vertical structure are in the appropriate boundaries for super cooled cloud.

Figure 16. "Fog" product image at night showing water cloud bands over the Great Lakes during the night of December 7, 1995.

4.c.1c Super-cooled cloud during daytime

As mentioned above, because of differences in particle size and phase, it is relatively simple to distinguish between ice and water cloud during daytime using 3.9 µm imagery. Use of visible, Figure 4, 3.9 µm, Figure 17, and 10.7 µm channel imagery revealed that the "hazy" region is Figure 4 is thin super cooled fog and stratus. The 3.9 µm image in Figure 17 is displayed as reflectivity: a) cold clouds, ice, ice clouds and snow appear as dark; b) warm surfaces and water clouds appear as light to bright; c) land surfaces can be both hot and reflective and may appear as very bright.

Figure 17. (not available) 3.9 µm image taken at the same time as the visible image in Figure 4b. See text for discussion

Comparing Figure 4 and Figure 17 reveals that the filmy region (visible) over Hudson Bay appears as bright (3.9 µm) while the surrounding snow and ice is darker (3.9 µm) - this is a thin super-cooled water cloud with small droplets which makes it so reflective (10.7 µm imagery showed the cloud region to be slightly colder than the surrounding snow and ice). The black region (3.9 µm) to the south that extends from west to east is cirrus. The great lakes are dark, while the surrounding ground is brighter - this is because the ground is warmer (remember we're displaying the image as reflectivity so that warm is bright and cold is dark).

4.c.1d Snow and ice cover and cloud over snow during daytime

The discussion centered around Figure 17 has already touched on the topic of locating water cloud over snow, and pointed out that snow cover would appear as dark in 3.9 µm imagery, while appearing relatively bright (depends on snow cover age) in visible imagery. The use of visible channel imagery to locate areas of snow cover were discussed earlier. Another example, over the Central United States, is shown in Figures 18a and 18b. Notice in the visible image in Figure 18 that much of the image is bright: the area is covered with both snow and cloud. The question is which is which? Recalling that ice particles are poorly reflective at 3.9 µm, discussed in detail previously, allows a user to separate the water cloud region from snow on the ground. While the information above has dealt mainly with separating water cloud from snow, how does one deal with cirrus? This problem is tricky, but may be approached in the following way. The nighttime fog product may be used to locate regions of cirrus cloud. Continuity into daytime and use of the 3.9 µm channel imagery allows a user to follow the movement of cirrus, thus helping separate moving cirrus regions from snow and ice surfaces.

Figure 18a. GOES-8 visible image taken on December 14, 1995. Both snow and cloud cover much of the image; see Figure 18b for clarification.

Figure 18b. Same as Figure 18a, but at 3.9 µm.

4.c.2 Cumulus phase during daytime

As might be expected, the potential for identifying cumulus that have entered the ice phase exists using both 3.9 µm and visible imagery. Inspection of Figure 19a shows a large thunderstorm complex along the Louisiana Gulf Coast. Well organized bands of cumulus and cumulus congestus are over the Gulf and feed into the storm area. When this same storm is viewed with 3.9 µm imagery, displayed as reflectivity, many of the some of the cumulus appear as black, while others nearby do not. The most likely reason for this is that the cumulus have entered the ice phase. This type information may prove to be important for identification of growing storms with the potential for electrical activity: this is an significant area for future investigations.

Figure 19a. Visible images of a large thunderstorm complex along the Gulf Coast on May 31, 1994. The aspect ratio for the GOES-8 (left) image is 0.57:1, while the aspect ratio of the GOES-7 (right) image is 0.87:1.

Figure 19b. 3.9 µm images of the thunderstorm shown in Figure 19a, and at the same time. The aspect ratio for the GOES-8 (left) image is 0.57:1, while the aspect ratio of the GOES-7 (right) image is 0.87:1. Ice clouds are black.

4.c.3 Fires and biomass burning

As discussed previously, imagery at 3.9 µm is very sensitive to sub-pixel hot spots. Because of this, large fires, perhaps as small as 200 acres (recall that there are 640 acres in a square mile) and 500 degrees K, can be detected using enhanced 3.9 µm imagery. While AVHRR is better suited for fire detection, it is limited by its polar orbiting time scale. Again this allows for a strong partnership between AVHRR and GOES; especially since GOES can follow the diurnal cycle in biomass burning.

