In the simulated imagery presented here, AVHRR and GOES-7 imagery were taken within a few minutes of each other and mapped to GOES-7 projection for direct comparison. The AVHRR data were assumed to be noise-free and were treated as though they were 1-km x 1-km resolution at all pixel locations (only true at the subpoint); these assumptions should lead to a simulated image that is inferior to an actual GOES-I image.
FIG. 6. GOES-7 6-bit and simulated GOES-I 10-bit visible images from near 2011 UTC on 26 April 1992, illustrating the expected improvement in detection of cloud-top features using higher bit depth visible imagery. The GOES-7 image (a) and simulated GOES-I image (b) have been enhanced to show detail at cloud top.
FIG. 7. GOES-7 and simulated GOES-I infrared window images of the same thunderstorm top at the same time as Fig. 6. The GOES-7 image (a) and simulated GOES-I image (b) have been color enhanced to show detail at cloud top. Brightness temperatures are superimposed.
Figures 6a,b illustrate the expected improvement in detection of cloud-top features using 10-bit GOES-I visible imagery. Notice the detail in the overshooting top area and other regions of the anvil in the simulated GOES-I image. The storm with the well-defined overshooting top just south of the Kansas-Nebraska border (-67C in Fig. 7b) was a supercell that produced hail, damaging winds, and a series of F2 and F3 tornadoes. Figures 7a,b are infrared images from the same time as Figs. 6a,b and show the same thunderstorm area. The figures illustrate the expected improvement in detection of cloud-top features due to the improved resolution of GOES-I at 10.7 µm. Cold overshooting top areas are easily detected in the simulated GOES-I imagery. The storm along the border, discussed above, has a well-defined cold top and downstream warm wake that are easily detected in the simulated GOES-I imagery. Such features have been associated with severe thunderstorms (Heymsfield et al. 1988). The coldest cloud-top temperature was measured at -74C in the original AVHRR image; GOES-I shows-67C, while GOES-7 shows only-56C. It is interesting that the cold top (-64C in Fig. 7b) in Nebraska has a similar anvil structure that is readily detected in the simulated GOES-I image but not in the GOES-7 image. That storm produced large hail and an F1 tornado.
FIG. 8. GOES-7 and simulated GOES-I visible images over Florida shortly after sunrise on 21 December 1992. The AVHRR image from which the GOES-I image was simulated (b) was taken about 15 min. earlier than the GOES-7 image (a).
FIG.9. The 3.9-µm images for the same times as Fig.8 comparing GOES-7 imagery(a) with simulated GOES-I imagery(b).These images are displayed as reflectivity.
Figures 8a,b were taken shortly after sunrise on 21 December 1992; comparison of the images illustrates the ability to detect low-light features in GOES-I imagery. The GOES-7 and simulated GOES-I images are enhanced to show maximum detail in dark regions. Notice how well the light sand beaches, cloud detail, and shadows show up in the simulated GOES-I image. Cloud shadows enable computation of very accurate cloud heights, which promise to be valuable ancillary information for cloud-drift winds and ASOS cloud products.
Figures 9a,b are 3.9-µm images for the same times as Figs. 8a,b, and compare GOES-7 imagery with simulated GOES-I imagery. Since 3.9-µm imagery has both reflected and emitted radiation, a choice must be made on how to display that imagery. Usually, 10.7-µm imagery is displayed with warm scene temperatures (large brightness temperature) as darktones and cold scene temperatures (low brightness temperature) as bright tones; the reverse is true for visible imagery where bright surfaces such as clouds are displayed as bright tones and dark features are dark tones. These 3.9-µm images are displayed as reflectivity: at 3.9µm, cirrus (large particle ice clouds) are poorly reflective and cold (very dark tones), while low clouds with small water droplets are bright and relatively warm (lighter tones). Comparison of Figs. 8b and 9b allows discrimination between low and high clouds, and perhaps cloud phase. When the 10.7-µm band is used to add cloud-top temperature information, the potential exists to isolate regions of supercooled cloud.
