Extended version of paper presented at: The EUMETSAT Meteorological Satellite Data Users' Conference, 16 - 20 September 1996, Vienna, Austria ( Last modified: 29 November 1996 )

Spatial and temporal characteristics of convective storm tops in the 3.7/3.9 µ spectral bands are examined from meteorological satellite imagery. Previous observations from Europe, based on NOAA/AVHRR data, have shown that some of the storms exhibit a significant increase of cloud top 3.7 µ reflectivity. Recent observations have shown that similar cloud top phenomena exist in the NOAA/AVHRR data for convective storms over the U.S. Great Plains. Launch of the GOES-8 and GOES-9 geostationary satellites has enabled a study of development of storm cloud top structures in a similar band (3.9 µ) with high temporal resolution. This imagery shows that the smallest areas of increased 3.9 µ reflectivity appear and fade on the scale of few minutes, although the larger ones can persist for tens of minutes to several hours.

Selected cases of convective storms exhibiting an increase in the 3.9 µ reflectivity have been studied with respect to internal storm structure as observed by NEXRAD Doppler radars. This has revealed that majority of spots of increased 3.9 µ reflectivity can be found above no significant (or weak) radar echo region, though close to storm's core. However, some of these spots have appeared above a mesocyclone at the time (or close to it) of associated tornado touchdown, which suggests that these spots might be of different origin. One case of high 3.9 µ reflectivity over entire storm tops has been recorded on 22 May 1996 simultaneously from GOES-8 and GOES-9. Some aspects of bidirectional scattering are evident from the differences in reflectivity observed from these two satellites.

1. INTRODUCTION

The GOES-8/9 channel 2 (3.78 - 4.03 µ) and the NOAA/AVHRR channel 3 (3.55 - 3.93 µ), all located in a spectral band that includes both emitted and reflected solar radiation during the daytime, represent a unique observational tool for studies of convective storms' cloud top microphysics. Given the very low temperatures of anvil tops, the emitted component in these channels is almost negligible when observing storm tops. Hence, the reflected component plays the major role in appearance of convective storms in these channels in daytime hours (Setvák, 1989).

Earlier NOAA/AVHRR observations of convective storms over Europe have shown that some of these exhibit a significant increase of the 3.7 µ cloud top reflectivity and that the observed features fall into two broad classes (Setvák and Doswell, 1991):

- areas which are either spot-like, typically located close to overshooting tops (as determined from visible or thermal IR imagery) or more widespread with fuzzy or blurred edges. These features range in size from that of a single AVHRR pixel to the extent of the entire anvil top.

- a plume like shape, emanating from almost a pixel-size source, typically located downwind from the coldest tops (Levizzani and Setvák, 1996). This form has been observed much less frequently than the spot-like class (for an example see Figure_1).

Present work extends the research to U.S. Great Plains storms, which offers several significant advantages. First, comparison of the U.S. and European storms can generalize the observations from Europe, and inferences made from them. Second, the broader range of observational tools available in the U.S. can help link the observed cloud top features to internal storm processes and to accompanying weather phenomena. Finally, introduction of the GOES-8/9 satellites (Menzel and Purdom, 1994) enables determination of the evolution of these cloud top phenomena with temporal sampling from 15 to 1 minute.

2. DATA SOURCES AND PROCESSING

2.1. NOAA/AVHRR

The AVHRR/HRPT data from NOAA11 and 12 for 1994, NOAA 12 and 14 for 1995 and 1996 have been obtained from the NOAA/NESDIS Satellite Active Archive in level-1B format. The AVHRR data sets were processed (calibrated and georeferenced) by software written at the Czech Hydrometeorological Institute (CHMI) for MS-DOS platforms. Daytime data from the AVHRR channel 3 have been converted into 3.7 µ reflectivity by an algorithm developed at CHMI (Setvák and Doswell, 1991). An advantage of using 3.7 µ reflectivity instead of the more commonly used channel 3/4 brightness temperature difference as a characteristic of a cloud top is that the reflectivity is normalized by incident solar radiation, which is not true for the ch3/ch4 difference.

2.2. GOES-8

Most of the GOES-8/9 imagery were obtained from the NOAA/NESDIS archive. In addition, data were supplied from the NESDIS Regional and Mesoscale Meteorological Branch (RAMM) and from the NCAR Cooperative Program for Operational Meteorology, Education, and Training (COMET). Data were processed using the University of Wisconsin Man Computer Interactive Data System (McIDAS).