Figure 20. 3.9 µm image of south central Brazil showing areas of fires. The fire areas are bright white in this reflectivity image.

4.c.3a Smoke detection

Figure 21 is over the same region and at the same time as the 3.9 µm image shown in Figure 20. This image has been specially enhanced to show medium brightness areas (smoke); regions with cumulus and cumulonimbus are saturated and bright white while ground is very dark. Notice the smooth creamy area near picture center; that area is smoke. But how can we be sure? The answer lies in the multispectral details found in the new GOES imagery. By using the 10.7 µm and 6.7 µm infrared channels, we can determine if this is an area of thin cirrus (section 4.c.6) - it is not. What about a low level water cloud? Recalling that the 3.9 µm channel is very reflective to small water droplets and inspection of Figure 20 reveals that this is not the case. In addition, smoke is composed of relatively large particles and appears dark in 3.9 µm during the daytime because it is poorly reflective.

Figure 21. Visible image over the same area and for the same time as in Figure 20.

4.c.4 Sea surface temperature

As with the tracking of sea ice, the ability for 30 minute updates to help clear clouds places the new GOES into a strong partnership with NOAA's polar system for evaluation of SST. The potential for improved sea surface temperature monitoring is possible using GOES 3.9, 10.7, and 12 µm images because of improved spatial resolution and spectral characteristics. During the daytime, use of the 10.7 µm and 12 µm channels allows for correction of surface temperature values that are corrupted by low level water vapor absorption in those channels. At night, 3.9 µm can provide valuable additional information. That channel has less water vapor absorption than at either 10.7 µm or 12 µm, its resolution is higher because of less diffraction (more energy entering the sensor from within an IGFOV) and it is less sensitive to clouds than either 10.7 µm or 12 µm.

4.c.5 Urban heat islands

The ability to locate urban heat islands under clear sky conditions, especially at night, is easily done using enhanced 3.9 µm imagery, as in Figure 22. As was pointed out earlier (and has been documented by a number of studies) there is inherent variability in land surface temperatures on a scale of fractions of a kilometer, especially in and around cities and their suburbs. GOES IGFOV's over cities may be thought of in the same sense as partially filled fields of views for fires or clouds. Thus, higher surface temperatures should be measured at night over cities at 3.9 µm vs 10.7 µm because of stronger response at 3.9 µm to the warm areas as well as more energy coming from within one IGFOV than at 10.7 µm.

Figure 22. 3.9 µm image from a clear night. Notice how well urban heat islands (bright) are depicted in this enhanced image.

4.c.6 Thin cirrus

As mentioned earlier, in regions of cirrus, where the cloud is thick, the resulting 3.9 µm image is very noisy. In areas of thin cirrus, since it is often patchy and since 3.9 µm is sensitive to warmer temperatures, means that thin cirrus area will have a higher energy signal reaching the satellite than a corresponding scene at 10.7 µm. At nighttime, for the nighttime fog product, that area will yield a difference of the opposite sign as for the water cloud area, helping locate the thin cirrus area. During daytime, the difference in reflection between ice and water cloud makes it fairly simple to distinguish between thicker cirrus and water cloud: cirrus is both cold and non-reflective, making it dark in a 3.9 µm image. However, in thin cirrus areas, energy from below will appear to increase its reflectivity, and other precautions are advised - like look at the 6.7 µm channels imagery (section 5) and animation.

4.c.7 Sun Glint

For many years the location and extent of sun glint has been done using visible imagery. Its importance in locating regions of smooth seas and weak surface winds is well recognized. Opportunity for new investigations in this region exist for two reasons: 1) the high quality imagery available with the new GOES visible channel gives users ability to enhance that imagery and have the resulting image remain near noise free, revealing detail in the sun glint region; and, 2) sun glint is also evident at 3.9 µm and is much more intense and extensive than at visible wavelengths. An unenhanced visible image of sun glint is shown in Figure 23a, with a companion image from the 3.9 µm channel in Figure 23b. Notice the greater extent of the sun glint in the 3.9 µm image.

Figure 23a. Sun glint as detected by GOES-9's visible channel while the satellite was located at 90 West. Glint region is at right-most center.

Figure 23b. Sun glint as detected by GOES-9's 3.9 µm channel while the satellite was located at 90 West (displayed as reflectivity). Note extensive glint area as compared to visible image above.

