Abstract

A detailed study on global oceanic precipitation is carried out using the simultaneous TOPEX and TMR (TOPEX Microwave Radiometer) data. It is motivated by the success of a series of feasibility studies based on a few years of TOPEX–TMR data, and the availability of a decade-long new dataset that spans 1992–2002. In this context, a previously proposed rain probability index is improved by taking into account the difference of the dynamic range of the TOPEX-measured backscatter coefficients at the Ku and C bands and the latitudinally complementary sensitivities of the TOPEX and TMR rain detections, leading to a refined joint precipitation index, which is generally consistent and quantitatively comparable with existing precipitation climatologies from the Global Precipitation Climatology Project (GPCP) and the Comprehensive Ocean–Atmosphere Data Set (COADS). The new TOPEX–TMR precipitation climatology, on the one hand, confirms the fundamental features of global oceanic rainfall with additional details, and, on the other hand, reveals a number of interesting characteristics that are previously unknown or poorly defined. 1) The spatial variability of the western Pacific “rain pool” (the atmospheric counterpart of the oceanic warm pool) is characterized by an interannual zonal migration, an annual cycle of meridional seesaw, and a semiannual cycle of expansion and shrinking. 2) The Pacific, Atlantic, and Indian Ocean intertropical convergence zones (ITCZs) all have an annual cycle of cross-basin oscillation with east and west stops in JJA and DJF, respectively. 3) A well-defined prominent rainy zone is observed in the southeast China Seas around Taiwan Island, connecting with the Pacific rain pool in the south. 4) Between El Niño and La Niña years, there is a systematic sign reversal of the geographical distribution of precipitation anomaly, which exists globally rather than in the tropical oceans only. 5) On a global basis, interannual and annual precipitation variabilities are of the same magnitude, but the interannual (annual) component is more important for the Southern (Northern) Hemisphere. 6) For the tropical oceans, “season” defined by rainfall usually has a one-quarter delay with respect to the corresponding meteorological season. For the “marine deserts” in the subtropical oceans, however, the rain-based season is found to be anticorrelated with the meteorological season. In addition, the annual cycle of the Atlantic precipitation is nearly 180° out of phase with respect to that of the Pacific and Indian Ocean for the same hemisphere.

1. Introduction

It is well known that rainfall can vary by orders of magnitude in intensity in a matter of minutes or even seconds and over distances of literally tens of meters. Although more than enough has been said on the importance of rainfall in shaping regional weather and global climate, accurate measurement of this geophysical parameter is extremely difficult owing to its enormous variation in space and time. Surface-based observations are inadequate for estimating rainfall over oceanic areas. In fact, major advances in describing and understanding oceanic precipitation have occurred since the advent of the satellite era in the 1960s. Following a clear recognition of the unique role that spaceborne sensors play in rainfall detection, the first dedicated precipitation satellite, the Tropical Rainfall Measuring Mission (TRMM), was successfully launched in November 1997, setting a milestone for rainfall monitoring (Simpson et al. 1988). During the past few years, tremendous efforts have been made to explore the TRMM dataset, leading to substantial progress in many aspects of rainfall characterization [see, e.g., the TRMM special issues in the Journal of Climate (2000, vol. 13, no. 23) and the Journal of Applied Meteorology (2000, vol. 39, no. 12)].

Despite the success of the TRMM mission, it is clear that other satellite rainfall derivation approaches need to be advanced in pace with the dedicated mission in order to maximize its potential. This can be understood for several reasons. First, given the intermittent and highly variable nature of precipitation, as well as the orbit and sampling scheme of present satellites, any spaceborne sensor can only capture a small fraction of rain events that occur globally. Additional rain information from nondedicated satellites is always valuable in terms of expanding spatial and temporal sampling, as evidenced in, for example, Huffman et al. (1997). Second, TRMM is designed to be an experimental rather than an operational satellite. It has a limited lifetime of several years and covers only the tropical and subtropical areas of the ocean within ±35°. In contrast, some of the other satellites with rain detection capacity, such as Special Sensor Microwave Imager (SSM/I), Microwave Sounding Unit (MSU) of National Oceanic and Atmospheric Administration (NOAA), and Geostationary Operational Earth Observing Satellite (GOES), have accumulated nearly 10 or 20 yr of continuous data with near-global coverage (Spencer 1993; Wilheit et al. 1994; Joyce and Arkin 1997). Third, direct validation of satellite-measured rainfall is difficult, not only because of inadequate field devices but also because of the poor representativeness of point measurement for an areal mean. These make cross-validation or intercomparison of simultaneous satellite observations highly desirable (e.g., Adler et al. 2001). Therefore, it can be concluded that all available sources of rain information could be complementary; a comprehensive sensing of the global rain system can only be achieved by integrating measurements from various satellite sensors and field devices.

An alternative method for rainfall estimation over the ocean is to use jointly data from the TOPEX altimeter and the TOPEX Microwave Radiometer (TMR). The feasibility of the TOPEX–TMR approach has been demonstrated in several previous studies (Quartly et al. 1996, 1999; Chen et al. 1997, 1998; Tournadre and Morland 1997). Similar to the initial design of the TRMM (and also the proposed future precipitation satellite) sensor package (Kummerow et al. 2000), the TOPEX–TMR system consists of a dual-frequency (5.3 and 13.6 GHz) active microwave component (the altimeter) and a three-frequency (18.0, 21.0, and 37.0 GHz) passive microwave component (the radiometer). Such a system has the advantage of integrating rain information from five microwave frequencies, leading to complementary sensing of the rain effect. By August 2002, the excellent performance of the TOPEX mission has resulted in a 10-yr continuous dataset with a broad coverage up to ±66° (nearly twice the coverage of TRMM, albeit for a narrower swath). Another advantage of the TOPEX–TMR system is that it can provide coincident measurements of several important geophysical parameters along with rainfall. These include sea surface height, sea surface wind speed, significant wave height, etc. Joint analysis of these parameters is very useful in revealing many aspects of the thermal and dynamic processes of the global ocean–atmosphere system (Quartly et al. 1999).

The work described in this paper is an improvement over, and an extension of, an earlier investigation on the same topic (Chen et al. 1997). In the present study, a modified algorithm is proposed to estimate the joint TOPEX–TMR precipitation index (section 2), on the basis of which, a decadal climatology of global oceanic rainfall is constructed and compared with similar products derived from the Global Precipitation Climatology Project (GPCP; Huffman et al. 1997) and the Comprehensive Ocean-Atmosphere Data Set (COADS; da Silva et al. 1994) (section 3). The precipitation variabilities at timescales from seasonal to decadal are also analyzed and discussed. In section 4, a harmonic analysis is performed to extract the annual, semiannual, and quarterly components of the rainfall variation, and meanwhile, to determine the “seasons” of the global ocean by rain. Finally, a summary with conclusions is presented in section 5.

