Polarimetry and Radiometry of the Special Sensor Microwave Imager (SSMI) Written by Paul J. McCrone LT, US Navy HQ Air Force Global Weather Central Product Improvment Branch Satellite Applications Section (AFGWC/DOAS) September 25, 1995 ABSTRACT Currently, advanced microwave sensors aboard meteorological satellites are providing data about the atmosphere on an unparalleled level. Aboard spacecraft of the Defense Meteorological Satellite Program (DMSP), there are three such microwave sensors . This paper will focus on the newest of these sensors: the Special Sensor Microwave Imager (SSMI). The intent of this paper is to provide the reader with a basic overview of the sensor, including some of the physics behind the design of the sensor. INTRODUCTION All meteorologists are intimately familiar with the visual and infra-red data from such conventional satellites as the US NOAA TIROS, GOES series, European METEOSAT, and Japanese GMS. The GOES, METEOSAT, and GMS are all GEOSTATIONARY satellites: as they orbit the earth, they maintain same location over the earth (or nadir ) at all times. They orbit directly over the equator, and have a velocity that allows them to maintain the same nadir continuously. This is a great advantage, since a person can readily put images from this type of spacecraft together and form an animation (using a system like McIdas), since it always has the same field of view (FOV). By contrast, the NOAA-TIROS satellite roughly orbits the earth from the North pole to the South pole about 14 times or so a day. The satellite orbit is referred to as SUN-SYNCHRONOUS- meaning that the plane of the satellite orbit maintains a constant angle with respect to a plane containing the earth's rotational axis and a line drawn from the center of the earth to the sun. While maintaining this orbit, the field of view continuously changes. While an image animation is more difficult with this type of satellite, much higher resolution imagery is obtained. This discussion will focus on the Polar orbiting, Sun Synchronous satellites, and in particular , a type of spacecraft of the US Department of Defense, the Defense Meteorological Satellite Program (DMSP). Aboard DMSP spacecraft, there are three microwave sensors: the Passive Microwave Temperature Sounder (SSM/T or SSM/T-1), the Passive Microwave Temperature Sounder [Water Vapor] 2 (SSM/T-2), and the Special Sensor Microwave Imager (SSM/I or SSMI). The SSM/T-1 and SSM/T-2 are primarily focused on the retrieval of temperatures in the upper troposphere / lower stratosphere. The SSMI is focused on the retrieval of microwave energy emitted from the surface of earth. The main thrust of this paper is to discuss the SSMI and explain its functionality, design, and applications. A brief overview of the DMSP satellite will be provided, then a description of the SSMI will follow. Each of the sensor attributes will be explained in some detail, followed by a look at the environmental products that can be derived from the SSMI. I. THE DMSP SATELLITE The DMSP provides environmental data too both DoD and civilian organizations world wide. The satellites are launched into nearly polar, sun- synchronous orbits and phased so that they pass over a given point on the earth at the same local time each day. See Figure 1 for an example of a DMSP spacecraft. A. General Information The DMSP spacecraft has numerous sensors aboard, chief among them is the Operational Linescan System (OLS), a visible and infrared sensor capable of producing imagery in two different resolutions: SMOOTH ( 1.5 nautical miles) and FINE (0.3 nm [nautical miles]). The spacecraft flies in a sideways orientation, with the longer axis perpendicular to the direction of motion. As the spacecraft continues, the OLS scans from left to right of the track (perpendicular to the direction of motion), similar to the sensors on the NOAA TIROS. In order to view the earth at all times, a constant pitch rate is induced on the satellite, making the OLS face down toward earth. Flying at an altitude of 833 km, the DMSP satellites take 102 minutes to orbit the earth, resulting in ~ 14 orbits every day. The actual orbit of the DMSP is at an inclination angle of 98.8 degrees, causing the satellite to just miss the polar caps. Normally, there are at least two DMSP satellites on orbit at all times, and sometimes more. At this writing, two of the satellites, F11 and F13, are still fully