Western Pacific Moisture Analysi� as Observed from DMSP SSM/I Measurements

DMSP SSM/I data are used in this study to investigate the global dis· tribution of moisture, and to study the variations of brightness temperatures with the changes of the intensity of typhoons during September 1987. The rain rate of typhoons is also determined based on Olson et aL's (1990) algo­ rithm. It is noted that the SSM/I data could provide realistic patterns and magnitudes of total precipitable water, cloud water content, and rain rate for the globe. They are comparable to the previous findings and other model analysis. The characteristics of SSM/I data are documented in this study for the three typhoons (Gerald, Freda , and Holly) over the western Pacific Ocean. The detailed structure of the typhoon can be identified by the SSMJI data, and the estimated rainfall of the typhoon appears to be reasonable. The development of the algorithms to derive atmospheric variables for Taiwan and its vicinity is suggested, and the validation of the algorithms has to be executed based on radar. data, and upper and surface observations. The intercomparions of various algorithms also can be performed, if ground truth data are not available.

estimated atmospheric and surface variables , such as total precipitable water, cloud (liquid) water content, rainfall, surface wind speed, and surface temperature.Several specialized field experiments have also been conducted to validate the algorithms that determine the values of these variables.Previous studies have shown that the passive microwave method could provide realistic global patterns •Of moisture.Researchers have also found that the intensity of tropical cyclones and evolution of midlatitude cyclones can be determined accurately by using various algorithms based on microwave data.
The higher microwave frequencies with better resolution aboard satellites were not available until the launch of DMSP (Defense Meteorological Satellite Program) F8 satellite on June 19,1987; that was equipped with the Special Sensor Microwave/Imager (SSMII).
SSM/I offers a data set of unprecedented quality and temporal coverage to study global distribution of moisture and tropical cyclones.
The present study's purpose is to use DMSP FS SSM/I data to investigate the global distribution of moisture, and to study the intensity and rainfall of three typhoons over the western Pacific Ocean during September 1987.The objective is to promote the understand ing of the characteristics of microwave data, and to initialize the future research on the applications of microwave data.

PA SSIVE MICROWAVE MEASUREMENTS
At microwave frequencies, remote sensing instrument can "see" through clouds to ob serve surface features with rain (large liquid and ice hydrometers) being the major source of variations; the surface features (such as sea ice, snow cover, ice type, surface wind speed, roughness of ocean surface, vegetation covers, and others) can be determined adequately and accurately.Most moisture variables in the atmosphere, such as total precipitable water, cloud (liquid) water content, and rainfall intensity can also be derived using various algo rithms.However, most atmospheric variables can only be determined over oceans, where the background brightness temperatures are nearly constant.The low emissive brightness temperatures are distinct from the brightness temperatures of clouds and precipitation.
For the microwave frequencies bel .ow 37 Ghz, rain may be evidenced by the greater brightness temperature due to increased emission of the liquid and ice hydrorneteors; while above 60 GHz, rain can be detected by decreased brightness temperature due to the scattering by ice hydrometeors.At frequencies between 37 and 60 GHz, the combination of both emission and scattering prevails.
Large raindrops and ice particles, such as graupel or hail, 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 hydrometeors are equally effective over water and land.More detailed review on the passive microwave measurements is discussed in Huang and Liu (1992).
The relationship, though indirect, between this scattering process and surface rain rate, enables the SSM/I 85.5 GHz observations to play a key role in identifying precipitation.This capability has already been shown by using aircraft microwave data at 92 GHz (Wilheit et al., 1982� Harkarinen and Adler, 1988� Heymsfield and Fulton, 1988), by analysis of coupled radiative transfer/numerical cloud models (Adler et al., 1988), and by analysis of SSM/I data (Spencer et al., 1989).This relationship can also be seen in the calculations of Wu and Weinman (1984).The orbit provides complete coverage of the earth, except two small circular sectors of 2.4° centered on the North and South poles.Table 1 lists the orbital parameters of DMSP.
The DMSP SSMJI has actually 7 separate total-power radiometers, each simultaneously measuring the microwave emission coming from the earth and the intervening atmosphere.Dual polarization measurements are taken at 19.35, 37.0, and 85.5 GHz, and only vertical polarization is observed at the 22.235 GHz water vapor channel.The spatial resolutions of these chann els vary from about 15 km (85.5 GHz) to 60 km (19.35 GHz) depending on the frequency.Table 2 summarizes the characteristics of SSM/I.
Each observation is taken during a 102.4° segment of the rotation when the SSM/I is looking in the aft direction, as is shown in Figure 1.The 102.4 ° arc is centered on the spacecraft subtrack and corresponds to a 1394 km wide swath on the earth's surface.
During each scan, the 85.5 GHz channels are sampled 128 times over the 102.4° arc.The integration period for a single sample is 3.89 ms.This sampling scheme results in 128 vertical polarization footprints and 128 horizontal polarization footprints having an effective 3-dB resolution of about 15 km for 85.5 GHz channels.A more detailed description of the SSM/I can be found in Hollinger et al. (1987) and Huang and Liu (1992).
The coverage of SSM/I is far better than its predecessor, SMMR (Scanning Multichannel Microwave Radiometer) on Nimbus-7 and Seasat satellites.The mosaics of all 14 ascending and descending swaths in one day are shown in Figure 2. It should be noted that these swaths are asynoptic (not synoptic).Each swath is 101.42 minutes before or after the neighboring swaths, and the last swath of the day can be almost 23.5 hours after the time of the first swath of the day.The first swath of the day is to the west of the last swath of the day by 25.4° longitudes.The diamond-shaped gaps in one day are centered at about 20°N and 20°S.This coverage is adequate for global-and synoptic-scale studies, especially as the average of a few days is desired.Nevertheless, the temporal resolution may not be sufficient for meso-scale research.
On the average the intersections of ascending and descending nodes are about 780 km apart; they become much closer together as latitude increases, until they are 220 km apart for the most poleward intersections.This pattern migrates westward and nearly repeats at six-day intervals.The 1394 km swath width gives daily coverage of the equator, double coverage (i.e., twice daily) for half of the subtropics, and quadrupled coverage polewardof about 60°N and 60°S.