Daytime reflectivity at 3.9 µ (GOES-8/9 channel 2) was computed from measured radiance at 3.9 µ and 11 µ (channel 4) following the same methodology used for the AVHRR data. Visible (channel 1) data have been used for verification of optical thickness of the anvils to exclude "false spots" where gaps in cloud top might appear.

2.3. OTHER DATA

Radar data (WSR-88d) were obtained from the National Climatic Data Center (NCDC) archive for individual NEXRAD radar sites which recorded data in level-II format. Reflectivity and radial velocity data were displayed using the Radar and Algorithm Display System (RADS) developed at the NSSL.

Supplementary surface and rawinsonde observations obtained during the 1995 VORTEX field experiment in the southern U.S. Plains (see Rasmussen et al.1994) augmented routine meteorological data during this period. Surface observations of hail, high winds, and tornados were obtained from the log of severe weather maintained by the NOAA Storm Prediction Center and displayed using software Severe Plot on a PC.

3.OBSERVATIONS

3.1. NOAA/AVHRR OBSERVATIONS

Examination of about 30 AVHRR data sets from 1994 and 1995 has shown that 3.7 µ features, observed for European storms, can be found also over some of the U.S. Great Plains convective storms. Though the number of investigated cases is much smaller (to date) compared to the European observations, preliminary results indicate significantly higher frequency of plumes than in Europe, though the majority of these is detectable in visible and near infrared channels only (showing no increase of the 3.7 µ reflectivity). However, this may result from the fact that most of these plumes were found on early evening NOAA-12 images, when low elevation of the Sun does not provide enough 3.7 µ radiance to illuminate the scene sufficiently. Nevertheless, in general these observations have proved similar appearance of storm cloud tops in the AVHRR 3.7 µ channel for both continents.

3.2. COMPARISON OF THE AVHRR AND GOES 8 DATA CALIBRATION

As a first step, brief cross-calibration tests have been performed for processed AVHRR/HRPT NOAA-14 and GOES-8 data. Comparison of thermal IR channels (AVHRR channel 4 and GOES-8 channel 4) has shown close agreement between these. The brightness temperatures from both instruments are within about a half degree Kelvin at both ends of temperature range (warm ground surface and cold tops of storms). This is in much better agreement than found previously from AVHRR/NOAA versus Meteosat cross-calibration tests (Levizzani et al, 1992) where the differences were as high as 7 K in the low temperature range (~ 200-220 K), owing to resolution differences and calibration uncertainty at low temperatures on Meteosat.

Comparison of the 3.7/3.9 µ reflectivities of various objects shows significant differences between the two satellites. Highest differences were observed for storm cloud tops, while for surface targets the differences were somewhat smaller. These differences likely result from variations in viewing geometry and relative solar angle, although differing resolution and center wavelength of these channels can also be factors. This subject is planned for future study utilizing simultaneous observations from GOES-8 and GOES-9 to provide some insight on magnitude of bidirectional scattering effects at cloud tops.

3.3. STORM CLOUD TOPS IN THE 3.9 µ GOES-8 CHANNEL 2

Seven GOES-8 data sequences, showing severe storm development, have been examined for the presence of spots or plumes of increased 3.9 µ cloud top reflectivity. As expected, spots or larger areas of increased 3.9 µ reflectivity were found over some of these storms; however, not so clearly as in the AVHRR images for the same time periods. Since most of the investigated cases involved storms that developed on VORTEX operations days, ground observations were available for some of these storms. Based on these observations, one of the first results of this study was rejection of speculations based on earlier European observations, that indicated a possible direct link between hail and increased 3.7 µ reflectivity (Setvák, 1989). Some of the storms that are known to have produced significant hail on VORTEX days, have shown no significant increase of 3.9 µ reflectivity at all. However, many of the storms did produce spots of varying size, persistence, and magnitude. Lifetimes of these spots ranged from few minutes (as determined from 1 minute data scans) to about two hours. The size of these varied from one single GOES-8 channel 2 pixel (4x4km) up to about 20 - 30 km across. Highest recorded 3.9 µ reflectivity was about 0.20, while typical "background" of the anvils was about 0.02 - 0.04 (all at brightness temperatures below 215K). No direct link between the 3.9 µ reflectivity and brightness temperature minima was found, which indicates that overshooting tops can be excluded as preferred areas where the spots of increased 3.9 µ reflectivity occur.

