Attempting to Turn Night into Day; Development of Visible Like Nighttime Satellite Images

Visible satellite images have long been used in aviation flight planning. The visible satellite images show a variety of phenomena of interest to aviation, including fog, low clouds, thunderstorms, etc. Since one's eyes detect visible light, the visible satellite image is easier for untrained personnel to interpret than other bands. The biggest problem with the visible images is that they are not available at night. However, other channels on the satellites can be used to develop a derived satellite product which looks very much like a visible satellite image. This derived product can then be inserted into the nighttime portions of the visible images to allow for a continuous “day/night” visible image. The original nighttime derived product was the “fog” images (Ellrod, 1995) generated from the difference between the 3.9 micron channel and the 11 micron infrared channel on the GOES satellites. At night the difference between the 3.9 and 11 micron channels detects emissivity differences rather than absolute temperature (Ellrod ,1995). These emissivity differences are related to the size of the cloud particles, so small droplets (such as occur in fog) can be readily distinguished from larger ice crystal clouds or the ground. Hence low clouds can be detected at night even if they are at the same temperature as the ground. Taking the difference between the two channels will generate a low nighttime cloud image which appears white, but the cirrus clouds will appear black, with the ground being gray. While the visible/fog image is useful, it has problems for untrained users. One of the big problems with visible satellite images for untrained users is distinguishing between high and low clouds. Mosher (2006) developed a visible/fog image with the clouds above 18,000 feet having a blue tint with the clouds below 18,000 feet being black and white. Rather than have the nighttime high cirrus clouds black, Mosher (2006) inserted the infrared channel into the nighttime image for the sections above 18,000 feet. While this helped make a more visible like nighttime product, there were still problems. High thin cirrus with low emissivity would appear warmer and lower, so black cirrus clouds would still show up at night in the low cloud section. Another problem with using the infrared image for the portion of the image above 18,000 feet is that active thunderstorms are difficult to detect relative to inactive cirrus clouds. Active thunderstorms have overshooting tops and towers do not have much temperature variation because of the nearly isothermal vertical temperature structure near the tropopause. An enhanced version of the day/night visible satellite images have been developed to make the product even more visible like. The problem of thin low emissivity cirrus clouds having too warm apparent temperatures was addressed using a correlation between the IR and Water Vapor channels. The warm appearance of the low emissivity cirrus cloud is due to the lower level radiation going through the cloud and being mixed with the radiation from the cloud top. Since the Water Vapor channel has radiation coming from the water vapor molecules in the atmosphere rather from the surface or low clouds for the IR, the Water Vapor channel apparent temperature is closer to the true cirrus temperature than the IR apparent temperature. A correlation between the IR and WV channels is run. Pixels having a correlation between the two channels, and the pixel being brighter than the running mean are identified as being cirrus. The IR temperature is replaced with the WV temperature for those pixels identified as cirrus. The problem with the thunderstorm tops not showing up sufficiently was addressed using the difference between the 11 micron IR channel and the 13.5 CO2 channel. The CO2 channel detects the radiation from the cloud and the radiation from the CO2 molecules in the atmosphere. The difference between the two channels is a function of the depth of the atmosphere, so high clouds will have a lower difference than lower clouds. Since the difference is dependent on atmospheric depth, not cloud temperature, towering clouds will show up against the thunderstorm top, even if they have similar IR temperatures. The new day/night visible images are generated by first doing brightness normalization on the visible by dividing by the cosine of the solar zenith angle. Then the visible image is divided into two brightness ranges for cloud above and below 18,000 feet. An enhancement table is then used to tint the higher clouds blue. The IR temperatures are corrected for thin cirrus, and then used in the cloud height determination. For the lower section (below 18,000 feet) of the nighttime image, the 3.9-11 micron difference fog image is generated. The nighttime upper sections of the image are generated using the IR-CO2 difference and for dark pixels of the lower section. For the tropics, the low level moisture also generates dark surface images for the 3.9-11 difference. Rather than replacing these dark low level pixels with the IR-CO2 pixels, and pixels below 6,000 feet are reinserted back as the original 3.9-11 micron fog pixels. The net result is a nighttime image which looks very much like a daytime visible image. Figure 1 shows an example of the derived image in the Texas region for June 8 at 11:32Z. Visible pixels are on the right while the derived nighttime pixels are on the left. Figure 1. Day/Night Derived image for June 8, 2012 at 11:32Z. The visible image is on the right with the derived nighttime image on the left. References: Ellrod, G. P., 1995: Advances in the detection and analysis of fog at night using GOES multispectral infrared imagery. Wea. Forecasting, 10, 606-619 Mosher, Frederick R.: “Day/Night Visible Satellite Images”, Preprints, 14th Conference on Satellite Meteorology and Oceanography, Atlanta, Georgia, Feb. 2006.