operational , with the remaining satellites (F10,F12) in a partially operational mode. The DMSP system is actually more than just the spacecraft : an extensive network of ground receiving and relay stations enable personnel at the Air Force Global Weather Central (AFGWC) to download the entire collection of imagery and data coming from the spacecraft. This data is stored in an operational environment at AFGWC to support DoD weather requirements , and , is also stored at the National Geophysical Data Center (NGDC) in Boulder, CO, and is available ( in part ) via the internet (see the following address on the World Wide Web: http://web.ngdc.noaa.gov/dmsp/dmsp.html). Finally, another significant player in the processing of SSMI data is the U.S. Navy's Fleet Numerical Meteorology and Oceanography Center (FNMOC) in Monterey , CA, where some of the products from the SSMI are implemented into the Naval Operational Global Atmospheric Prediction System (NOGAPS), a global numerical weather prediction (NWP) model, similar to the Medium Range Forecast (MRF) model of NOAA's National Meteorological Center (NMC). B. Special Sensor Microwave Imager (SSMI) The SSMI is a conical-scanning, multi-channel, multi-polarization microwave radiometer aboard DMSP spacecraft. The SSMI can be seen in figures 2 and 3. Figure 2 shows the sensor by itself, and figure 3 shows where the sensor is located on the bus (or structure ) of the satellite. As demonstrated in figure 4, the sensor rotates axially through 360 degrees. However, it only actually scans through 102 degrees of the rotation in the back of the satellite. This is because of (1) an electronic need to recalibrate the SSMI every time around, and (2) because of an unavoidable physical obstruction to the sensors' field of view caused by the bus of the spacecraft itself. This 102 degree limitation is important because it cuts down the amount of area the sensor ' could cover ( SSMI has half the spatial coverage of the DMSP OLS). 1. Sensor Channels and Polarizations The SSMI has four frequencies: 19.35 GigaHertz (GHz), 22.235 GHz, 37.0 GHz, and 85.5 GHz. The sensor is a passive suite: it only measures the radiation coming up from the earth. It does not send out a signal , such as a RADAR. The 19, 37, and 85 GHz frequencies are split into two channels each : one channel detects vertically polarized microwave energy, while the other receives horizontally polarized energy. The 22 GHz frequency only has a vertically polarized channel. Thus, the SSMI has seven total channels, summarized below in Table 1. ______________________________________________________________ Channel Frequency (GHz) Resolution(km) Bandwidth (MHz) Wavelength(mm) 19V 19.35 50 240 15.5 19H 19.35 50 240 15.5 22V 22.235 50 240 13.5 37V 37.0 25 900 8.1 37H 37.0 25 900 8.1 85V 85.5 13 1400 3.5 85H 85.5 13 1400 3.5 ______________________________________________________________ Table 1: Channels of the SSMI Note: 'V' means 'Vertical' and 'H' means 'Horizontal' Polarization The sensitivity of these channels ranges from +/- 0.37 degrees K to 0.73 K. The SSMI channels measure a quantity known as brightness temperature . Brightness temperature, or Tb, is the radiometric temperature of the earth. It is directly related to the actual temperature, via equation (1): Tb = e T (1) where T is the Temperature, and e is the emissivity of the earth surface. Emissivity is a relative, dimensionless measure than varies from 0 to 1. Objects with a low emissivity (such as water ) will appear colder to the SSMI than they actually are. Conversely, objects with a high emissivity will appear closer to the actual temperature. In the ideal case, an object with a emissivity of 1 (referred to as a Planck blackbody) would emit a brightness temperature which exactly equalled the surface temperature. This is a perfect emitter. With this in mind , one can say identify two primary functions of the SSMI : it conducts radiometry (the measure of radiant energy) and polarimetry (the measure of polarization properties of a scene) of the earth's surface. 