2. Data and methodology

a. Data editing

The satellite altimeter is a nadir-pointing radar that transmits pulses vertically downward and measures the shape, power, and slope of the backscattering signals. From these signals many important geophysical properties of the ocean and the atmosphere can be derived (Fu and Cazenave 2001). The TOPEX/Poseidon satellite is a joint U.S.–French mission carrying a state-of-the-art radar altimetry system (Fu et al. 1994). It samples the ocean surface between 66°S and 66°N at a 1-s interval (corresponding to a 5.8-km resolution on the ground track) for each of the 254 ascending and descending passes that make up a 9.9156-day cycle. The ground track is maintained constant from cycle to cycle to within 1 km in a fixed pattern. Nearly ten years of TOPEX–TMR data spanning December 1992 through June 2002 are used in this study (AVISO 1996). A careful data screening is performed to eliminate measurements with abnormal instrumental or environmental conditions, as well as those with off-nadir pointing angles larger than 0.12°, resulting in an edited dataset containing time, longitude, and latitude of each valid sample, the coincident measurements of TOPEX backscatter coefficients for the Ku and C band, and the TMR brightness temperatures at 18.0, 21.0, and 37.0 GHz. It should be mentioned that in subsequent figures of this paper, a fixed ice mask is not used to blank out the seasonal sea ice zone, as it was in Chen et al. (1997), because we feel that the quality flags of the present generation dataset work reasonably well to ensure a rigorous classification of water and ice.

b. TOPEX precipitation index

The basic principle for rain detection by the TOPEX altimeter is based on the fact that rain attenuation in the Ku band is greater than that in the C band by an order of magnitude under a wide range of rain rates (Olsen et al. 1978). It is found that the mean relationship between TOPEX measured σ0C and σ0Ku remains very stable from cycle to cycle. Departure of (σ0C, σ0Ku) from the mean relationship with σ0Ku being lower than expected may then indicate a possible rain effect.

A “normal” relationship between σ0C and σ0Ku is calculated using all available data pairs. In doing so, the σ0Ku values are binned into 0.1-dB intervals with respect to σ0C over the range of 0–30.0 dB. The normal relationship, f(σ0C), can then be obtained by

 
formula

where i is the bin index and N(σ0C,i) is the number of data within bin i. The corresponding standard deviation, s(σ0C), is given by

 
formula

Graphic illustrations of f(σ0C) and s(σ0C) are shown in Figs. 1 and 2 (the thick black curves), respectively. Also overlaid on Fig. 2 is a normalized probability distribution function (pdf) of σ0C (the thin solid curve). The σ0Kuσ0C relationship basically takes a “two stick” form with a kink at σ0C ≈ 15.3 dB (or σ0Ku ≈ 12 dB, see Fig. 1) that corresponds to an intermediate wind speed according to Witter and Chelton (1991), indicating that wind speeds below or above the medium value essentially result from two scattering regimes. The standard deviation in Fig. 2 appears to be quite noisy for σ0C ≤ 11.5 dB and σ0C ≥ 23 dB due to insufficient data. The rest of the curve is supposed to be statistically significant. An interesting feature to be noted is the unusual bump around 14 dB, which is slightly shifted from the peak location of the pdf at about 14.5 dB. We suspected, at the beginning, that it was mainly due to rain effect. But after removal of the rain-contaminated data the bump still occurs, though less distinct. Nevertheless, given the smooth and pronounced nature of this feature, it is likely to be caused (perhaps jointly) by some sort of “systematic” effect resulting from the atmosphere and/or the ocean (possibly rain, swell, current, and so on). Note that, if the bump is removed and the gap is filled with a cubic spline (see the shadowed area in Fig. 2), the minimum of the smoothed s(σ0C) coincides perfectly with the maximum of the pdf.

Fig. 1.

The thick black curve indicates mean relationship between σ0ku and σ0C for TOPEX cycles 8–351. The grayscale area indicates TOPEX precipitation index computed using Eq. (3)

Fig. 1.

The thick black curve indicates mean relationship between σ0ku and σ0C for TOPEX cycles 8–351. The grayscale area indicates TOPEX precipitation index computed using Eq. (3)

Fig. 2.

Standard deviation (thick solid curve) and probability distribution function (thin solid curve) of σ0C for TOPEX cycles 8–351. The dashed curve bounding the shadowed area indicates the expected standard deviation after removal of the bump

Fig. 2.

Standard deviation (thick solid curve) and probability distribution function (thin solid curve) of σ0C for TOPEX cycles 8–351. The dashed curve bounding the shadowed area indicates the expected standard deviation after removal of the bump

Taking into account the rain-induced departure of σ0Ku from the normal relationship, f(σ0C) and the varying degree of sensitivity of rain effect on σ0Ku with respect to σ0C, a TOPEX precipitation index, PA, is introduced as

 
formula

where N1 = –2.5 is a constant. It is believed that the inclusion of a standard deviation to define a dynamic index is necessary because the range of s(σ0C) (from 0.1 to 0.7 dB) is too large to be neglected. Also note that the positions of σ0Ku and σ0C are shifted between Eqs. (3) and (1) of Chen et al. (1997), given that σ0Ku may be more sensitive to rain compared to σ0C. The distribution of PA as functions of σ0Ku and σ0C is plotted in Fig. 1 (the grayscale area). The figure shows a distinct two-regime pattern corresponding to low and medium-to-high wind speeds, respectively. It thus points to the need for introducing a dynamic, rather than a fixed, precipitation index with respect to σ0C.

c. TMR precipitation index

The definition of a TMR-based precipitation index is given by

 
PR = Lz/N2,
(4)

where Lz (in µm) is the atmospheric liquid water content expressed as a quadratic polynomial of the three corrected TMR brightness temperatures, Tb18, Tb21, and Tb37 (Keihm et al. 1995). In Eq. (4) N2 is set to 600 µm. Thus PR = 1 corresponds to the rain flag applied in the TOPEX Geophysical Data Record (GDR) product (Quartly et al. 1999).

d. Joint precipitation index

Following a similar scheme proposed by Chen et al. (1997), a joint TOPEX–TMR precipitation index, PJ, is defined as

 
PJ = w1PA + w2PR,
(5)

where w1 and w2 are weighting functions. In Chen et al. (1997), w1 and w2 are set to 0.75 and 0.25, respectively. In this study, we reconsider the form of these functions by examining the effectiveness of PA and PR with respect to latitude. If P ≥ 1.0 is regarded as a probable rain event, the zonal distributions of rain frequency triggered by PA or PR alone are shown in Fig. 3a. The solid line denotes PA ≥ 1.0 and PR < 1.0, while the dashed line denotes PR ≥ 1.0 and PA < 1.0. Apparently, the TOPEX-based index is much more effective for the tropical oceans as well as for latitudes higher than 60°. This confirms an earlier result by Quartly et al. (1999), who find that the TOPEX-based algorithm underestimates the rain rate by some 25% for the Tropics while by some 50% for the extratropics (see their Fig. 10). On the contrary, the TMR-based index has a much better performance at midlatitudes around ±45°. These are probably a manifestation of the unequal sensitivity of the two sensors to stratiform and convective events progressing from the equator to the poles. Given the highly complementary nature of the altimeter and the radiometer-based precipitation indices, two latitude-dependent weighting functions are proposed:

 
formula

where ϕ is the latitude of the sample. It can be seen from Fig. 3b that w1 (the solid curve) and w2 (the dashed curve) are normalized and phase opposite. These two formulas are employed in our analysis because they are analytically simple and geophysically meaningful. The so defined weights of w1 and w2 aim to take advantage of the altimeter sensitivity at the equator and near the poles and that of the radiometer sensitivity at midlatitudes. The final precipitation index, P, is derived by a prescribed empirical adjustment of the amplitude to match the GPCP zonal distribution (using data of “normal” years only, see next section), yielding

