Regional and Mesoscale Meteorology Branch

GOES satellites measure energy in spectral regions ranging from the visible portion of the electromagnetic spectrum to the far infrared. At visible wavelengths, that energy is only reflected solar radiation (radiation from the sun which is reflected by the earth’s surface and clouds); at far infrared wavelengths, that energy is only emitted terrestrial radiation. However for the short wavelength infrared channel, the 3.9 um spectral band, energy measured by the satellite can be a mixture of solar radiation that is reflected by the earth’s surface or clouds and radiation that is emitted by the earth’s surface or clouds.

Figure 2a shows the Planck blackbody radiance curves for the sun (6000 K) and the earth (300 K). The energy received from the sun at the top of the atmosphere is represented by the area under the left-hand curve, and energy emitted by the earth is represented by the area under the right-hand curve. If all the sun’s energy reaching the earth were reflected back to the satellite, a satellite detector would sense the values represented by the solar curve (the left side of Fig. 2a). However, about 50% of the sun’s energy is selectively absorbed by various atmospheric constituents (ozone, water vapor, molecular oxygen, carbon dioxide, certain aerosols) and the earth’s surface. The remainder is scattered back to space by aerosols and reflected by clouds and the earth’s surface. That scattering and reflection is a function of wavelength and the particular constituent (cloud phase/droplet size, soil type, etc.) with which the interaction is occurring. This reflected and back scattered solar energy can be detected by a satellite sensor. The vertical lines in the figure locate the spectral region sensed by GOES in the 3.9 um band. Satellite detectors do not measure energy at a single wavelength, the GOES imagers’ 3.9 um channel extends from 3.78 – 4.04 um. In the figure, notice that satellite measurements in the 3.9 um band are a combination of earth emitted and solar reflected radiation.

Figure 2b shows a close-up view of the portion of Figure 2a (previous page) where the Planck curves for the sun and the earth overlap. Curves for solar reflection at 50% and 20% are shown as well. The exact combination of solar and terrestrial energy measured by the satellite at 3.9 um depends on the time of day as well as the reflectance and emissivity of the underlying surface. This combination of emitted terrestrial energy and reflected solar energy during daytime, combined with information on cloud and surface characteristics, is one of the things that make interpretation of imagery at 3.9 um so interesting.

The contribution to the measured radiance temperature at 3.9 um, due to the reflected solar component, may be determined from Fig. 2c. The set of curves shows the radiance temperature that would be measured by a satellite for cloud tops at several temperatures (with albedo = 0 and emissivity = 1.0), with respect to increasing amounts of reflected solar radiation. Notice that the radiance temperature measured at 3.9 um begins to converge near an upper limit, around 350 K. This convergence occurs because as cloud albedo increases, the addition of reflected solar radiance far outweighs the cloud’s emitted radiance (evident in Fig. 2b). When inspecting Fig. 2c, keep in mind that cloud top temperatures above 285 K are rare.

The GOES-8 3.9 um sensor gain is set to saturate at 335 K, and 3.9 um saturation for GOES-9 occurs at 325 K. Saturation for this sensor in future GOES Imagers, like GOES-8, will be at 335 K.

Emissivity is a function of both wavelength and surface type. Fig. 3a is a plot of emissivity vs. wavelength for four different surfaces. Notice that the emissivity for soil types is more variable near 3.9 um than 10.7 um, e.g., dry sand has an emissivity near 0.8 at 3.9 um and 0.95 at 10.7 um. Imagery of sandy areas appears cooler at 3.9 um than at 10.7 um when there is no reflected solar radiation (e.g., at nightime, in a dry atmosphere). This “apparent” temperature difference affects the interpretation of derived image products

Referring again to Fig. 3a, strong absorbers at a particular wavelength are also strong emitters at that wavelength. Furthermore, during the daytime, radiation that is not absorbed is reflected by the surface. Because of this, in the 3.9 um imagery, land surfaces will appear different, depending on their composition. This is especially apparent over regions like White Sands, NM, where the soil type is mainly sand (g = 0.75 to 0.85) and across north-central GA, AL and MS, where more reflective soils are found. The imagery shown in the “fire detection” section also demonstrates this characteristic.