GLOBAL DISTRIBUTION OF MOISTURE .
The global distribution of moisture is studied for August 31-Sep tember 27, 1987.This period is further divided into two 14-day periods,•namely, August 31-September 13, and September 14-27, 1987, respectively.The estimated variables are the total precipitable water, cloud (liquid) water content (both are based on the algorithms of Alishouse, 1983), and rain rate as estimated by Olson et al. (1990).
In the first period (August 31-September 13), three typhoons (Gerald, Freda, and Holly) occurred almost at the same time over the western Pacific Ocean.The detailed analysis of these typhoons will be discussed in Sections 5 and 6.In the second period (September 14-27), it represents a tranquil state over the western Pacific Ocean.
The SSM/I data obtained from the NASA WetNet program are in image format with brightness temperatures at pix.els.Each global composite image is 640 (called elements or columns) by 320 (called lines or rows) pixels, and full resolution swath of brightness temperatures 'is 128 (elements or columns) by 1610 (lines or rows) pixels.The image data are processed in the PC-McIDAS (Personal Computer-Man-computer Interactive Data Access system) software package, that was developed by the Space Science and Engineering Center of the University of Wisconsin-Madison, Wisconsin.The digitalized pixel values are transferred to the VAX microcomputer at the Center for Space and Remote Sensing Research (CSRSR), National Central University to perform statistical analysis, because PC-McIDAS limits the pixel values to between 0 and 255.All negative values will be considered as zero, and all values more than 255 become 255.