 
P = 24.0N3PJ cos(ϕ),
(7)

where N3 = 2.0. Combining Eqs. (1)–(7), an estimate of the joint precipitation index at nadir can be obtained. Note that, in common with the TRMM product, the TOPEX–TMR precipitation index also includes information from an active radar, which measures the two-way attenuation of rain, leading to an improved efficiency in the observations compared to a passive system. Moreover, in a recent work by McMillan et al. (2002) on the validation of the TOPEX rain algorithm against ground-based radar, it is concluded that the TOPEX precipitation climatology could be improved by incorporating the radiometric information from TMR, albeit that altimeter-only algorithm is still in use and produces reasonable results (Quartly et al. 1999, 2000).

Fig. 3.

(a) Rain frequency as a function of latitude determined by PA (solid curve) or PR (dashed curve) alone. (b) Weighting functions for PA (solid curve) and PR (dashed curve) in Eq. (6)

Fig. 3.

(a) Rain frequency as a function of latitude determined by PA (solid curve) or PR (dashed curve) alone. (b) Weighting functions for PA (solid curve) and PR (dashed curve) in Eq. (6)

3. Statistical analysis

a. Decadal climatology and interannual variability

A 9.5-yr climatology of global oceanic precipitation is constructed using TOPEX–TMR data from December 1992 to June 2002, as shown in Fig. 4a. A GPCP precipitation climatology for the same period is also presented in Fig. 4b for comparison. Figure 4c is the difference between the two climatologies. A quick look at Figs. 4a and 4b indicates that the basic features are almost identical: The principal rain zones in the tropical western Pacific and eastern Indian Ocean, the predominant zonal rain belts above the equator and in the midlatitudes of the Pacific, the Atlantic, and the Southern Ocean, as well as the six marine deserts in the eastern subtropical areas of the three ocean basins. The most heavily raining regime over the equatorial western Pacific is, in a sense, an atmospheric counterpart of the warm pool in the ocean, and is therefore called “rain pool” for simplicity. A somewhat surprising characteristic is that two pronounced rainbands in the North Pacific and North Atlantic [called North Pacific rainband (NPRB) and North Atlantic rainband (NARB)], as well as a circumpolar rain belt (CPRB) along 50°S of the Southern Ocean, appear to be better defined and much intensified compared to the well-known South Pacific convergence zone (SPCZ) (Fig. 4a). These rainbands/belts correspond to the storm tracks of the two hemispheres. For the CPRB, however, precipitation intensity is strongest in the South Atlantic, as opposed to storm activity, which is strongest in the southern Indian Ocean (Trenberth 1991). More generally, one can find [by comparing Fig. 4a with Fig. 1 of Chen et al. (2002), for instance] that the locations of maxima in rain and wind intensity are often shifted, or even anticorrelated. This might be related to the subtropical and polar jet streams centered over zones of rising air near ±30° and sinking air near ±60°. Recall that it is the location of the boundary between (rather than the centers of) subtropical and polar air that is most important for the evaporation and precipitation activities in those regions. A shifted pattern of wind and rain intensity as observed is therefore less surprising.

Fig. 4.

Global oceanic precipitation climatology for 1992–2002 derived from (a) TOPEX–TMR and (b) GPCP and (c) the difference between (a) and (b)

Fig. 4.

Global oceanic precipitation climatology for 1992–2002 derived from (a) TOPEX–TMR and (b) GPCP and (c) the difference between (a) and (b)

A closer examination of the top two maps in Fig. 4 allows several significant differences to be identified. First, most of the rain zones are shifted in position between Figs. 4a and 4b. For example, the center of the western Pacific rain pool in the TOPEX–TMR map is approximately 10° west compared to that in the GPCP map. The NARB in Fig. 4a is to the northeast of that in Fig. 4b by a considerable distance. Second, some marked features are observable in one map but are absent in the other, such as the significant rain zone around Taiwan and a well-defined rainy belt in the Atlantic sector of the Southern Ocean on the TOPEX–TMR map alone (Fig. 4a), and a narrow band of heavy precipitation near 240°E of the equatorial Pacific on the GPCP map alone (Fig. 4b). Third, the marine deserts in Fig. 4b appear to be much drier compared to Fig. 4a. The difference between the TOPEX–TMR and GPCP climatologies exhibits a largely random pattern, with the former being slightly higher on average (Fig. 4c). Interestingly, the TRMM-derived oceanic precipitation is also found to bias positively (∼10%) versus GPCP (Adler et al. 2000). Given that GPCP is the pre-TRMM state-of-the-art precipitation estimate, it can be concluded that our TOPEX–TMR precipitation result is generally consistent with existing climatologies.

The anomalously high level of the TOPEX–TMR estimates over the East China Sea (Fig. 4a) deserves further examination. This feature is reported by Spencer (1993) as “a persistent maximum seen in all seasons, apparently marking the beginning of the western Pacific storm track.” In fact, similar features are also identifiable (though less obvious) in other climatologies based on Jaeger (1976) and Legates and Willmott (1990) [see the middle and bottom panels of Fig. 3 in Huffman et al. (1997)]. Concerning the reason of this feature, Spencer (1993) suggests that the East China Sea experiences an unusually high cloud/rainwater ratio, which might well affect any microwave emission estimates in this region. But on the other hand, land precipitation maps of eastern Asia (not shown) indicate that the rainfall distribution is bow-shaped over central and east China, increasing from northwest to southeast with highest values observed near the Zhejiang coast. This pattern seems to extend to the Korean peninsula as well as to the Japanese and Taiwan islands (a picture somewhat similar to Fig. 8f below) since the western areas of Japan have almost the same level of annual rainfall as southeast China despite their 10° latitudinal difference. In other words, southeast China and west Japan might belong to the same precipitation regime, at least for some seasons. From a continuity point of view, the existence of an enhanced rain belt between the two regions can be more or less expected. Of course, storm activities may also contribute to (modify) the rain system in this area.

Fig. 8.

Annual anomaly of oceanic precipitation with respect to a decadal (1992–2002) climatology derived from TOPEX–TMR: (a) 1993, (b) 1994, (c) 1995, (d) 1996, (e) 1997, (f) 1998, (g) 1999, (h) 2000, and (i) 2001. Negative anomalies are denoted in blue

Fig. 8.