3.4. COMPARISON OF THE 3.9 µ OBSERVATIONS WITH NEXRAD DATA

Given the observation that spots of increased 3.9 µ reflectivity do not develop at any specific location with respect to the brightness temperature field and, therefore, are not linked exclusively to overshooting tops. An attempt to determine their possible sources was performed using NEXRAD radar reflectivity and radial velocity data. Though only four data sets (07 May 1995, 23 May 1995, 02 June 1995 and 08 June 1995) have been examined so far (combining radar and 3.9 µ data), it is obvious now that number of categories of increased 3.9 µ reflectivity is much broader than was suspected before.

Most of the smaller scale spots (up to a few GOES 8 channel 2 pixels) appear above areas with relatively weak radar reflectivity, lasting from a few minutes up to somewhat less than one hour. In cases when spots appear above storms organized in lines, they typically develop on the westward side of a ridge of overshooting tops and later drift westward (storm-relative) into the "stratiform" part of anvil. Nevertheless, even such spots may have a pronounced "core" from which the material seems to spread out into the surrounding area. The generating mechanism of such spots remains uncertain.

Another category of spots appears linked to mesocyclones and cores of high reflectivity or bounded weak echo regions (BWER) aloft. Even for this type of spot, their behavior and co-location with a nearby mesocyclone can vary significantly. For example (Figure_2), on 07 May 1995 a spot of higher 3.9 µ reflectivity (between 0.78 to 0.87, over "background" values at about 0.35 to 0.40) appeared above a mesocyclone at 2145 UTC, which also was the touchdown time of an associated tornado. Since the previous 3.9 µ image that is available from 2130 UTC shows no trace of this spot, an uncertainty up to 15 minutes remains for the time of its onset. Nevertheless, the spot persisted in the anvil for about the next two hours, disappearing after 2330 UTC. In this case, the spot began instantly to drift away from its "parent" cell as determined by radar observations.

A second example of a spot above a mesocyclone occured on 02 June 1995. In this case, the spot first appeared at about 2330 UTC (the mesocyclone was first detected at 2246 UTC). It had a similar appearance at 2345 UTC and then increased in magnitude and size by 0015 UTC, attaining a diameter of about 15 km. The spot remained in close proximity to the mesocyclone during this period. (No GOES-8 image is available for 0000 UTC - thus exact time of this increase is again uncertain). A tornado touchdown was reported at 2300 UTC, followed by 3 more between 2343 and 0000 UTC. There was no trace of the spot on the next image taken at 0045 UTC, and the mesocyclone disappeared around 0030 UTC.

In a third case, 08 June 1995, one of several 3.9 µ spots again appeared to be associated with a mesocyclone, although in this case the mesocyclone did not produce a tornado. This spot persisted for at least 30 minutes from 2045 to 2115 UTC. Most of the other spots above this storm had lifetimes of only 15 - 30 minutes, which made linking their presence to radar-observed structures difficult. They were found nearby strong reflectivity cores, as well as over weaker echo regions. It is not clear whether those spots formed in their observed position or simply drifted there from elsewhere. This stresses the need for 1-min data to resolve the evolution of short-lived features such as these.

Thus, on the basis of satellite imagery alone, it is impossible to distinguish between different types of spots of increased 3.9 µ reflectivity. However, it should be stressed that examination of a significantly larger sample of cases is required to determine the fraction of 3.9 µ spots associated with strong updrafts and/or rotating features.

There might be a possible link between 3.9 µ spots and "stratospheric" cirrus as observed from aircraft flying at anvil top levels (Fujita, 1982). This cirrus, "jumping up" above anvil tops downwind of overshooting towers as these collapse (Fujita's proposed explanation of stratospheric cirrus generation), is likely to be composed of smaller particles than was their original distribution within the anvil top (due to gravitational settling). Although Fujita (1982) has reported the stratospheric cirrus to extend great distances from its source (25 km or more), one of authors of this paper (Setvák) has observed (on 24 May 1996, Alabama, Georgia) similar "jumping cirrus" to be a quite frequent phenomenon on somewhat smaller scale. Many storm tops on that day in that area displayed a "crest" of cirrus, with horizontal cross-section somewhat smaller than the horizontal size of nearby overshooting tops (Figure_3). Persistence of these "crests" was on the order of a few minutes or more (the upper limit was above the time of plane's fly-by). Location of these and their typical duration are close to those of some of the 3.9 µ spots. Perhaps some of the observed 3.9 µ spots could alternatively be attributed to pileus clouds.