4.d 6.7, 10.7 and 12.5 microns

Figure 24a presents a high resolution atmospheric absorption spectrum and comparative blackbody curves for temperatures ranging from 200 K to 300 K. This spectrum was observed by one of the NIMBUS series satellites clearly shows the effect of various atmospheric gasses on what is observed by the satellite. At 6.7 µm notice that most of the radiance received by the sensor comes from very cold temperatures; this is because water vapor is a very active absorber at that in that portion of the spectrum, and thus any radiation reaching the satellite comes from emission of water vapor that is very high in the atmosphere (in fact the GOES channels saturate somewhere around 0.5 mm of water vapor). It can also be seen that most of the energy radiated from the surface, or cloud, around the 10.7 µm region reach the satellite, thus the term window since the temperature measured is close to scene temperature (recall the discussion about emissivity earlier). The window region around 12 µm, especially out toward 12.8 µm is contaminated by low level water vapor, and thus is called the Dirty Window. Notice also the window region around 4.0 µm (expanded in Figure 24b), this window is the "cleaner" than either 10.7 or 12.0 µm, but as mentioned earlier, it is contaminated by solar reflection during daytime. As with the channel at 3.9 µm, the channels at 6.7, 10.7 and 12.5 µm all have their characteristic measurement uncertainty ranges. These are presented in Figures 24c, d &e. In those figures notice how the measurements of temperature are less accurate at colder temperatures, as should be expected from previous discussions. Also notice how the channel at 6.7 µm is very accurate at cold temperatures.

Figure 24a. High resolution atmospheric absorption spectrum from 3.0 to 18.0 µm.

Figure 24b. As in Figure 24a, but expanded to show the region around 3.9 µm.

Figure 24c. Plot of radiance versus temperature at 6.7 µm.

Figure 24d. Plot of radiance versus temperature at 10.7 µm.

Figure 24e. (not available) Plot of radiance versus temperature at 12.0 µm.

The set of images that follow compare GOES-8 imager channels with GOES-VAS channels for the channels at 6.7 µm, 10.7 µm and 12.5 µm. These are at the same time, and may be compared with the visible and 3.9 µm images shown in Figures 19a and 19b. The full set of images are the focal point of the RAMM/CIRA tutorial "INTRODUCTION TO GOES-8," and the reader is referred to that tutorial for detailed color image comparisons. Several features are worthy of note when inspecting Figures 25a,b &c. First, in Figure 25a, which compares 6.7 µm imagery, note how much better features can be detected in the GOES-8 versus GOES-7 image. This is allowing for better detection of mesoscale features imbedded within synoptic circulations. For example, small scale standing waves have been detected in the water vapor imagery, but not other channel imagery, down wind from mountain ranges - very likely a locator of clear air turbulence. Furthermore, based on the work of Velden et al. (1992) high density water vapor winds, depicting atmospheric motions in clear regions, are now feasible on a routine basis. Finally, comparison of GOES-8 imagery over the thunderstorm area at 6.7 µm versus 10.7 µm and 12.0 µm reveals that the overshooting top temperatures are very nearly the same: this is most likely due to the effect of diffraction mentioned earlier. The ability to assess thunderstorm top temperatures very accurately with the 6.7 µm channel makes it a candidate for thunderstorm rainfall nowcasting, partly since it is a single detector and will not exhibit the striping that may occur at 10.7 µm. Furthermore, since radiation reaching the satellite from this channel comes from higher in the atmosphere that radiation at 10.7 µm (normally), thin cirrus will appear as colder at 6.7 µm versus 10.7 µm further aiding in its detection.

Figure 25a. Comparison of GOES-8 (left) and GOES-7 (right) imagery at 6.7 µm. The aspect ratio for the GOES-8 image is 0.57:1, while the aspect ratio of the GOES-7 image is 0.87:1.

Second, in Figure 25b, which compares 10.7 µm imagery, note the difference in appearance of the imagery over the open oceans. GOES-8 imagery appears slightly more distinct than the GOES-7 imagery. This is in part due to the higher sampling frequency and smaller IGFOV on GOES-8. In the thunderstorm region there is little difference in cloud top temperature, believed to be partly due to diffraction effects mentioned earlier.