Annual anomaly of oceanic precipitation with respect to a decadal (1992–2002) climatology derived from TOPEX–TMR: (a) 1993, (b) 1994, (c) 1995, (d) 1996, (e) 1997, (f) 1998, (g) 1999, (h) 2000, and (i) 2001. Negative anomalies are denoted in blue

An intercomparison of the TOPEX–TMR, GPCP, and COADS precipitations is conducted from a zonal perspective, as shown in Fig. 5. Recall that the GPCP (Huffman et al. 1997) is a combination of several spaceborne (excluding TOPEX–TMR) and ground-based precipitation observations, whereas the TOPEX–TMR and COADS (da Silva et al. 1994) are purely satellite and in situ observations, respectively. The three sources of precipitation are therefore thought to be totally independent, each representing a disparate category. The general impression of Fig. 5a is that the TOPEX–TMR and GPCP results are rather similar in terms of major zonal features, while the COADS result shows large discrepancies (some 30% higher) around the ITCZ at about 7°N, and at high latitudes beyond 60°S. As noted before, the GPCP and COADS results exhibit opposite trends toward the Antarctic (Chen et al. 1997), while a sort of “compromise” is reached in our result. Moreover, the COADS precipitation appears to be higher than the two others for most latitudes south of 10°N, and the reverse is true for other latitudes. In addition to the well-known uncertainties associated with satellite rain measurements, those unique to ship observations are also responsible for the differences in the comparison. For example, the extremely inhomogeneous sampling in space and time, systematic misclassification by inexperienced observers, “fair” or “foul” weather bias, mislocation, etc. [see Petty (1995) for a detailed discussion]. Quantitatively, the scatter of the three climatologies is less than 0.5 mm day–1 for a majority of latitudes, but can reach a factor of 3 or 4 around 7°N or beyond 60°S (Fig. 5b). Geographically, the standard deviation of the three precipitation regimes varies from 0.43 mm day–1 in the Atlantic to 0.64 mm day–1 in the Indian Ocean, with a global mean reduced to 0.50 mm day–1 (see Table 1). These statistics are in general agreement with previous comparisons (e.g., Adler et al. 2001; Quartly et al. 1999). Also, it is possible (with the COADS distribution in common) for the result of the present algorithm to be compared with that of Chen et al. (1997, their Plate 2). The major difference is that the previous rain index seems to be too high in the midlatitudes of the Northern Hemisphere relative to their Southern Hemisphere counterparts. Note that the good agreement between TOPEX–TMR and COADS curves in their Plate 2 resulted from a subjective match, as stated in the caption. It is also important to remember that the length of data is 1 year and 9 1/2 yr for the previous and present studies, respectively.

Fig. 5.

(a) Zonally averaged oceanic precipitations derived from TOPEX–TMR (thick solid curve), GPCP (thin solid curve), and COADS (dashed curve) climatologies. (b) The corresponding standard deviation of (a)

Fig. 5.

(a) Zonally averaged oceanic precipitations derived from TOPEX–TMR (thick solid curve), GPCP (thin solid curve), and COADS (dashed curve) climatologies. (b) The corresponding standard deviation of (a)

Table 1.

Comparison of standard deviation of the TOPEX–TMR, GPCP, and COADS precipitation climatologies

Comparison of standard deviation of the TOPEX–TMR, GPCP, and COADS precipitation climatologies
Comparison of standard deviation of the TOPEX–TMR, GPCP, and COADS precipitation climatologies

As a relative measure of the interannual variability of global oceanic precipitation, the spatial distribution of the standard deviation of the annual average TOPEX–TMR indices divided by corresponding decadal mean climatology is shown in Fig. 6. Two areas with extraordinarily large interannual variations are apparent in the western equatorial Pacific and the eastern equatorial Indian Ocean, respectively. The one in the Indian Ocean coincides well with its rain pool, implying that the Indian Ocean rain pool is stable in location but variable in intensity. The center of the interannual variability high in the equatorial Pacific, however, is shifted to the east by approximately 20° from that of the rain pool, suggesting a zonally unstable nature of the Pacific rain pool (which can be further identified in the annual precipitation maps in Fig. 7 below). Other tropical interannual variability highs are found in the central and eastern Pacific, as well as in the surrounding waters of Indonesia. Note that the only area with high values in the extratropical ocean is observed in the eastern China Seas. This seems to be consistent with the dramatic climate variability in this region during the past decade. To give an example, China suffered massive flooding (which is the second worst to hit the country in more than 130 years) in the southeastern areas along the Yangtze River during the summer of 1998 (as evidenced in Fig. 8f below).

Fig. 6.

Relative interannual variability of global oceanic precipitation for 1992–2002 derived from TOPEX–TMR precipitation index

Fig. 6.

Relative interannual variability of global oceanic precipitation for 1992–2002 derived from TOPEX–TMR precipitation index

Fig. 7.

Annually averaged oceanic precipitation index derived from TOPEX–TMR: (a) 1993, (b) 1994, (c) 1995, (d) 1996, (e) 1997, (f) 1998, (g) 1999, (h) 2000, and (i) 2001

Fig. 7.

Annually averaged oceanic precipitation index derived from TOPEX–TMR: (a) 1993, (b) 1994, (c) 1995, (d) 1996, (e) 1997, (f) 1998, (g) 1999, (h) 2000, and (i) 2001

b. Annual means and anomalies

The TOPEX–TMR-derived annual precipitation maps for 1993–2001 are presented in Fig. 7. As can be seen, the global pattern of oceanic precipitation experiences notable changes from year to year. This is particularly evident for the zonal migration of the western Pacific rain pool. It moved to the easternmost position with a minimum overall intensity in 1997 when the strongest El Niño of the twentieth century occurred (McPhaden 1999) and returned to its westernmost position with a maximum overall intensity in 1999 when a major La Niña event occured (Figs. 7e and 7g). Since the atmosphere receives three-fourths of its energy from the release of latent heat of precipitation and more than half of this precipitation falls in the western tropical Pacific, such a pronounced change of the rain pool in space and intensity is responsible, to a very large extent, for the worldwide abnormal climate during that period. The evolution of the eastern Indian Ocean rain pool is also remarkable: The almost disappearance of this zone in 1997 is followed immediately by a decadal peak in its area and intensity in 1998 (Figs. 7e and 7f). The year-to-year variation of oceanic precipitation in areas other than the rain pools is also significant but appears to be mild and gradual in general, such as the wet zone in the equatorial eastern Pacific and the marine deserts in all ocean basins. These results clearly demonstrate that the basic climate signals have been well detected by the TOPEX–TMR precipitation dataset.