2. Scan Geometry The SSMI has a scanning mechanism that consists of a broad-band, corrugated, offset parabolic reflector. This is the larger disc shaped object sitting atop the sensor in figure 2. The microwave reflector is an ellipse, 61 X 66 cm, with a 45 degree down angle (nadir angle). The projection of the scan spot , or 'shadow' on the earth will be an approximate 51 degree incidence angle (See figures 5 and 6) . Of course, the sensor itself is strictly passive: it merely receives upwelling radiation and detects it. No microwave signal is sent out from the spacecraft. Thus, when one refers to the 'projection' or 'shadow' of the antenna, one is merely referring to the area on the earth that the sensor is truly 'looking' at. The sensor rotates at 31.6 revolutions per minute, with a 1.9 second period and sweeps a swath 1395 km wide. II. SENSOR DESIGN PRINCIPLES The previous section gave useful information regarding the functionality of the SSMI. However, no information was given as to the purpose for the sensor to be designed this way. The design of the sensor will be discussed in the following section. The SSMI design is deeply rooted in radiative physics. Thus, some time will be spent discussing the physics behind the SSMI. In general, there are numerous processes we are concerned with in atmospheric radiation such as reflection, scattering, absorption, and emission. With respect to each of these, there can be ( and often is ) great dependence on frequency (and, thus, wavelength), viewing angle, emissivity, temperature, size and type of material, etc. The resultant path traversed by the radiation can be complex indeed. These different paths are representative of different interactions that the microwave energy will have with mass (such as water vapor, ice, oxygen, etc.). Figure 7 shows the significant microwave interactions associated with the four SSMI frequencies. Again, in general, we can characterize the radiation using the concept of a Planck blackbody mentioned earlier. Ideally, a blackbody can be represented by equation (2) below, known as the Planck blackbody function: 2phc2 B(l,T)= __________________ (2) l5 [exp(ch/lkT)-1] Where B = Blackbody brightness (Watt /m2 sr) l = Wavelength h = Planck's constant k = Boltzmann constant c = Speed of light T = absolute temperature (degrees Kelvin) In the microwave region, the quantity (ch/lkT) << 1 . Therefore, an approximation can be made to equation (2) below: 2pckT B(l,T)= __________________ (3) l4 Equation (3) is the Rayleigh-Jeans Approximation to the Planck blackbody formula. It is the functional equation for applications with the SSMI. A. The Purpose of each channel The SSMI has seven channels, but only four distinct frequencies: the 19, 22, 37, and 85 GHz channels. The reason for each frequency is based on the absorption, emission, and scattering properties of the earth's surface and atmosphere. The radiometric aspects of the SSMI were chosen to accomplish specific objectives. This absorption of microwave energy is displayed in figure 7 (courtesy of Aerojet Electro Systems). Looking at this curve, note the 22 GHz region. Here, there is a relative maximum in the absorption. This is caused by water vapor in the atmosphere. Thus , the 22V channel was chosen to provide an estimate of water vapor in the atmosphere. The 85 GHz channels were chosen to provide an estimate of the precipitation from a given cloud complex. The following excerpt from Fitzpatrick [1994] provides a discussion of this phenomenon: "Large raindrops and ice particles scatter microwave radiation, effectively lowering the background brightness temperature by scattering emitted radiation away from the satellite. This effect is most pronounced at high frequencies at which the particle size becomes comparable to the wavelength. This lowering of the brightness temperature due to scattering by precipitation- sized particles is equally effective over water and land. This relationship enables the 85.5 GHz observations from the SSM/I to play a key role in identifying precipitation." Thus , the 85 GHz channels take advantage of the scattering properties of the raindrops and ice particles in the atmosphere, since at a wavelength of 3.5 mm, the radiation is close to the size of water droplets and ice particles. For the 19 GHz channels, the intention was to find a regime that would permit radiation from the surface of the earth to pass directly to the sensor. If the 85 GHz (3.5 mm) channel was close to the size of water droplets, then the 19 GHz (15.5 mm) would be an order of magnitude larger than these species, and as a direct result, the least affected by the water vapor/droplets/ ice. Instead, the 19 GHz region is sensitive to the characteristics of the earth's surface, such as ocean surface