Day/night visible satellite images

Visible satellite images are very useful for a wide variety of users. In particular, they are helpful in identifying areas of clouds and fog for general aviation pilots who must fly within sight of ground. However visible satellite images have several major drawbacks, such as at night the visible pictures are black. Another problem is that it is sometimes difficult to distinguish between high clouds and low clouds. While the infrared channel can be used at night, frequently the low clouds and fog are near the temperature of the ground, so low clouds do not show up well on infrared images. Starting with the GOES-8 geostationary satellite series, the 3.9 micron channel has been available. At night the difference between the 3.9 and 11 micron channels detects emissivity differences rather than absolute temperature, so low clouds can be detected at night even if they are at the same temperature as the ground. Ellrod (1995 ) has developed a difference product (available at http://www.orbit.nesdis.noaa.gov/smcd/opdb/aviation/fog.html) which can be used at night for low cloud detection. A similar product is available on the AWIPS workstations used by NWS forecasters. However, during the day the sun has some radiance at the 3.9 micron frequency, so during the day there is a blend of the emissivity differences and the reflected solar energy. Hence there is a difference in appearance between the day and night which make interpretation of the 3.0-11 micron images difference images difficult, especially for users not trained in satellite image interpretation. To try to improve the interpretation problem, the Aviation Weather Center has a vis/fog product (available at http://aviationweather.gov) that inserts the visible image into the daytime portion of the image. While this is an improvement for low cloud surveillance, there still is a major discontinuity in appearance of the high clouds between day and night.

The problem of day/night difference in appearance also limits the potential to make visible global mosaics. A number of groups make global infrared and water vapor composites (such as SSEC at http://www.ssec.wisc.edu/data/composites.html), but no one makes a visible global composite. Not all of the satellites scan at the same time or at the same rate, so there are time differences between sections of the mosaic. For infrared and water vapor, the cloud displacements are relatively minor, but for the visible the sun is moving at 1000 mi/hr, so minor time differences between images make large appearance differences. To solve this problem, one needs the nighttime section of the images to have the same approximate appearance as the daytime. This requires a synthetic nighttime “visible” image and a brightness normalized visible image during the day. This purpose of this day/night visible effort has been to develop such a product.

The nighttime portion of the image is constructed from 11 and 3.9 micron channels. The difference between the images is stretched with a digital enhancement that leaves the low clouds with small droplets white, the ground gray, and the high clouds with large ice particles black. The dark, ice cloud pixels are then replaced with infrared image pixels, as are any pixels higher than 18,000 ft. The 18,000 ft. temperature threshold is determined from a modified standard atmosphere. The image is then divided into two images, those pixels above 18,000 ft. and those below. The brightness of the pixels above 18,000 ft are stretched into the gray scales between 1 and 64. The brightness of the pixels below 18,000 ft are stretched into the gray scales between 65 and 255. The two images are then recombined into one image. For the display an image enhancement is used that stretches the 1-64 gray scales into a 0-255 blue scale with the 65-255 being stretched into a 0-255 black and white scale. The net result is an image with the higher clouds having a blue tint, and the lower cloud looking white.

The daytime portion of the image is generated from the visible and infrared channels. The visible pixels are first brightness normalized by dividing by the cosine of the solar zenith angle and then utilizing an empirical correction for the departure from a lambertion reflector. Then the brightness normalized visible image pixels are divided into two images, above and below 18,000 ft. as were the nighttime pixels. The visible pixels above 18,000 ft. are stretched into the 1-64 brightness scale while the lower pixels are given the 65-255 gray scales. The images are then reconstructed into a single daytime image. The nighttime and daytime images are then reconstructed into a single image that is then displayed with an enhancement table that gives the clouds above 18,000 ft a blue tint. The following is an example of the sunrise transition from night on the left to day on the right.

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