Some of the aforementioned features are more easily identifiable from the annual anomaly maps as shown in Fig. 8. Examining the nine panels in Fig. 8, one finds that large positive/negative anomalies are always associated with El Niño/La Niña episodes. The El Niño years of 1993, 1994, and 1997 all display a positive (negative) anomaly in the east (west) portion of the western Pacific rain pool (Figs. 8a, 8b, and 8e). The magnitude of these anomalies seems to be proportional to the intensity of the El Niño. An inverse pattern with sign-opposite anomalies in this area is visible during the La Niña years of 1999 and 2000 (Figs. 8g and 8h). By comparing Figs. 8a and 8g one recognizes that the anomaly reversal between El Niño and La Niña years may exist globally. 1998 is a special year of mixed El Niño and La Niña, the anomaly of which shows a unique structure with unusually large amplitude of positive values in the central and eastern equatorial Pacific and eastern equatorial Indian Ocean, as well as in the equatorial Atlantic (Fig. 8f). The remaining years, 1995, 1996, and 2001, are normal years whose anomalies are nearly neutral and quite homogeneous (Figs. 8c, 8d, and 8i). A joint analysis of Figs. 7 and 8 suggests that the decadal variation of global oceanic precipitation is closely linked to the El Niño–Southern Oscillation (ENSO). Two types of phase-opposite but geographically correlated anomaly patterns are observed during El Niños and La Niñas, respectively. They imply a zonal shift of major equatorial rain zones between the two episodes, which is obviously related to the zonal migration of the western Pacific warm pool (Matsuura and Iizuka 2000). Many previous researchers have investigated the ENSO-related rainfall anomaly in the tropical Pacific. Joyce and Arkin (1997), among others, point out that, during a warm episode, the spatial distribution of the precipitation anomaly exhibits a dipole pattern with a positive maximum near the date line and a negative minimum over Indonesia. Our results confirm the existence of a dipole in El Niño years (see Figs. 8a, 8b, and 8e), but suggest that the positive maximum moves no farther than 170°E for moderate El Niños. It seems that the date line can only be reached during extraordinary El Niños such as in 1997 (McPhaden 1999). The 10° displacement in dipole position between our results and those in Joyce and Arkin (1997) and Spencer (1993) might result from the different periods of data that have been used (note that the aforementioned previous results are based on data acquired between 1986 and 1994 when El Niños prevailed, while our data cover 1993–2001 when El Niños were less dominant in terms of frequency of occurrence). Another impressive aspect in Figs. 8e and 8f is that the dipole appears to tilt clockwise while extending/translating eastward during 1997–98. Such an unusual distortion of the dipole is, to our knowledge, previously unseen.

c. Seasonal means

Unlike the sea surface air temperature whose seasonal change is dominated by an annual cycle with a hemispheric phase opposition, the seasonality of oceanic precipitation is characterized by zonal shifts of major rain zones and regional fluctuations in rain intensity (Fig. 9). The intense rain belt over the Pacific ITCZ migrates eastward from DJF to MAM (Figs. 9a and 9b) and westward from JJA to SON (Figs. 9c and 9d). The annual oscillation of a well-defined wet zone in the equatorial Atlantic is particularly evident (Figs. 9a–9d). These well-known tropical features agree fairly well with similar results from other precipitation climatologies (e.g., Spencer 1993; Joyce and Arkin 1997; Huffman et al. 1997; Xie and Arkin 1997), given that they are based on disparate data sources from different years. Focusing on the western equatorial Pacific, we find that the world’s largest rain pool undergoes a semiannual cycle of expansion and shrinking in area and strengthening and weakening in intensity with a primary peak in JJA (Fig. 9c) and a secondary peak in DJF (Fig. 9a). This characteristic is hardly recognizable in other referenced rain maps. In the extratropical areas of the Northern Hemisphere, the storm tracks of the North Pacific and North Atlantic move northeastward in MAM and JJA (Figs. 9b and 9c), while southwestward in SON and DJF (Figs. 9d and 9a), consistent with the findings of Spencer (1993). However, they are found to be different from the results of Huffman et al. (1997) in that an axis of enhanced precipitation appears to connect the storm tracks to the low-latitude maximum over Southeast Asia during the cold rather than warm seasons of the Northern Hemisphere. The intensity variations of both the NPRB and NARB are dominated by an annual cycle. But surprisingly, they are found to be phase opposite, with the NPRB following a typical boreal behavior while the NARB has an antiseasonal cycle. Such a large-scale incoherency in oceanic precipitation is somewhat unexpected and is not seen in either GPCP (see Fig. 4 of Huffman et al. 1997) or other available climatologies.

Fig. 9.

Three-year (1995, 1996, and 2001) averaged quarterly oceanic precipitation derived from TOPEX–TMR. (a) DJF, (b) MAM, (c) JJA, and (d) SON. The three selected years are considered to be normal years without significant El Niño or La Niña events

Fig. 9.

Three-year (1995, 1996, and 2001) averaged quarterly oceanic precipitation derived from TOPEX–TMR. (a) DJF, (b) MAM, (c) JJA, and (d) SON. The three selected years are considered to be normal years without significant El Niño or La Niña events

The Indian Ocean rainfall system is unique in many aspects compared to the other two basins as a result of its inland nature in the north. A slightly tilted east–west axis of maximum precipitation extends from Sumatra to the northern tip of Madagascar in DJF (Fig. 9a). The rain core in its west weakens rapidly while moving northeastward in MAM (Fig. 9b). By JJA, it has merged with the monsoon-induced heavy rain zone in the Bay of Bengal (Fig. 9c). In SON, the Indian Ocean rain pool controlled by the monsoon rotates clockwise by nearly 90°. Meanwhile, it extends southwestward systematically (Fig. 9d), setting a favorable condition for the winter reenhancement to occur (Fig. 9a). This complex evolution is basically a reflection of the interaction between the ITCZ and the Indian monsoon. A similar process is largely evident in Spencer (1993, see his Fig. 18), but with a coarser resolution (2.5° vs 1°) and hence less detail (see, e.g., the ITCZ-related rain core to the northeast of Madagascar).

d. Time series

The monthly time series of oceanic precipitation index averaged over global and regional oceans are plotted in Fig. 10. Figure 10a shows the globally and hemispherically averaged results. The global time series (the thick curve) displays a small annual amplitude on top of a large interannual variability, which peaks around late 1998 and early 1999. There is a general phase reversal between the two hemispheres, with the Northern Hemisphere considerably exceeding the Southern Hemisphere for most of the time. A significant increase in hemispherically averaged annual amplitude is observed for the 1997–98 El Niño period. As far as individual ocean basins are concerned, the Pacific Ocean is the one that most resembles the global pattern in terms of time evolution (Fig. 10b). Specifically, the double maxima pattern is found to result from the inconsistency between the ITCZ annual cycle in the central and eastern Pacific and the cross-equator meridional oscillation of the rain pool in the western Pacific. This annually based inconsistency appears to be overtaken by the ENSO-induced interannual variability during the period 1997–99, but begins to recover somewhat in 2000. The Indian Ocean has a unique feature that the amplitude of its annual variation is three times larger for the Northern Hemisphere than for the Southern Hemisphere due to the monsoon effect (Fig. 10c). But the overall interannual variability of this ocean is generally small. A weak minimum is reached in late 1997, followed immediately by a decadal maximum in mid-1998. Note that, during this period, the three basins all show a wetter-than-usual and a drier-than-usual condition in the Northern and Southern Hemisphere, respectively. The Atlantic Ocean has the smallest annual and interannual amplitude among the three basins (Fig. 10d). The ENSO impact also seems to be the least for this ocean. It would be helpful to compare our results with those in Quartly et al. (1999, see their Fig. 5) and Quartly et al. (2000, see their Fig. 2), which are based on TOPEX altimeter data alone. The comparison reveals several interesting differences. For example, our result indicates a trough-to-peak jump in global oceanic rainfall from 1997 to 1998, while their result suggests a broad high throughout 1997–98. The double-peak structure is more evident in our Pacific plot than in their Indian Ocean plot. They observe lowest values of north Indian Ocean rainfall in early 1997 (which is argued to be a possible precursor of El Niños) for the period 1993–99, but we observe this feature for both 1996 and 1997. Given the diverse nature of the differences identified, there seems to be no single reason that can provide a satisfactory explanation. But possible contributing factors may include the incorporation of TMR data in our analysis, the differences in precipitation algorithms, as well as a small shift in setting the ocean boundaries [note that 280°E is used as a dividing line between the Pacific and Atlantic in our analysis, whereas 285°E was used by Quartly et al. (1999, 2000)].