3.5. PLUMES ABOVE STORM TOPS

European observations, summarized in (Setvák and Doswell, 1991) and (Levizzani and Setvák, 1996), though based on "snapshots" of the NOAA/AVHRR instrument only, have shown several important characteristics of the plumes that occasionally develop over convective storms' anvils:

• plumes can be found either in the 3.7 µ daytime imagery, or (less often) in visible/near infrared data; not always can they be traced out in all of these channels simultaneously;
• plumes are in most cases vertically separated from top of the anvil (in visible/near IR band they cast shadows on anvil tops); in some cases, however, they seem to be embedded in the anvil;
• typically, plumes are very thin, details of underlying anvil structures can be glimpsed through them in visible/near infrared imagery;
• more frequent are cases when the plume is split into two diverging streams rather then forming a single one;
• source of plumes is often very small, if the plume can be followed up to its source (which means that the source is still active at the time of observation) -- its size is comparable to a AVHRR pixel size (or a few of them);
• once revealed in above-mentioned spectral bands, it is usually impossible or difficult to find a trace of the plume in thermal IR data (AVHRR channels 4 & 5, 10.5 - 12.5 µ);
• source of plumes is always shifted downwind from the coldest tops, magnitude of this shift ranges from few to 10 - 20 km; if embedded IR warm spot is present, source of the plume is located within (or above) it;
• since the length of a plume can exceed several tens of km and at the same time the plume preserves a high degree of smoothness along its length, the source of such plume has to stable over a longer period of time (tens of minutes).

So far the best pronounced plume of increased 3.7 µ reflectivity, found in the U.S. AVHRR data sets, is shown on NOAA 11 images from 26 April 1994, 22:50 UTC (Figure_4) . In the AVHRR channel 3 reflectivity image, the plume exceeds well over the edge of storm's anvil, reaching peak reflectivity values of around 0.05 to 0.07, whereas the anvil's mean "background" is around 0.03 - 0.04. Notice the almost "point like" source of this plume - highest 3.7 µ reflectivity within it reaches 0.106.

Tornadic storms with a distinct plume-like structure and overall high 3.9 µ cloud top reflectivity developed over northeastern Colorado and southwest Nebraska on 22 May 1996 (Figure_5). Observations from GOES-8 and GOES-9 show that the 3.9 µ reflectivity was significantly higher than that of other nearby storms from the very beginning of the storm's onset (for details see Fig._5f and Fig._5g). Also a plume (well pronounced in visible imagery, not as well in the 3.9 µ imagery) started to form soon after the storm appeared. Highest values of the 3.9 µ cloud top reflectivity within the plume were observed at the onset of the plume's formation, the high reflectivity persisted within that part of the plume (being carried away from its source) till sunset. As can be seen from Fig.5f and Fig.5g, the reflectivity values change with time owing to varying scattering conditions. The 3.9 µ reflectivity begins to be noticeably higher about 2 hours before sunset for GOES-8 compared to GOES-9 (due to more favorable forward scattering angles), whereas earlier that afternoon the storm top reflectivity was almost the same for both satellites. (This case is being further studied from various approaches, thus this part will be modified soon.)

4. SCATTERING COMPUTATIONS AND MICROPHYSICAL MODELING

The potential of theoretical and experimental studies on the optical properties of non-spherical atmospheric ice crystals of the last two decades is yet far from being fully exploited. Theoretical solutions and numerical computations of the scattering by arbitrarily oriented non-spherical particles were proposed, among others, by Liou (1972a,b) and Asano and Sato (1980) for ice cylinders and spheroidal particles, respectively. Satellite multispectral techniques for the analysis of cloud microphysics were developed: interesting examples are the works by Ou et al. (1993) and Rao et al. (1995) using either the 3.7 µ channel alone or its combination with the 10.9 µ one. Radiative transfer models were coupled to radiometric observations for the identification of cirrus and stratiform cloud optical properties, which are of great relevance for the radiation budget and global warming issues. We will only references the works of Stephens (1980) on cirrus cloud properties in the infrared, Stone et. al (1990) on thin cirrus clouds in the near infrared and infrared from satellite with simultaneous observations with lidar and lirad, and Kleespies (1995) on marine stratiform cloud in the 3.9 µ channel.

Note that little of the above-mentioned activity has been devoted to convective storms cloud top studies. Scattering computations and radiative transfer theory represent at present the only available "probe" for the identification of the microphysical composition of the 3.7 µ channnel plumes. A radiative model of plumes is presently under construction for the simulation of the 3.7 µ channel response to varying sun-satellite-cloud geometries. Numerical modeling of the storm's structure is needed to investigate the dynamical and microphysical origin of the plumes and their evolution in time. The Wisconsin Dynamical Microphysical Model (WISC-DYMM) (Johnson et al., 1993, 1994) will be used, given its very detailed microphysical parameterization.