•
The average of variables of all the ascending and descending swaths for the studied period yields the mean state.The standard deviation is the square root of sum of squares of deviation (differ enc e between daily values and mean state) divided by number of occurrences minus 1.It r epresents daily variations of the weather systems.The final products are converted back to PC-McIDAS for displaying and checking, and eventually to ERDAS image processing software at CSRSR for making hard copies.
Figure 3 shows the mean and standard deviation of total precipitable water of the first period between 60°S and 80°N.Tr opical moisture in the Pacific maximizes west of the dateline, and extends along the equator into the Indian Ocean, Atlantic Ocean, and the eastern Pacific Ocean.This high concentration of moisture coincides with the ITCZ (InterTropical Convergence Zone).The extension of moisture maximum intruding from the • east of dateline into the Southern Hemispheric midlatitudes characterizes the SPCZ (South Pacific Convergence Zone), whose counterpart over the South Atlantic Ocean, the SACZ (South Atlantic Convergence Zone), is inactive during this period.The drier air over the high latitudes is noted for its cold temperature.
By comparing Figure 3 to the monthly means of total precipitable water of Septem ber 1987 as analyzed by NMC (National Meteorological Center) and ECMWF (European Centre for Medium-range Weather Forecasts) (see Figure 4), we found striking quantitative similarities.SSM/I estimates, however, show ri1ore details than those of NMC and ECMWF.
The moisture variability at the time scale of transient features can be studied in the standing deviation of daily SSM/I to the 2•week mean.Transient activity maximizing east of the midlatitudinal continents clearly identifies middle latitude storm tracks.The largest variability exists east of Japan, but significant fluctuations extend across the north Pacific to 150°W.The presence of synoptic transients in the SPCZ and SACZ is also apparent.The high variability over northern Australia signifies the migrating cyclones in this region.As for the local maximum off Baja, Califonnia, it is possibly caused by the variab ilities of fog and low stratiform clouds in this region.The small variability of moisture over the ITCZ is also noted as expected.Pacific Ocean.The migrating extratropical cyclones, on the other hand, do not yield large rain rate due to their transient features and weaker convection.Local maxima associated with these cyclones, however, can still be recognized in both mean and standard deviation maps.
It should be also noted that difference color enhancement schemes are used in maps of mean state and standard deviation.The magnitudes of standard deviation are almost twice as large as that of mean state.
The mean and standard deviation of total precipitable water, cloud (liquid) water con tent, and rain rate for the second period (September 14-27) are presented in Figures 7-9.
Basically they are similar to those of the first period (August 31-September 13) in pattern and magnitudes.The notable differences are the lower values over the western Pacific Ocean, and more active convection over the India Ocean.The SPCZ is also more pronounced dur ing this period.The maximum variation of the eastern coast of Japan is possibly due to the transient features of extratropical cyclones.SSM/I data consist of ascending and descending swaths, that represent morning passes (6:12 am crossing-equator local time) and evening passes (6:12 pm crossing-equator local time).It is of interest to investigate the diurnal variation of convection and rainfall as suggested by Albright et aL (1985), Hartman and Recker (1986), and Petty and Katsaros (1992).Figures 10 and 11 shows the mean rain rate of morning (upper panels) and evening (lower panels) of the first and second periods, respectively.The main features of convection are similar, but different in magnitudes.It appears to confirm a tendency toward larger areal extent and greater mean intensity of oceanic rainfall in the morning than in the evening, over much of the world's oceans.This finding is consistent with other research (e.g .. Petty and Katsaros,1992). •