Fig. 10.

Monthly time series of oceanic precipitation for (a) global oceans, (b) Pacific Ocean, (c) Indian Ocean, and (d) Atlantic Ocean. The thick, thin, and dashed curves denote the mean values of the global ocean, the Northern Hemisphere, and the Southern Hemisphere, respectively. (e) Same as above but for the North Pacific (thick curve), North Atlantic (thin curve), and North Indian Ocean (dashed curve), respectively

Fig. 10.

Monthly time series of oceanic precipitation for (a) global oceans, (b) Pacific Ocean, (c) Indian Ocean, and (d) Atlantic Ocean. The thick, thin, and dashed curves denote the mean values of the global ocean, the Northern Hemisphere, and the Southern Hemisphere, respectively. (e) Same as above but for the North Pacific (thick curve), North Atlantic (thin curve), and North Indian Ocean (dashed curve), respectively

Next, it would be interesting to further examine the phase relationship of the three Northern Hemisphere rainfall variations. In doing so, the time series of the precipitation index for the North Pacific, North Atlantic, and North Indian Oceans are replotted in Fig. 10e. Clearly, the Pacific and Indian Ocean curves oscillate with slightly shifted phases, while the Atlantic Ocean is nearly 180° out of phase with them. This striking feature is not discernible in a similar plot of Quartly et al. (1999) (see their Fig. 5b) nor is it visible in the GPCP-based result (not shown). It is felt that further study is needed before a convincing explanation can be given.

4. Harmonic analysis

In this section, a harmonic analysis is performed to extract the annual, semiannual, and quarterly components of oceanic precipitation from the decadal TOPEX–TMR dataset and, meanwhile, to determine the “seasons” (the timing of wet and dry quarters) of the global ocean by rain.

a. Amplitude

Figure 11 shows the amplitudes of precipitation variability obtained by harmonic analysis. Obviously, the variability is overwhelmingly dominated by the annual component, which has a zonally banded structure with alternating narrow belts of lows and highs poleward from the equator (Fig. 11a). Minima are found at about 0°, ±20°, and ±50°, while maxima are around ±10° and ±40°. Such a regular pattern confirms that the global oceanic rainfall is basically associated with synoptic atmospheric circulation. Notable features in Fig. 11a are the highs in the Asian monsoon regions, the tropical northeast Atlantic, and the area southwest of Mexico. Higher frequency harmonics are more complicated in spatial distribution with much less zonal preference (Figs. 11b and 11c). As far as the semiannual component is concerned, high values are concentrated in the monsoon and the NPRB areas. In contrast, the primary maximum is observed at the NPRB, and secondary maxima at the SPCZ and SACZ for the quarterly component. The globally averaged annual, semiannual, and quarterly harmonics are 1.51, 0.79, and 0.58 mm day–1, respectively, implying that the fluctuation of rain is mostly at an annual frequency.

Fig. 11.

Harmonics of global oceanic precipitation variability. (a) Annual component, (b) semiannual component, and (c) quarterly component

Fig. 11.

Harmonics of global oceanic precipitation variability. (a) Annual component, (b) semiannual component, and (c) quarterly component

Next, we would like to quantify the relative importance of interannual and annual variabilities by plotting the geographical distribution of the ratio of their amplitudes, as illustrated in Fig. 12. The global average of this ratio is estimated to be 1.01, suggesting that the overall annual and interannual variabilities are of the same order of magnitude. For an arbitrarily given location, however, this ratio varies from 0.1 to 2.4. Relatively, the Northern Hemisphere has an annual dominance (ratio = 0.83) while the Southern Hemispher has an interannual dominance (ratio = 1.16), as can be recognized in Fig. 12. The Asian monsoon region appears to have a very low year-to-year variability in precipitation. It is somewhat surprising that even the rain pools in the western Pacific and eastern Indian Ocean do not exhibit a marked interannual dominance. In contrast, very high ratios are observed in the Pacific sector of the Southern Ocean, especially to the south and southeast of Australia.

Fig. 12.

Geographical distribution of the ratio of interannual vs annual amplitude of oceanic precipitation

Fig. 12.

Geographical distribution of the ratio of interannual vs annual amplitude of oceanic precipitation

b. Phase

We now examine the marine seasons defined by precipitation. The spatial distributions of the months in which maximum and minimum annual precipitations occur are plotted in Figs. 13a and 13b. An overall impression is that the geographical pattern is characterized by a zonally oriented structure with no systematic phase oppositions between either the wet and dry quarters or the two hemispheres. As far as the tropical oceans are concerned, however, the Southern and Northern Hemispheres reach a rainfall maximum (minimum) and minimum (maximum) during the same period of MAM (SON). This partially explains the phenomenon that, for the western Pacific rain pool, the southern tongue leads the northern tongue in area and strength in MAM (Fig. 9b), while the reverse is true in SON (Fig. 9d). The rainfall peaks of the northern Pacific and northern Indian Ocean occur in JJA (Fig. 13a), which are consistent with the annual cycle of air temperature. A majority of the Southern Hemisphere oceans reach their annual precipitation maximum in MAM, having a quarter delay compared to the austral summer. As to the annual minimum, the Northern Hemisphere largely displays a MAM–JJA–MAM pattern for low, middle, and high latitudes, respectively (Fig. 13b). Comparatively, the Southern Hemisphere oceans are more localized and scattered in reaching their dry season. It should be pointed out that, for most of the six marine deserts, the rain-based season is anticorrelated with the season defined by air temperature, although some of them are less evident owing to the noisy nature of rain statistics.

Fig. 13.

Geographical distribution of the timing of (a) wet and (b) dry seasons defined by rain. Wet and dry seasons here refer to a maximum and a minimum quarterly precipitation, respectively

Fig. 13.