5. CONCLUSIONS

Preliminary studies have revealed a veritable "bestiary" of phenomena at the tops of deep convective clouds, including plumes and various types of spots of increased 3.7/3.9 µ reflectivity. These features do not seem to be tied to any particular form of severe weather. At this point, we do not have any definitive explanations for the observations, either. Any effort to understand the meaning of such signatures must include: 1) a study of the radiative transfer properties of storm tops, 2) a multisensor look at the storms that exhibit these features and those which do not, to look for clues about what distinguishes signature producing storms from those that do not, and 3) an accurate knowledge of the weather events the storms have produced.

ACKNOWLEDGEMENTS

This research was supported by the U.S./Czechoslovak Science and Technology Program, grant #94067, the Italian Space Agency (ASI) and the Italian National Council of Research under the Progetto Strategico for the Mesoscale Alpine Programme (MAP. The authors wish to thank Pavel Hampl from CHMI for his software support when processing AVHRR data, Julie Adolphson and Patrick Dills of UCAR/COMET, Jim Purdom and John Weaver of NESDIS/RAMM for providing archived and real-time GOES-8 data.

Asano,S. and M. Sato, 1980: Light scattering by randomly oriented spheroidal particles. Appl.Opt., 19, 962-974.

Fujita T.T., 1982: Principle of stereographic height computations and their and their applications to stratospheric cirrus over severe thunderstorms. J. Meteor.Soc. of Japan, 60, 355-368.

Johnson,D.E., P.K. Wang and J.M. Straka, 1993: Numerical simulations of the 2 August 1981 CCOPE supercell storm with and without ice microphysics. J. Appl. Meteorol., 32, 745-759.

______, ______, ______, 1995: A study of microphysical processes in the 2 August 1981 CCOPE supercell storm. Atmos. Res., 33, 92-123.

Kleespies, T.J., 1995. The retrieval of marine stratiform cloud properties from multiple observations in the 3.9-µ window under conditions of varying solar illumination. J. Appl. Meteorol., 34, 1512-1524.

Levizzani, V., M. Setvák, J. Kracmar, F. Porcu and F. Prodi, 1992: Multisensor Remote Sensing Analysis of Deep Convective Storms' Structure over Continental Europe and Mediterranean. Proceedings, 11 ^th Int. Conf. on Clouds and Precipitation, IAMAP - Montreal, 1075 - 1078.

______, and ______, 1996: Multispectral, high-resolution satellite observations of plumes on top of convective storms. J. Atmos. Sci., 53, 361-369.

Liou, K.-N., 1972a: Electromagnetic scattering by arbitrarily oriented ice cylinders. Appl.Opt., 1, 667-674.

______, 1972b: Light scattering by ice clouds in the visible and infrared: A theoretical study. J. Atmos. Sci., 29, 524-536.

Menzel, W.P. and J.F.W. Purdom, 1994: Introducing GOES-I: The first of new generation of Geostationary Operational Environmental Satellites. Bull. Amer. Meteor. Soc., 75, 757-781.

Ou, S.C., K.-N. Liou, W.M. Gooch and Y.Takano, 1993: Remote sensing of cirrus cloud parameters using advanced very-high-resolution radiometer 3.7- and 10.9-µ channels. Appl.Opt., 32, 2171-2180.

Rao, N.X., S.C. Ou and K.N. Liou, 1995: Removal of the solar component in AVHRR 3.7-µ radiances for the retrieval of cirrus cloud parameters. J. Appl. Meteorol., 34, 482-499.

Rasmussen, E.N., J.M. Straka, R. Davies-Jones, C.A. Doswell III, F.H. Carr, M.D. Eilts and D.R. MacGorman, 1994: Verification of the Origins of Rotation in Tornadoes EXperiment (VORTEX). Bull. Amer. Meteor. Soc., 75, 995-1006.

Setvák, M., 1989: Convective storms - the AVHRR channel 3 cloud top reflectivity as a consequence of internal processes. Proceedings, 5^th WMO Conf. on Weather Modification and Applied Cloud Physics, Beijing, WMO TD - No.269, 109-112.

______, and C.A. Doswell III, 1991: The AVHRR channel 3 cloud top reflectivity of convective storms. Mon. Wea. Rev., 119, 841-847.

Stephens, G.L., 1980: Radiative properties of cirrus clouds in the infrared region. J. Atmos. Sci., 37, 435-446.

Stone, R.S., G.L. Stephens, C.M.R. Platt, and S. Banks, 1990: The remote sensing of thin cirrus cloud using satellites, lidar and radiative transfer theory. J. Appl. Meteorol., 29, 353-366.