Reflection at 3.9 um is sensitive to cloud phase and is very sensitive to particle size, as is shown in Fig. 3b. Notice how water droplets are more reflective than ice particles of the same size. In clouds, water droplets are normally between 5 and 20 um in diameter, depending upon cloud type, while ice crystals are usually an order of magnitude larger. Reflection that is detected by a satellite is from multiple scattering (Fig. 3b represents a single scatter) and each reflection further reduces the amount of energy returned to the satellite. For example, three reflections each, by ice at 100 um and water at 10 um, yield scattering values of 0.16 and 0.73 respectively.

During the daytime, clouds with small water droplets, such as cumulus, fog and stratus over land, are much brighter when viewed at 3.9 um than are ice clouds, which are very poorly reflective and hence, dark. Marine stratocumulus, with their larger water droplets, appear relatively dark when compared to cumulus or stratus over land.

Aside from emissivity and reflectivity, several other factors are responsible for differences in the appearance of imagery between the 3.9 um and 10.7 um bands: 1), different responses to scene radiance make possible the detection of sub-pixel hot regions at 3.9 um that are not detected at 10.7 um; 2), the exaggeration of noise at cold temperatures in the 3.9 um band makes it virtually useless for thunderstorm top analysis; 3), because of diffraction effects, the 3.9 um and 10.7 um bands view slightly different areas; and 4), criteria for displaying 3.9 um imagery may differ between day and night since the 3.9 um band also contains reflected solar energy during the daytime.

For more in-depth discussion of these differences, please review the following topics:

Temperature Responsivity

Sub-pixel Response

In Fig. 4a, satellite-measured radiance temperatures at 3.9 and at 10.7 um are compared at night-time, when there is no reflected radiation at 3.9 um, for a FOV where the ground at 300 K becomes increasingly covered with clouds at 260 K. Also, in this case, surface emissivities are assumed to be equal to 1.0 and atmospheric absorption is assumed to be negligible.

Notice that when the FOV is either totally clear or totally cloud covered, the radiance temperatures at 3.9 and 10.7 um are the same. However, as fractional cloud cover increases, the radiance temperature at 3.9 um becomes greater than the radiance temperature at 10.7 um due to the stronger response at 3.9 um to the warm portion of the partially filled FOV.

Figure 4b shows the radiance temperature difference,

between the 3.9 um and the 10.7 um wavelengths as a function of cloud fraction. In this case (ground at 300 K and cloud at 260 K) the radiance temperature difference reaches a maximum of nearly 6 K, at about 65% cloud cover. The difference in radiance temperature may be of value in determining sub-pixel information about the clouds. If uniform cloud and ground cover (whose temperatures are known) can be assumed, then the fractional amount of cloud can be determined (if the effects of diff

Noise

Figs. 4c & 4d are plots of radiance versus temperature for the GOES channels at 3.9 um and 10.7 um respectively. The accuracy of the radiance measurements at each wavelength is constant and is shown by the horizontal dashed lines in each figure. However, temperature measurement accuracy, shown by the corresponding vertical dashed lines in each figure varies with respect to scene temperature. Notice that the radiance at 10.7 um (Fig. 4d) is fairly linear with temperature compared with radiance versus temperature at 3.9 um (Fig. 4c). This means that 10.7 um radiance temperatures may be determined very accurately for both warm and cold scene temperatures. In the 3.9 um figure, notice how radiance increases rapidly with increasing temperature. Also notice how “flat” that curve is at cold temperatures. Since the inherent measurement accuracy of the GOES instrument at 3.9 um is constant, the result is a much less accurate temperature measurement at cold versus warm scene temperatures. Interpretation of this figure shows that GOES’ 3.9 um imagery is less useful for analyses in cold temperature regions, such as thunderstorm tops; however, for measurements of warm surface temperature the 3.9 um channel does a fine job.

Fig. 4e is a plot of noise equivalent temperature as a function of scene temperature for the 3.9 and 10.7 um channels. The figure shows that the accuracy with which temperature can be measured at 3.9 um is worse than 2 K for scene temperatures below about 250 K. On the other hand, at 10.7 um, the accuracy is always better than 0.5 K for all possible scene temperatures. Another way of showing the effect of low scene radiance at 3.9 um is by inspecting the signal-to-noise ratio (S/N) versus temperature, Fig. 4f. Signal-to-noise, as used here, is the ratio of the scene radiance divided by GOES’ radiance accuracy. Fig. 4f shows that the 10.7 um channel has a much better S/N than the 3.9 um channel. In fact, the S/N at 3.9 um decreases to where there is no signal above noise at temperatures below about 230 K, making this channel of little use at colder scene temperatures.

Diffraction