Brightness temperatures of typhoons
The full resolution swaths (128 by 1610 pixels) of brightness temperatures of SSM/I frequencies are first chosen, if they are in the studied period of typhoons (September 4-17, 1987) and fall within the studied domain (9.5°-37°N, and 105°-162.5°E).The daily mosaics of the ascending swaths and descending swaths are produced separately in PC-MclDAS.All the pixels values in the mosaics are converted to the VAX microcomputer for computing rain rates.Although all the frequencies of both vertical and horizontal polarization are grocessed, only the brightness temperatures of vertical polarization will be presented.T he 8 lat.(95 lines) by 8° long.(89 elements) box centered at the center of the Typhoon is selected in the Cartesian coordinate system .The brightness temperatures in this box are used to study the intensity of the typhoons.
During the studied period (August 31-September 27, 1987), four typhoons (Gerald, Freda, Holly, and Ian) and one tropical depression (17 W) occurred over the western Pacific Ocean (Hoffman et al.,Anhual Tropical Cyclone Report, 1987).Only the first three typhoons were studied because they were completely covered in this period (Typhoon Ian occurred during September 23-0ctober 1,1987).The tracks of Gerald, Freda, and Holly are shown in Figure 12.
Typhoon Gerald (September 4-19, 1987) matured within the monsoon trough, but did not detach from it.Typhoon Gerald, having a unusually large eye, moved northwestward from the east of Philippines, sweeping the south of Taiwan, then crossing over Formosa Strait, and it finally dissipated in the Fujian Province of China • Typhoon Freda (September 4-17, 1987) was the first tropical cyclone during September and was the middle (geographically) of a three-storm situation (the other tropical cyclones being Gerald and Holly).Freda was unusual because it traversed fewer than 10 degrees of longitude, but 25 degrees of latitude.
Super Typhoon Holly (September 5-15, 1987) was the third tropical cyclone to develop from the active monsoon trough which also spawned Gerald and Freda.Freda and Holly  The area is go latitudes by go longitudes box centered at the center of the typhoon.It encompasses eye, eye wall, most rain bands, and the environment of a typhoon.The area averaged brightness temperature should reflect the changes of the structure and intensity of the typhoon.However, if the typhoon is too large and/or asymmetric, this 8° latitudes .bygo longitudes area may not be adequate to cover all the main features of a typhoon; consequently the results may be biased.Therefore the general conclusion should be inferred with caution.
The responses of 19.35 and 37 GHz are almost identical, while 19.35 GHz shows lower brightness temperatures, because it does not respond to the presence of liquid water in the clouds and water vapor in the atmosphere.The 22.235 GHz has high brightness temperatures due to the emission of microwave energy from the abundant water vapor within and near the typhoon.These brightness temperatures do not vary much and show little relationship with the intensity of the typhoon.At g5.5 GHz, the presence of scattering of precipitation ice particles produces lower brightness temperatures as discussed in Section 2. The areal averages of 85.5 GHz brightness temperatures respond to the changes of intensity of typhoons.It appears that higher values of brightness temperature correspond to the weaker period of typhoon, the lower values to the intensifying period.This relationship is best illustrated in Typhoon Freda for its extensive coverage by the SSM/I.However, more cases are needed to confirm this relationship, and different radii may be explored for best representation of the typhoon's features.The method of defining a threshold or using a percentage of convection in the chosen area may prove to be effective in studying the variation of intensity of a typhoon.
The cross sei;tions of brightness temperatures at SSM/I frequencies for typhoons at the chosen times are shown in Figures  TYPHOON HOLLY ,A;•:*.TAO, Vol.3, No.4, Dec. 1992 the north of the typhoon with maximum estimated rain rate being 6.5 mm/hr (see Figure 17).
The surface maximum wind speed was about 95 knots.Weakening Typhoon Freda at 0914 UTC September 12, 1987 was at about 19.5°N and 138.5°E with moderate rain occurring to the southwest of the eye (see Figure 18).Figure 15 shows that a large eye (about 85 km in diameter) is identified by the warm brightness temperatures region at 85.5 GHz between 3.2° and 5.0° (4° being the center of the typhoon).
The convection to the east and south of the eye is noted for its low brightness temperatures at 85.5 GHz.The cross sections of Typhoon Holly at 0757 UTC September 10, 1987 is presented in Figure 16.Holly was at about 20°N and 155.5°E with extensive rainfall to the north of the eye (see Figure 19).The maximum surface wind speed was about 130 knots.The lower brightness temperatures of 85.5 GHz at eye wall are noted as expected.The symmetric structure in the east-west cross section is recognized, while more convection is found to the east of the eye.