Geographical distribution of the timing of (a) wet and (b) dry seasons defined by rain. Wet and dry seasons here refer to a maximum and a minimum quarterly precipitation, respectively

5. Summary and conclusions

Following a feasibility study of deriving oceanic precipitation from TOPEX and TMR (Chen et al. 1997), a modified algorithm for estimating the precipitation index is proposed in this paper, resulting in several significant improvements. First, contrary to the previous definition [see Eq. (1) of Chen et al. 1997], the present TOPEX precipitation index is defined as a primary function of σ0ku, and a secondary function of σ0C [see Eq. (3)]. This is believed to have increased the sensitivity of the estimated index as a result of a wider dynamic range associated with rain-affected σ0ku. Second, a latitude-dependent weighting function is applied for both TOPEX and TMR, which considerably enhances the spatial consistency of the derived precipitation climatology. Although the complementary nature of TOPEX and TMR in rain sensing with respect to latitude is, to some extent, recognized by Chen et al. (1997, see their Fig. 4), it is not taken into account until the present work is carried out. Figure 4c suggests that, with this critical change, geographically correlated bias between the TOPEX–TMR and the GPCP precipitation climatologies becomes almost insignificant. Third, the TOPEX–TMR precipitation index is quantitatively comparable to other rain climatologies such as GPCP and COADS in terms of rain rate. This characteristic facilitates its integration into existing climatologies and its assimilation into numerical models.

Despite the above improvements, the potential effect of the latitude-dependent weighting scheme on some of the obtained results should be noted. For example, the midlatitude zonal features in Fig. 11c may relate to a particular seasonal sensitivity of the radiometer algorithm (which is dominant at these latitudes). Similarly, the zonal band of higher interannual variability at 45°S in Fig. 12 and the change in seasonal minimum/maximum rainfall with latitude in Fig. 13 may also be partially attributable to the algorithm employed. Therefore it is fair to consider the present forms of the weighting functions as first-order approximations, which will be further explored and validated using coincident satellite and field measurements.

The availability of a decade-long TOPEX–TMR dataset allows us to produce a new and independent precipitation climatology based on the improved methodology. The results appear to be generally consistent with the comprehensive GPCP climatology (Fig. 4), implying that the suite of sensors onboard the TOPEX and its follow-on missions (such as JASON and ENVISAT) are capable of providing valuable by-products on global oceanic precipitation. Furthermore, the unique combination of the sensor package (a dual-frequency radar plus a three-frequency radiometer) along with its non-sun-synchronous sampling scheme make TOPEX and TMR highly complementary to other rain observation approaches. Our main findings on global oceanic precipitation are summarized below.

  1. On a decadal scale, the global oceanic precipitation climatology is dominated by two rain pools (the western Pacific rain pool and the eastern Indian Ocean rain pool), five rainbands (ITCZ, SPCZ, SACZ, NPRB, and NARB), six marine deserts (dry zones in the eastern subtropical portion of the three ocean basins), and a circumpolar rainbelt in the Southern Ocean. Although these features are also identifiable for other precipitation climatologies such as GPCP, they are found to have a geographical shift of 10°–20° in terms of central location and are very often inconsistent in spatial structure and relative intensity. Typical examples include the western Pacific rain pool, the ITCZ, the NPRB, and the SACZ (see Figs. 4a and 4b). An unusual feature that appears in the TOPEX–TMR result is the predominant rain zone in the southeast China Seas surrounding Taiwan. Such systematic differences between various precipitation maps suggest that there is still a long way to go before achieving, even in a climatological sense, a consistent and reliable rainfall estimation over the ocean.

  2. As far as the interannual variability of oceanic precipitation is concerned, the primary influential factor is ENSO. It is found that the magnitude of the annual precipitation anomaly in the tropical oceans is proportional to the intensity of the El Niño/La Niña events. Moreover, there appears to be a systematic reversal of the global pattern of precipitation anomaly between El Niño and La Niña years (Fig. 8). At a regional scale, very large interannual changes are associated with the two rain pools (Fig. 6). However, the high variability of the Pacific rain pool is mainly caused by its spatial instability, while that of the Indian Ocean rain pool mainly results from its intensity fluctuation. Again, the eastern China Seas exhibit a relatively high interannual variability. The 1.01:1 ratio of the interannual versus annual variation suggests that they are nearly of the same importance in shaping the global precipitation, although a south–north tilt of their influence is evident (Fig. 12). In terms of time evolution, an unexpected feature is that the Atlantic precipitation is nearly 180° out of phase with respect to that of the Pacific and Indian Ocean for the same hemisphere (Fig. 10e). Such a systematic phase reversal is, to our knowledge, previously unreported.

  3. Traditionally, seasonality refers to geophysical variations with respect to the meteorological seasons. In this context, the seasonality of oceanic precipitation is characterized by zonal migrations of major rain zones and regional fluctuations in rain intensity. The Pacific, Atlantic, and Indian Ocean ITCZs all have an annual cycle of cross-basin oscillation with east and west stops in JJA and DJF, respectively (Fig. 9). The Pacific rain pool undergoes a semiannual cycle of expansion and shrinking accompanied by coherent changes in intensity (primary and secondary peaks in JJA and DJF, respectively). The Indian Ocean rain pool has a bimodal nature: a dramatic strengthening in JJA and SON, and an abrupt weakening in the other half of the year. In a broader sense, however, seasons can also be defined with a given geophysical parameter that has a clear annual cycle. As far as the wet and dry seasons of rainfall are concerned, a zonally oriented structure with no systematic hemispheric opposition is observed (Fig. 13). For the tropical oceans, the rainfall seasons usually have a quarter delay compared to the corresponding meteorological seasons. For the marine deserts in the subtropical oceans, however, their rain-based seasons are found to be anti-correlated with the corresponding meteorological seasons.

Given the encouraging results of precipitation climatology and variability derived from TOPEX–TMR, it can be concluded with confidence that a dual-frequency altimeter in conjunction with a multifrequency radiometer is one of the most effective tools in observing oceanic precipitation. As Dr. Joanne Simpson, the founding Project Scientist of TRMM, points out: “The most innovative part of TRMM was and is to use rain radar and passive microwave together so that the joint measurement is better than either one alone” (Tao et al. 2000). Combined active/passive remote sensing systems, such as TOPEX–TMR, JASON, and ENVISAT (which serves as a simplified TRMM in some aspects), may therefore have a greater part to play in a post-TRMM era before the next dedicated rainfall measuring mission comes into being.

Acknowledgments

This work is cosponsored by the Natural Science Foundation of China (Project 40025615), the National High-Tech Project on Ocean Monitoring Technology (Project 818-AA63-03-06), and the Teaching and Research Award Program for Outstanding Young Teachers in Higher Education Institutions of MOE, PRC. The authors are very grateful to Dr. Graham D. Quartly at the Southampton Oceanography Centre (UK) for his thorough review on an earlier version of this paper.