Figure 10a shows a GOES-VAS (GOES-7) derived product image on 28 June 1993 at 0000 UTC of total PW calculated using radiance measurements from the longwave split window and the water vapor bands. Radiosonde reports are superimposed, and general agreement in synoptic trends is evident. Figure 1Ob shows the simulated GOES-I imager derived product image for 28 June 1993 at 0020 UTC of PW using the comparable imager spectral bands. The GOES-I PW image is cleaner and depicts moisture gradients (e.g., across Texas) on smaller scales than GOES-7; this is a direct result of GOES-I improvements in spatial resolution and signal to noise.
FIG. 10a. GOES-7 derived product image on 28 June 1993 at 0000 UTC of total precipitable water using the longwave split window and the water vapor bands; radiosonde reports are superimposed to provide a comparison.
FIG. 1Ob. Simulated GOES-I derived product image on 28 June 1993 at 0020 UTC of total precipitable water using the longwave split window and the water vapor bands.
FIG. 11. GOES-7 visible imagery for 0000 UTC 22 July 1993.
Principal conclusions are that the GOES-I sounder provides significant improvements in both coverage and accuracy compared to the current VAS product. Furthermore, results show that the data should be beneficial to the numerical forecast products, especially with regard to moisture. This conclusion is based on statistics that show that the retrievals are generally more accurate than the 12-h forecast (whereas current VAS is slightly worse in temperature and modestly better in moisture). Naturally, caveats apply. These results are based on simulation both in terms of what is believed to be the true state of the atmosphere and realistic radiance measurements. Neither is perfect, and history attests that a simulation of this type is almost always optimistic. To prepare for operational use of GOES-I sounder data, NMC and NESDIS are planning to conduct joint tests with simulated GOES-I moisture soundings in the NMC Eta model.
Figure 11 shows the GOES-7 visible image for the retrieval domain at 0000 UTC on 22 July 1993. The weather situation is typical of a summer afternoon. Figure 12a shows the VAS and Fig. 12b shows the GOES-I sounding coverage, the former plotted over the observed infrared window and the latter over the simulated infrared window. The retrieved lifted index is indicated at the location of each sounding. The VAS is processed at an 11 x 11 FOV density and the GOES-I at a more dense 5x 5 FOV. Retrievals are attempted when 25 and 9 FOV are estimated to be cloud-free for the VAS and GOES-I, respectively. The improved coverage and the ability of the GOES-I to find holes in the vicinity of clouds is obvious.
FIG. 13b. GOES-I simulated 7.0-µm measurements for 0020 UTC 22 July 1993.
Figure 13a presents the VAS 6.7-µm measurements (the water vapor band) as obtained in sounding as opposed to imaging) mode. The striping is caused by skipping every other pair of lines in order to increase latitude coverage for soundings in a given time period (necessary because the GOES-VAS imaging and sounding functions are time shared). The cloud clearing suffers accordingly. Figure 13b presents a similar (slightly lower in the atmosphere) band simulated for GOES-I. The moisture patterns are very similar, though better defined with GOES-I, which has a better signal-to-noise definition . The cloud contamination simulated at this wave-length appears to be reasonable. It should be noted that clouds are simulated at only one level, certainly not realistic but adequate for the purpose of defining fields of view that are not clear, and that is the criterion for rejection in the current retrieval algorithm (i.e., no partly cloudy FOVs are considered).
The rms statistics of the retrieved temperature and moisture profiles with respect to radiosondes show that the temperature and dewpoint at several levels of the atmosphere for the GOES-I retrievals improve upon the 12-h NMC forecast by 0.5-1.0 degrees C. This is especially apparent for moisture. At some levels, temperature accuracy is within 1.5C, but one should not lose sight of the fact that these excellent numbers are obtained chiefly because the forecast first guess is already excellent. Since the forecasts will likely continue to improve in both accuracy and resolution, so too must the sounders if they are to remain useful. An appropriate conclusion from these comparisons is that the GOES-I sounder and processing system is much better than VAS and should be of great benefit to numerical weather prediction responsibilities of the next decade. However, with the thrust of the NWS modernization directed toward improved accuracy of forecasts (local, mesoscale, synoptic scale, and long range), future improvements in the sounder beyond that of GOES I-M will be required.