RAINFALL ESTIMATES OF TYPHOONS
The inference of rainfall properties such as intensity and areal extent are very important to the understanding of global and regional hydrological cycle and water budget.Direct rainfall observations can be taken over the land and coastal areas, islands, and by ships.Oceanic rainfall has to be determined from the remote sensing data.Useful results have been obtained from visible and/or infrared radiometers although these instruments do not directly sense the precipitation (e.g., Kilonsky and Ramage, 1975;Arkin, 1979;Adler and Negri, 1988).
Microwave radiation can detect the presence of liquid water and precipitation ice in the clouds, so it can be applied for rainfull estimation as suggested by Wilheit et al. (1976),Wein man andGuetter (1977), and others.The Nimbus-7 SMMR launched in 1978 has produced useful data for more than ten years.Using an algorithm described in Wilheit and Chang (1980), Gloersen et aL (1984) found reasonable agreement between coastal rain gage observations and SMMR determi nations.Spencer et al. (1983) regressed SMMR radiances versus radar-derived rainfall rates over the Gulf of Mexico.Although a statistical regression of horizontally-polarized 37 GHz radiances against rainfall rates yielded a 72% explained variances for maritime rain fall rates, the data base was small,and no tropical cyclone data were included.Spencer and his co-investigators also concluded that a significant problem for determining rain rates for microwave radiometers was "beam filling", i.e., the instantaneous field-of-view was not uniformly filled with precipitation.Spencer (1986) later used SMMR 37 GHz channels in both polarizations to obtain esti mates of rainfall rate in Gulf of Mexico and South Atlantic precipitation systems.Prabhakara et al. (1992) utilized the lower SMMR frequencies (6.6 and 10.7 GHz) to estimate global distributions of total cloud water contents in precipitating clouds over the oceans.Total cloud content was then empirically related to rainfall rate.
Olson (19 87, 1991) investigated rain rate determinations for tropical cyclones.As part of that investigation radar data were used as input to a radiative transfer model which computed brightness temperatures for the SMMR frequencies and antenna patterns.Olson then retrieved fractional beam filling from these synthetic brightness temperatures.
Recently a simplifi ed signal channel emission-based algorithm for unpolarized 19 GHz measurements was developed by Smith and Mugnai (1988).A scattering-based algorit�m of Spencer et al. (1989) utilizes polarization differences at 37 GHz to measure the depo larizing effects of scattering by precipitation in terms of a polarization corrected brightness temperature (PCT), from which rainfall rates were then inferred.Alishouse et al. (1990) empirically derived equations for predicting radar reflectivity and fraction of the field-of-view of a microwave radiometer that is filled with rain.Oceanic rainfall then is inferred from the radar reflectivity.Their results for Hurricane David show that they can explain about 2/3 of the variance in predicting the radar reflectivity.Olson's (1991) method is based on the sta tistical regression of transformed brightness temperature parameters involving all four SSM/I frequencies to radar-retrieved rainfall data.Kummerow et al. (1991) utilizes the inversion of hydrometeor profile relationships to retrieved rainfall rates.Prabhakara et aL (1992) estimated oceanic rainfall from SMMR and SSMJI 37 GHz data by empirically developing a statistical relationship between the observed brightness temperatures at 37 GHz and rain rates.
In the present study, a simple al gorithm is applied to estimate rain rates from the measured SSM/I brightness temperature.The algori thm was developed by Olson (1987) The estimated rain rate for the typhoons at the selected Limes are depicted in Figs.
17-19.The rain rates appear to be reasonable in the vicinity of typhoons for all studied cases.The little or no rain areas are found at the eye and between rain bands, which can be identified clearly.The maximum rain rate is more than 6 mm/hr, which is comparable to the other investigations (Rao and Hutchison, 1992).
Because this algorithm can only be appl ied over the oceans, the rain rates over the land are not correct.It should also be noted that this algorithm tends to produce small rain rates (less than 1 mm/hr) over cloudfree regions.In the present study, the pattern and relative intensity of rainfall are more emphasized than the .accuracy of absolute values.The validation of the algorithm is needed , but there are only sporadic surface and upper observations over the oceans.The intercomparisons of various rainfall algori thms b:JSed on different theories seem to be feasible.

CONCLUDING REMARKS
Microwave data such as SSM/I provides unprecedented opportunity to study global aspects of moisture.The hydrological variables as estimated by the SSM/I are crucial for global water budget.The results of using passive microw ave measurements can also be useful for the verification of numeri cal model outputs.

•
The SSM/I data could provide realistic patterns and magnitudes of total precipitable water, cloud water content, and rain rate for the globe.They are comparable to the previous findings and other model analysis.The characteristics of SSM/I data are documented in 3. DMSP SSM/I DATA On June 19, 1987, Special Sensor Microwave/Imager (SSM/I) was launched abroad the Defense Meteorological Satellite Program (DMSP) block 5D-2 Spacecraft F8.The DMSP orbit is circular, sun-synchronous, and near-polar, with an altitude of 833 km and an inclina tion of 98.8°.The orbit period is 101.42 minutes (about 14.1 orbits per day).The local times for the ascending and descending equatorial crossing are 6:12 am and 6:12 pm.respectively.