Color Day–Night Visible GOES Satellite Images

The day/night GOES visible satellite images available at http://wx.erau.edu/erau_sat/ are full color images constructed so as to provide easily interpreted information on clouds both day and night. The goal of this product is to generate satellite images which appear similar both day and night. During the daytime a three channel color image is generated. The clouds above 21,000 feet are tinted light blue to distinguish the high clouds from the lower clouds. At night an image is derived from multiple infrared channels to generate an image similar to the daytime visible image.

During the day the color images are generated from bands 1 (blue), 2 (red), and 3 (vegetation) on the GOES-east satellite. The GOES satellite does not have a true green channel, so the .8 micron vegetation channel is used as a substitute green channel. Chlorophyll in green plants causes a weak reflectance (about 5%) in the green visible region of the spectrum, and a strong reflectance (about 30%) in the .8 micron near infrared spectrum. To tone down the vegetation channel into a green channel, a weighted average of the blue, red, and vegetation channels is performed. Weights of .29, .29, and .33 are utilized for the blue, red, and vegetation channels respectively to generate a “green” channel. The three bands are then combined into a color image.

In order to distinguish the high clouds (above 21,000 feet) from the low clouds a corrected infrared temperature is converted to cloud top height. Thin cirrus clouds have an emissivity less than one, so there is a mixture of radiation coming from the cloud and radiation going through the cloud from below. This causes the thin clouds to have an infrared temperature that is warmer than the true cloud temperature. To correct for this warm cirrus cloud bias, a correlation analysis is performed between the infrared temperature (band 13) and the water vapor temperature (band 9). Pixels which show clouds in both the water vapor and infrared channels have the water vapor temperature inserted into the infrared image. This corrected IR temperature is then used to determine the cloud height. The conversion of temperature to height is done using the 1966 seasonal US Standard Atmosphere Supplement soundings To generate the high/low cloud image, the original images are separated into two images; one with pixels above 21,000 feet and another with pixels below 21,000 feet. The higher cloud image brightness values ( 0-255) are compressed into the 191-255 brightness range. The lower cloud image bright values are compressed into the 0-190 brightness range and then merged back with the high cloud image. The two high/low brightness ranges are expanded back into the full brightness ranges using an enhancement table on the display, with the high cloud range having a little blue added to the enhancement.

The three channels are each brightness normalized. The brightness normalization corrects for the changing brightness caused by the changing sun angle. The basic correction is to divide the brightness by the cosine of the solar zenith angle (the angle from the overhead at a spot to the sun). However, this breaks down at sunrise/sunset when the cosine of the solar zenith angle is zero. The routine limits this effect in two ways. First, if the sun is within 3 degrees of the horizon, the processing is terminated. The second modification is a correction term based on multiple scattering in clouds. The correction is to multiple the normalized brightness by 1/(1+ (sun's zenith angle in radians)/pi). This reduces the cloud brightness near the terminator zone to better approximate the cloud brightness at higher sun angles. For the blue channel, in addition to the brightness normalization, a simplified Rayleigh scattering correction is applied. The term 27*(1-cos(sun angle)*(1-cos(sat angle)) is subtracted from the normalized brightness value. The value 27 was obtained from looking at the difference in the sea brightness value near the equator between local noon and local sunset for GOES-16 channel 1 normalized images. Both the normalization and Rayleigh Scattering corrections assume isotropic scattering (i.e. scattering equally in all directions). This isotropic scattering assumption is a reasonable assumption for most situations. However, for forward scattering situations the brightness normalized clouds can be too bright.

The nighttime portion of the image is constructed from multiple IR channels. The 10.3-3.9 micron 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. This is used for the low cloud portion of the nighttime image. The 10.3 – 9.6 micron difference is used to generate the high cloud part of the image. There also is a slight difference in the emissivity of clouds between the two channels related to the thickness of the clouds. This allows for thin cirrus to be depicted as dim clouds and thick thunderstorms to be depicted as bright clouds. This difference allows overshooting tops of thunderstorms to be more noticeable than a simple IR temperature image. This difference is used for all the high cloud portion of the image as well as for the dark portions of the lower cloud image. The low and high cloud images are merged to form the nighttime portion of the image product. The nighttime image is then “colorized” by adding 15 counts to the pixels with counts less than 80 for the green land channel and 30 counts to the pixels less than 80 to the clear water areas for the blue channel. The nighttime portion of the image is then merged with the daytime portion for the final products.

The attached figure shows the sunrise line with the left side of the image being in the dark and the right side being in the sunlight.