roughness, land surface moisture, land type, etc. Because of this feature, the 19 GHz is a phenomenal channel for a variety of applications: Ocean surface wind speed ( to be discussed later), presence of snow / ice on the surface, detection of flooded regions, and many others. The 19 GHz channel is perhaps the most useful, flexible, and exciting channel on the SSMI. The 37 GHz was selected for the most part as a direct result of previous experience from earlier NASA microwave sensors flown aboard satellites like the Nimbus series of the 1970's. Primarly, the 37 GHz channels are a good 'midpoint' between the 19 GHz and 85 GHz. These channels are insensitive to some atmospheric species like water vapor and oxygen, yet they are sensitive to rain, cloud water content, and surface ice cover. Because of these factors, the 37 GHz provides a good detector to help identify regions where selected atmospheric phenomena might degrade the determination of numerous environmental products, such as ocean surface wind speeds. B. The Purpose of each Polarization Having discussed the criteria used to select the frquencies of the SSMI, the discussion will now turn to the utility of the cross polarization channels at 19, 37, and 85 GHz. The manner in which we measure radiation might depend not only on the phenomena which emits it, but also on the spatial orientation in which it is emitted. Basically, the emissivity of an object or surface may be a function of polarization of the radiation. As stated in Grant [1991], "Measurement of the amount of polarization of the emitted can be performed by using two channels of the same frequency, one of which detects vertically polarized energy while the other detects horizontally polarized energy. Usually an EM [electromagnetic] wave is not polarized purely in the vertical or horizontal plane. In fact, it may not be polarized at all. In this case, both the vertical channel and the horizontal channel will each detect part of the energy from the wave. If the wave is closer to being vertically polarized than horizontally polarized, then the vertical channel will detect the most energy. The difference between the brightness of an object on the vertical channel and in he horizontal channel provides useful infor- -mation about it." There are four types of electromagnetic [EM] polarization that we are concerned with: Random Polarization, Linear Polarization, Circular Polarization, and Elliptical Polarization (see examples in figure 8). In random polarization, there is no preferred sense of polarization. In linear polarization, the E vector remains in a single plane. Two special cases of linear polarization are the horizontal and vertical polarization. "Horizontal polarization is .. defined as the state where the electric vector is perpendicular to the plane of incidence. The vertical polarization corresponds to the case where the electric vector is in the plane of incidence. " [Elachi, 1987]. Elliptical polarization is a case where two waves of the same frequency , different phase, different amplitude, and different polarization combine into a single wave. The result is the vector addition of the two E vectors. The combined E vector will seem to rotate in the pattern of an ellipse about the vector forming the direction of motion. Circular polarization is just a special case of elliptical polarization, where all the above holds true, except that the amplitudes are now both the same. The final electric field vector [E] precesses around the axis of motion in a circle. The emissions at 22 GHz come from atmospheric water vapor. Such emissions are usually randomly polarized, so it is not necesary to measure the horizontal polarization. This is the reason for only one vertical channel at 22 GHz on the SSMI. The other channels will experience different phenomenon that make polarimetry important. For example, Sea ice exhibits some distinctive differences in polarization at 85 GHz. Desert regions also display polarization differences at many frequencies. The work of Barrett [1990] observed: "Polarization of reflected, transmitted and emitted PM [passive microwave] radiation depends on molecular or cystalline properties and the surface roughness of a medium. Like emissivity, it is a function of angle of view. Water, having a high dielectric constant produces highly polarized radiation at oblique viewing angles : for example, at 37 GHz , calculated brightness