Figure 25b. Comparison of GOES-8 (left) and GOES-7 (right) imagery at 10.7 µm. The aspect ratio for the GOES-8 image is 0.57:1, while the aspect ratio of the GOES-7 image is 0.87:1.

Finally, in Figure 25c, which compares 12.0 µm imagery, note that the GOES-7 image appears slightly cooler (grayer) over clear regions. This is because the GOES-7 channel in this spectral region extended out to 12.8 µm, where there is more absorption due to water vapor. While the 12 µm channel for GOES-8 has less water vapor absorption than GOES-7, derived products (precipitable water and lifted index) which use it in combination with the 10.7 µm channel appear to be an improvement over those from GOES-7. This is because the 12 µm channel on GOES-8 has much improved signal to noise and its resolution matches that from the 10.7 µm channel.

Figure 25c. Comparison of GOES-8 (left) and GOES-7 (right) imagery at 12.0 µm. The aspect ratio for the GOES-8 image is 0.57:1, while the aspect ratio of the GOES-7 image is 0.87:1.

5.0 Example of an advanced product: Principal component analysis of GOES multi-spectral imagery,

The information that follows was provided by Dr. Donald Hillger of the NESDIS RAMM Branch at CIRA (he may be reached at the same e-mail address as the author, simply by using his last name in place of the authors). It points to a fruitful region for further investigation and advanced product development. With multispectral satellite imagery, many of the spectral channels or bands contain redundant information about the atmosphere. By using Principal Component (PC) analysis to transform multichannel satellite images, this redundancy can be reduced. PC Image (PCI) analysis finds the information that is common among the various channel images and puts that information into the PCIs in descending order of significance. The first PCI contains common information, leaving other (difference) information for higherordered PCIs. A second PCI is then formed containing information common to the channels other than that explained by the first PCI. The process continues until the number of PCIs is equal to the number of channels being transformed. If none of the original channels contained redundant information then a PC transformation would not be needed. However this is not the case. Because of channel redundancy the number of useful PCIs is almost always less than the number of channels input. The highest-order PCIs may contain only noise or slight differences between some of the images, however, these differences may be of importance, especially for less obvious meteorological features in the atmosphere or for quality control (such as striping between detectors).

Many interesting examples of PCIs have been created from GOES-8/9 Imager and Sounder data. In the case of the GOES Imager, the interpretation of the PCIs is fairly predictable if they are created on a large spatial scale. The main differences in interpretation occur between day and night when visible radiation is an additional factor. However, for the GOES Sounder, the interpretation of the PCIs is more complex, since up to 19 channels can be included as input. Different subsets of the Sounder channels results in different products, with emphasis upon either temperature or moisture features in the atmosphere, or features of the ground surface or clouds. Even more variability in outcome is the result of concentrating on smaller spatial areas which contain more interesting but subtle meteorological features.

Figure 26. PCI from PC 4 which clearly shows the location of the dryline west Texas.

Don's assessment that visible PCIs are "fairly predictable" may be a slight exaggeration. PCIs clearly show features like haze, pollution, drylines, blowing dust, and snow cover; see Figure 26. No doubt, along with sounder channel PCIs a new and exciting way of viewing digital satellite data in image form is upon us.

6.0 Practical Implications of New GOES Observing Capabilities - Imager

* Much more detail in imagery combined with routine 15 minute views of the United States provides better coverage for meteorologist (the NWS Western Region WSFOs now receive digital GOES multi-spectral imagery every 15 minutes for analysis on RAMSDIS systems (Molenar, 1996) and a number of value added users, such as TV meteorologists: movies that are seen by most Americans during the evening news casts will be better.

* Better synchronization with other observations. Separate imager and sounder allow for more flexible scan modes. We are currently engaged in a Great Lakes winter storm exercise to help determine how to best utilize satellite and Doppler radar to enhanced storm forecasting and nowcasting using 5 channel imager data and sounder products. The anticipated success of this program will provide improved weather services for this important sector of our country. Great Lakes shipping, heavy snows in Buffalo, etc. RAMSDIS units are in place at several forecast sites for this exercise.

* Better cloud drift and water vapor winds. Best water vapor (6.7 µm) imagery ever (order of magnitude improvement enables identification of mesoscale disturbances in synoptic scale features). Improved winds are important for a number of reasons. Hurricane motion, improvement of numerical models which will result in better forecasts - this has major impact in all areas of our economy and quality of life, and improved winds for aircraft route planning. With the water vapor imagery, we can even see mountain waves in areas where clouds are not forming - this should improve turbulence forecasts for aviation.

