The Cirrus Detection product is based on measurements from the NASA Moderate-resolution Imaging Spectroradiometer (MODIS). It is actually composed of two algorithms – one that works for day and the other for night. We can take advantage of additional information from solar reflection channels during the daytime orbits to produce a superior product. At night we resort to other channels for cirrus detection, but must be more conservative in order to avoid false alarms in the enhanced imagery.
The daytime algorithm hinges on the measurement of reflected sunlight in the 1.38 µm shortwave infrared channel mentioned above. This channel is special because the abundant water vapor of our planet’s atmosphere is strongly absorbing in this region, meaning that sunlight that travels too deep into the atmosphere has no chance of reflecting back to the satellite. Cirrus clouds, which reside high in the atmosphere and above most of the water vapor, reflect this 1.38 µm radiation with minimal effect of water vapor absorption, and are therefore sensed by MODIS. Low-level clouds and the surface, on the other hand, are “buried in the vapor” and effectively filtered from the scene. Reflection detected in the 1.38 µm channel above a certain threshold is considered “cirrus” and color-coded as blue.
To separate thin from thick cirrus, we use two more MODIS channels (6.7 and 11.0 µm) situated in the infrared part of the spectrum. The 6.7 µm channel corresponds to another one of those “water vapor absorption bands” just like the 1.38 µm channel, but since it’s in the infrared part of the spectrum (where sunlight has minimal contribution) here we are dealing with Earth/atmosphere emissions only. The 11.0 µm channel is in a so-called “clean window” region, in the sense that water vapor and other gases are mostly transparent to radiation having this wavelength (so from a satellite vantage point w could in principle see all the way down through the intervening atmosphere to the surface in this band, if a cloud is not in the way). Measurements from these two channels are expressed in units of temperature (Celsius or kelvins (= Celsius+273.15)), and we are interested in the difference between these two measurements as a way to find thick, deep clouds in the scene.
The idea is that low clouds will again be “buried” in the water vapor, and since the 6.7 channel sees the temperature of the cooler water vapor present above the cloud (since temperature decreases with height), the satellite-measured temperature will be lower than the actual cloud top temperature. The same cloud observed at 11.0 µm (the window channel) will yield a temperature much closer to the true cloud top temperature. So, the difference between the two measurements (6.7 vs 11.0) will be large. Conversely, for a high and thick cloud, there is not much water vapor above it to depress the relative difference between 6.7 and 11.0, so the values in this case are small. Essentially we look for small differences between these two channels as a proxy for where the high and thick clouds are. We color code these areas as red. In terms of color composites, anywhere where we have both cirrus (blue) and “thick high cloud” (red) forms a magenta color.
While the 6.7-11.0 µm difference technique described above also applies to night imagery, we no longer have access to the 1.38 µm information to identify any thin cirrus component. Fortunately, another channel difference between 3.7 µm and 11.0 µm works fairly well for this purpose. 3.7 µm is a special channel because it is sensitive to both sunlight and earth/atmosphere radiation. At night, the sun-component is removed and this channel is very useful for detecting warm surface emissions (this channel is the cornerstone of fire detection). Because it is so sensitive to heat, any small amount that transmits through thin cirrus to reach the satellite creates a strong signal. This is in contrast to the 11.0 µm measurement, which will report a cooler temperature. In this way, we look for large differences between 3.7 and 11.0 µm as a way to highlight thin cirrus, which gets color coded as blue just like in the daytime enhancement.
Fig. 2. Example of the differences in appearance between the daytime (left) and nighttime (right) cirrus product for the same thunderstorm complex viewed roughly 9.5 hours apart.
Figure 2 shows an example of the Cirrus Detection product during the day and night for a squall line passing over southern Texas. In both day and night imagery, thin cirrus clouds appear dark blue while thick cirrus/t-storm tops appear magenta. Areas devoid of cirrus are substituted with visible-channel imagery during the day time, or infrared imagery at night. These regions are represented in black & white to provide context of the meteorology in terms of other cloud formations that may be present in the current scene.
What to look for: Mature thunderstorms typically appear as oval shapes with bright magenta cores and potentially large fans of blue (thin) cirrus, depending on the strength of the upper level wind field. Cirrus associated with large, organized storm systems often appear as extensive shields ahead of the storm front. Jet stream cirrus, also commonly observed during the winter time months and associated with storm systems, typically appear as elongated chords following an undulating path along the upper level flow.
Things to watch out for: Since the 1.38 µm channel only applies to daytime scenes, the nighttime cirrus product lacks the same high sensitivity to thin cirrus as the daytime product. Variable surface properties (especially over land) and lower atmospheric water vapor make the problem even more challenging. At night the product sometimes fails to pick up all the thin cirrus. Occasionally there will be “gray rims” surrounding enhanced cloud structures; the product has failed to discern these clouds as cirrus. Figure 3 illustrates an example of these missed thin cirrus regions (some pointed out with white arrows) associated with the remnants of thunderstorm anvils over southern Texas and northern Mexico.
Fig. 3. Example of gray rims (white arrows) around optically thin cirrus clouds, which are clouds that the nighttime Cirrus Detection algorithm failed to identify.
Another potential issue arises on occasion for high, snow-covered terrain and relatively dry atmospheric columns. In these cases there may not be sufficient total column water vapor to completely mask the bright, elevated surfaces, which begin to appear in the Cirrus Detection product as false cloud features particularly in the daytime imagery. These features are usually easy to spot (dendritic patterns tied to the topography), and in the GOES-R era would be even more readily identified as static features in the dynamic field. However, for demonstration from MODIS (where very limited temporal refresh is available) identifying these artifacts may be a greater challenge, so users should be more cautious when assessing this product in mountainous regions.
Owing to variable cloud opacity, infrared imagery does not provide a direct means to isolating the high cloud cover in the scene. A key advantage of the Cirrus Detection product is the ability to quickly identify regions of high cloudiness and qualitatively assess their opacity via color information (blue = thin, magenta = thick).
Although an attempt is made to provide a day/night capability, the algorithms are indeed different.