REFERENCES

REFERENCES
Adler
,
R. F.
,
G. J.
Huffman
,
D. T.
Bolvin
,
S.
Curtis
, and
E. J.
Nelkin
,
2000
:
Tropical rainfall distributions determined using TRMM combined with other satellite and rain gauge information.
J. Appl. Meteor.
,
39
,
2007
2023
.
Adler
,
R. F.
,
C.
Kidd
,
G.
Petty
,
M.
Morissey
, and
H. M.
Goodman
,
2001
:
Intercomparison of global precipitation products: The Third Precipitation Intercomparison Project (PIP-3).
Bull. Amer. Meteor. Soc.
,
82
,
1377
1396
.
AVISO
,
1996
:
AVISO user handbook: Merged TOPEX/Poseidon products. 3d. ed.
AVI-NT-02-101-CN, Toulouse, France, 196 pp
.
Chen
,
G.
,
B.
Chapron
,
J.
Tournadre
,
K.
Katsaros
, and
D.
Vandemark
,
1997
:
Global oceanic precipitation: A joint view by TOPEX and the TOPEX microwave radiometer.
J. Geophys. Res.
,
102
,
10457
10471
.
Chen
,
G.
,
B.
Chapron
,
J.
Tournadre
,
K.
Katsaros
, and
D.
Vandemark
,
1998
:
A new look at the diurnal variation of global oceanic precipitation from the ocean TOPography Experiment (TOPEX) and the TOPEX Microwave Radiometer (TMR).
Int. J. Remote Sens.
,
19
,
171
180
.
Chen
,
G.
,
R.
Ezraty
,
C.
Fang
, and
L.
Fang
,
2002
:
A new look at the zonal pattern of the marine wind system from TOPEX measurements.
Remote Sens. Environ.
,
79
,
15
22
.
da Silva
,
A. M.
,
C. C.
Young
, and
S.
Levitus
,
1994
:
Algorithms and Procedures. Vol. 1, Atlas of Surface Marine Data,.
NOAA Atlas NESDIS 6, 83 pp
.
Fu
,
L-L.
, and
A.
Cazenave
,
2001
:
Satellite Altimetry and Earth Sciences.
International Geophysics Series, Vol. 69, Academic Press, 463 pp
.
Fu
,
L-L.
,
E. J.
Christensen
,
C. A.
Yamarone
,
M.
Lefebvre
,
Y.
Ménard
,
M.
Dorrer
, and
P.
Escudier
,
1994
:
TOPEX/Poseidon mission overview.
J. Geophys. Res.
,
99
,
24369
24381
.
Huffman
,
G. J.
, and
Coauthors
.
1997
:
The Global Precipitation Climatology Project (GPCP) combined precipitation dataset.
Bull. Amer. Meteor. Soc.
,
78
,
5
20
.
Jaeger
,
L.
,
1976
:
Monatskarten des Niedershlags für die ganze Erde.
Ber. Dtsch. Wetterdienstes
,
139
,
1
33
.
Joyce
,
R.
, and
P. A.
Arkin
,
1997
:
Improved estimates of tropical and subtropical precipitation using the GOES precipitation index.
J. Atmos. Oceanic Technol.
,
14
,
997
1011
.
Keihm
,
S. J.
,
M. A.
Janssen
, and
C.
Ruf
,
1995
:
TOPEX/Poseidon Microwave Radiometer (TMR): III Wet troposphere range correction algorithm and prelaunch error budget.
IEEE Trans. Geosci. Remote Sens.
,
33
,
147
161
.
Kummerow
,
C.
, and
Coauthors
.
2000
:
The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit.
J. Appl. Meteor.
,
39
,
1965
1982
.
Legates
,
D. R.
, and
C. J.
Willmott
,
1990
:
Mean seasonal and spatial variability in gauge-corrected, global precipitation.
Int. J. Climatol.
,
10
,
111
127
.
Matsuura
,
T.
, and
S.
Iizuka
,
2000
:
Zonal migration of the Pacific warm-pool tongue during El Niño events.
J. Phys. Oceangr.
,
30
,
1582
1600
.
McMillan
,
A. C.
,
G. D.
Quartly
,
M. A.
Srokosz
, and
J.
Tournadre
,
2002
:
Validation of the TOPEX rain algorithm: Comparison with ground-based radar.
J. Geophys. Res.
,
107
.
.
4038, doi: 10.1029/2001JD000872
.
McPhaden
,
M. J.
,
1999
:
Genesis and evolution of the 1997–98 El Niño.
Science
,
283
,
950
954
.
Olsen
,
R. L.
,
D. V.
Rogers
, and
D. B.
Hodge
,
1978
:
The aRb relation in the calculation of rain attenuation.
IEEE Trans. Antennas Propag.
,
26
,
318
329
.
Petty
,
G. W.
,
1995
:
Frequencies and characteristics of global oceanic precipitation from shipboard present-weather reports.
Bull. Amer. Meteor. Soc.
,
76
,
1593
1616
.
Quartly
,
G. D.
,
T. H.
Guymer
, and
M. A.
Srokosz
,
1996
:
The effects of rain on TOPEX radar altimeter data.
J. Atmos. Oceanic Technol.
,
13
,
1209
1229
.
Quartly
,
G. D.
,
M. A.
Srokosz
, and
T. H.
Guymer
,
1999
:
Global precipitation statistics from dual-frequency TOPEX altimetry.
J. Geophys. Res.
,
104
,
31489
31516
.
Quartly
,
G. D.
,
M. A.
Srokosz
, and
T. H.
Guymer
,
2000
:
Changes in oceanic precipitation during the 1997–98 El Niño.
Geophys. Res. Lett.
,
27
,
2293
2296
.
Simpson
,
J.
,
R. F.
Adler
, and
G. R.
North
,
1988
:
A proposed Tropical Rainfall Measuring Mission (TRMM) satellite.
Bull. Amer. Meteor. Soc.
,
69
,
278
295
.
Spencer
,
R. W.
,
1993
:
Global oceanic precipitation from the MSU during 1979–1991 and comparisons to other climatologies.
J. Climate
,
6
,
1301
1326
.
Tao
,
W-K.
, and
Coauthors
.
2000
:
Meeting summary: Summary of a symposium on cloud systems, hurricanes, and TRMM:Celebration of Dr. Joanne Simpson’s career—The first fifty years.
Bull. Amer. Meteor. Soc.
,
81
,
2463
2474
.
Tournadre
,
J.
, and
J. C.
Morland
,
1997
:
The effects of rain on TOPEX/Poseidon altimeter data.
IEEE Trans. Geosci. Remote Sens.
,
35
,
1117
1135
.
Trenberth
,
K. E.
,
1991
:
Storm tracks in the Southern Hemisphere.
J. Atmos. Sci.
,
48
,
2159
2178
.
Wilheit
,
T.
, and
Coauthors
.
1994
:
Algorithms for the retrieval of rainfall from passive microwave measurements.
Remote Sens. Rev.
,
11
,
163
194
.
Witter
,
D. L.
, and
D. B.
Chelton
,
1991
:
A Geosat altimeter wind speed algorithm and a method for altimeter wind speed algorithm development.
J. Geophys. Res.
,
96
,
8853
8860
.
Xie
,
P.
, and
P. A.
Arkin
,
1997
:
Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs.
Bull. Amer. Meteor. Soc.
,
78
,
2539
2558
.

Footnotes

Corresponding author address: Dr. Ge Chen, Ocean Remote Sensing Institute, Ocean University of China, 5 Yushan Road, Qingdao 266003, China. Email: gechen@public.qd.sd.cn