* Use of the infrared channels at night will lead to enhanced land surface temperature monitoring: this is important for agricultural purposes such as early frost warning.

* Combination of the infrared channels at night also allow for the detection of fog when there are no other clouds to obscure the view: this is important for aviation purposes (Fed Ex and similar enterprises have numerous nighttime routes) and marine activities (there is a major Marine Risk Reduction activity underway at Boston where we have a RAMSDIS and this product). This area is an important one for combined polar and geostationary products.

* Use of the infrared channels at night will lead to enhanced sea surface temperature monitoring: this is important for shipping since SST gradients are related to currents. Also, we can do SST during the daytime, but not using the 3.9 micron channel. This area is an important one for combined polar and geostationary products.

* Timely depiction of changes in atmospheric moisture and stability - better delineation of gradients. There is no other sensor that can monitor low level moisture gradients as well as this satellite. This is very important for severe storm (tornado) forecasting: heat and moisture are the fuels for intense thunderstorms.

* Using the visible, 3.9 micron and infrared window channels: Capability to distinguish ice and water clouds during the daytime, and to detect low cloud and fog versus snow cover. For stratiform clouds, distinguish between ice, water and super-cooled clouds: this is important for aircraft icing is a super-cooled cloud phenomena and is extremely hazardous to small aircraft.

* Improved detection of forest fires and biomass burning. We should be able to develop products that will help with fire weather warnings. There are RAMSDIS units at Seattle, Salt Lake City and Boise that we hope to test these products with next summer. This area is an important one for combined polar and geostationary products.

* Useful imagery well beyond the satellite's 60 degree zenith angle (as seen from the satellite looking at the full disk, a circle whose radius extends directly North or South to 52 degrees: for GOES-8 this is about at the southern end of Hudson Bay), we can see things clearly up to the Arctic (or Antarctic) circles. This will allow for improved tracking of icebergs and monitoring of ice and snow cover. This area is an important one for combined polar and geostationary products.

* Best depiction ever of atmospheric changes ever in one-minute interval imaging. Work is currently underway to determine optimum ways of combining Doppler radar and SRRSD data and information. With one minute interval imagery around hurricane Rosa, we were able to track low level clouds at 70 knots near the hurricane eye. Rosa was estimated to have a 75 knot maximum wind.

7.0 CONCLUSIONS

This brief overview has pointed out several areas of improved capability with respect to observing oceanic and meteorological phenomena using NOAA's new GOES satellite's multi-spectral imaging capability. These were not the end of the list, but only the beginning. With its advanced imaging capability, the new GOES should be used as a strong partner with NOAA's polar system for observation and analysis of oceanic and meteorological phenomena.

8.0. ACKNOWLEDGMENTS

This research was supported by NOAA Grant NA37RJ0202. Special thanks are due to Bernadette Connell, Carol Vaughn, Jack Dostalek and Kathy Fryer of CIRA and Roger Phillips and John Weaver of RAMM Branch for their help in preparation of the figures.

9.0 REFERENCES (Abbreviated)

Chesters, D., 1994: Internet http://climate.gsfc.nasa.gov/~chesters/goesds.html

Ellrod, G., 1992: Potential ... 3.9 µm infrared imagery.

6th Conf. on Sat. Met. and Oceanog., 184-187,

Huh, O.K., et al, 1995: Remote Sensing of turbid coastal ...water-type analysis. J. Coastal Res.

Menzel, W.P. and J.F.W. Purdom, 1994: Bull. Amer. Meteor. Soc., 75, 757-781.

Molenar, D., 1996: RAMSDIS...; 12th Conf. IIPS, AMS, Atlanta, Ga.,

RAMM, 1994/1995: Internet address-

http://www.cira.colostate.edu.RAMM.overview

Rao, P. K.,et al, 1990: Weather Satellites: Systems, Data, and Environmental Applications. Amer. Meteor. Soc., Boston, 503pp.

Velden,

C.S., C.M. Hayden, W.P. Menzel, J.L. Franklin and J.S. Lynch, 1992: The impact of satellite-derived winds on numerical hurricane track forecasting. Weather and Forecasting, 7, 107-118.