temperature differences between horizontally and vertically polarized radiation emerging from a standard atmosphere exceeds 30K.... Radiation emitted by atmospheric gases is not polarized, and there is no significant polarization in either cloud drops or raindrops. Differences in polarization are therefore a primary basis for distinguishing between dry land, wet land, and rain over land in some PM rainfall estimation schemes." One item to note is the fact that emissivity for water is very different from the vertical (0.6) to the horizontal (0.4) [Negri et al, 1989]. This can be used to detect changes in the sea surface due to the wind blowing over it. C. The Purpose of the scan geometry Related to the issue of polarization is the concern over the 'look angle' or scanning geometry of the SSMI. The basic problem is that a direct downward (nadir viewing ) scan method will not observe any significant difference in polarization. Elachi [1987] stated that '.... at an incidence angle of 37 degrees from vertical, an optical wave polarized perpendicular to the plane of incidence will reflect about 7.8% of its energy from a smooth water surface, while an optical wave polarized in the plane of incidence will not reflect any energy from the same surface. All the energy will penetrate into the water. This is the Brewster effect." In general, "... lower incidence angles show much less difference between the horizontal and vertical polarizations of a channel than higher ones [See figure 9]. At nadir, where the incidence angle is 0 degrees, the polarization difference disappears entirely" [Grant, 1991]. The only remaining issue then is to pick a scanning angle that optimizes the 19, 37 , and 85 GHz channels. Rough inspection of figure 9 will show the reader the rationale behind the selection of 51-53 degrees. As mentioned earlier, the conical scanning method with constant incidence angle makes the sensor 'footprint' on the surface a constant shape and resolution. This is a significant advantage over the DMSP OLS and NOAA AVHRR, since both those sensors sweep across nadir from left to right. III. PRODUCTS AVAILABLE FROM SSMI The SSMI is a robust sensor , capable of producing numerous products. As discussed earlier with equation (1), there is a direct relationship between the Tb of the SSMI and the actual earth's surface temperature. A general list of products derived from the SSMI follows: Ocean Surface Wind Speed Sea Ice Concentration, Edge, and age Precipitation Rate (Over land and ocean) Liquid Water Content ( Ocean and Land) Cloud Water Content ( Over ocean, land, ice, and snow) Atmospheric Vapor Content (Ocean ONLY) Surface Moisture over land (except heavy vegetation) Surface Temperature (many surfaces) Snow Water Content, Egde Cloud amount (Land and snow) Surface Characteristics (type) and others. Any attempt to fully discuss these products ( referred to as Environmental Data Records, or EDRs) would be nearly impossible. The author will attempt only to give a flavor of these EDRs and how they are used. One of the other uses of the SSMI is tropical cyclone reconnaissance. Looking at figure (10), one sees a tropical cyclone in the North Arabian Sea. The system exhibits a convective Central Dense Overcast (CDO) , blocking a clear view of the cyclones center. Figure (11) shows the same DMSP pass with the 85H channel from the SSMI. Note that the center of the cyclone is clearly evident. Surface winds are derived using four channels: 19V, 22V, 37V , and 37H. The primary channel is the 19V: as wind blows on the water, sea foam is generated. This foam has a high emissivity, and is detectable at 19V. One problem remains though: foam can be generated by a passing rainstorm: a convective rain will certainly stir up the water above its normal level, even if the winds a reasonbly calm. Thus, the potential for erros exists. Using the 37V/H channels both as a screen and as weighting factors in the wind speed retrieval algorithm, these errors can be eliminated, reduced, or at least identified. The primary way that the 37 GHz channels are used to screen the data is to take the 37 differential (D37 = 37V - 37H) and use it to assign weighting factors to the wind speed data using this D37. Generally, the higher values of D37 indicate greater accuracy ( and confidence, statistically) of the wind speed values. The current algorithm for surface winds is given below from Grant [1991] in equation (4): SW = 147.90 + (1.0969) Tb19V - (0.4555) Tb22V -(1.76)Tb37V + (0.786)Tb37H (4) For a complete discussion on the surface wind speed algorithm, see the work of Grant [1991] and Hollinger et al [1989]. See figure 13 for an example of the output associated with this data. Other examples of this data are also being made available through NGDC on the internet, and also by AFGWC (see http://afgwctst1.offutt.af.mil/ssmi.htm). Most of the algorithms in use today for SSMI EDRs are simply linear regression models, such as equation (4) . Robust algorithms on a global scale are difficult to derive, since the surface characteristics of the earth are so variable. One source of new research is the use of computer neural networks (NN) in EDR computation. These NN algorithms, though non-linear, are still regression oriented. At AFGWC, these new algorithms are being tested. The results are displayed in figure 14. The figure shows several different experimental surface wind algorithms. Bear in mind that colors of red are 30 knots or greater: orange, yellow and blue are all less that 30 knots. Notice the poor performance of the neural network algorithm on the bottom left. This is meant to demonstrate that regression-type algorithms are simply not 'getting the job done ' on a global scale (note : this same NN algorithm performs quite well in the extratropics). IV. CONCLUDING COMMENTS The SSMI is a tremendous new sensor that will change the way forecasters do their job in the next decade. Already, data from the SSMI EDRs are being used in the analysis fields of NWP models with NOAA and the US Navy. SSMI is being used to track tropical cyclones in remote , data sparse areas. The future is bright for this sensor. In the future, work will be needed to improve the accuracy of the EDRs that already exist, in addition to new algorithms for new EDRs being tested in the research community. However, one area of growing concern is the utilization of algorithms derived by linear regression. These methods are simply lacking the ability to determine true environmental conditions on a global scale. There is a need for better algorithms based on radiative physics, the true 'nuts and bolts' of the sensor in the first place. BIBLIOGRAPHY Barrett, Eric C., 1990: Passive Microwave Satellite Imagery for Improved Rainfall Monitoring and Forecasting over Sea Areas. Microwave Remote Sensing for Oceanographic and Marine Weather - Forecast Models. Kluwer Academic Publishers, 101 Philip Drive, Norwell, MA 02061 (USA). pp 239-252. Elachi, Charles, 1987: Introduction to the Physics and Techniques of Remote Sensing. John Wiley & Sons, New York. Fitzpatrick, Michael, 1994: Special Sensor Microwave Imager (SSM/I) Tropical Training Guide. Headquarters, Air Force Global Weather Central, 106 Peacekeeper Dr. STE 2N3, (Code : DOAS) Offutt AFB, NE 68113-4039. Glackin, David L. and Samuel Gasster, 1993: Earth Remote Sensing: Electro- optical and Microwave Remote Sensing. The Aerospace Corporation. Los Angeles, CA. Goody, R.M., 1964: Atmospheric Radiation: I Theoretical Basis. Oxford University Press, Amen House, London UK. Grant, Kerry and John Heinrichs, 1991: Special Sensor Microwave Imager (SSM/I) User's Interpretation Guide. Dept. of the Air Force, Headquarters, Space Division (AFSD) Los Angeles AFB, CA. (Contract No. F04701-87-0016). Hollinger, James P. et al, 1989: DMSP Special Sensor Microwave / Imager Calibration/Validation, Final Report , Volume I. Space Sensing Branch, Naval Research Laboratory, Washington DC 20375-5000. Negri, Andrew J., Robert F. Adler, Christian D. Kummerow, 1989: False- Color Display of Special Sensor Microwave/Imager (SSM/I) Data. Bulletin of the American Meteorological Society, Vol.70, No.2 February 1989, pp.146-51. Pedersen, L.T., 1990: Microwave Radiometers. Microwave Remote Sensing for Oceanographic and Marine Weather - Forecast Models. Kluwer Academic Publishers, 101 Philip Drive, Norwell, MA 02061 (USA). pp 177-190. Rao, P. Krishna, Susan J. Holmes, Ralph K. Anderson, Jay S. Winston, and Paul E. Lehr, 1990: Weather Satellites: Systems, Data, and Environmental Applications. American Meteorological Society, 45 Beacon St. Boston MA 02108. Savage, Richard C., 1976: The transfer of thermal microwaves through hydrometeors (Ph.D Thesis). University of Wisconsin-Madison. g Branch, Naval Research Laboratory, Washington DC 20375-5000. Negri, Andrew J., Robert F. Adler, Christian D.