Agreement between monthly precipitation estimates from TRMM satellite, NCEP reanalysis, and merged gauge satellite analysis

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2010jd015483, 2011 Agreement between monthly precipitation estimates from TRMM satellite, NCEP reanalysis, and merged gauge satellite analysis Dong Bin Shin, 1 Ju Hye Kim, 1 and Hyo Jin Park 1 Received 13 December 2010; revised 22 March 2011; accepted 23 May 2011; published 17 August [1] Global monthly precipitation is a critical element in understanding variability of the Earth s climate including changes in the hydrological cycle associated with global warming. The NCEP reanalysis (R1), GPCP, CMAP, and TMPA precipitation data sets are often used in climate studies. This study compares the data sets (R1, GPCP, CMAP, and TMPA) with the TRMM precipitation data sets derived from the TRMM precipitation radar (TPR), microwave imager (TMI), and combined algorithm (TCA) for 11 years ( ) over the satellite s domain (40 S 40 N). The domain precipitation estimates from seven data sets range from 2.44 to 3.38 mm d 1 over the ocean and from 1.98 to 2.83 mm d 1 over land. The regional differences between the TPR and the other data sets are analyzed by a paired t test. Particularly, statistically significant differences between TPR and GPCP and between TPR and CMAP are found in most oceanic regions and in some land areas. In general, there exists substantial disagreement in precipitation intensities from the precipitation data sets. Therefore, significant consideration is given to the uncertainties in the data sets prior to applying the results to climate studies such as estimations of the global hydrological budget analyses. Meanwhile, the anomalies from all the data sets agree relatively well in their variability patterns. It is also found that the dominant mode of interannual variability which is associated with the ENSO pattern is clearly demonstrated by all precipitation data sets. These results suggest that all considered precipitation data sets may produce similar results when they are used for climate variability analyses on annual to interannual time scales. Citation: Shin, D. B., J. H. Kim, and H. J. Park (2011), Agreement between monthly precipitation estimates from TRMM satellite, NCEP reanalysis, and merged gauge satellite analysis, J. Geophys. Res., 116,, doi: /2010jd Introduction [2] Precipitation plays an important role in the Earth s climate system. Its variations directly affect the intensity of the global water cycle, thus modifying the global energy balance as precipitation releases latent heat to the atmosphere. In particular, latent heat released from tropical precipitation is a major driver of large scale atmospheric circulation. Accurate measurements or estimations of precipitation are therefore crucial for advancing our understanding of climate system variability and improving climate predictions. Global precipitation has been estimated from low Earth orbiting and geostationary satellites using various microwave, infrared and visible sensors, with the use of microwave sensors yielding the most accurate precipitation measurements. Microwave sensors can measure thermal emission from rain within and below clouds, which allows direct estimation of the surface rain rate particularly over the ocean. For this reason, a series of microwave sensors have been developed beginning with the 1972 deployment of the Electrically Scanning Microwave Imager on board the Nimbus 5 satellite. 1 Department of Atmospheric Sciences, Yonsei University, Seoul, South Korea. Copyright 2011 by the American Geophysical Union /11/2010JD [3] The global coverage afforded today by satellites fitted with various sensors allows us to generate precipitation data sets by blending different measurement sources. Global Precipitation Climatology Project (GPCP) data [Adler et al., 2003] and the National Oceanic and Atmospheric Administration s (NOAA s) Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) [Xie and Arkin, 1997] are examples of data sets created by blending data from a variety of satellites and gauges. However, as reported by Yin et al. [2004] and Trenberth et al. [2007], the two data sets can yield different conclusions especially related to climate trends. Such differences suggest that the characteristics of differences between precipitation data sets be carefully understood before investigating climate system variability from a specific data set. Monthly precipitation data are frequently used to understand the global water cycle, its variability and its relationships with many climate variables. This study compares the differences in the mean and anomaly fields of seven monthly precipitation data sets which are based on different sensors and methodologies. 2. Monthly Rainfall Data [4] Three instantaneous (level 2) surface rainfall data sets (so called 2A12, 2A25, and 2B31 in TRMM data classification) from the TRMM precipitation radar (TPR) and 1of8

2 microwave imager (TMI) are used in this study. The rainfall data from the TMI (2A12) are retrieved from the Goddard profiling (GPROF) algorithm [Kummerow et al., 2001]. The GPROF uses three dimensional cloud models to construct a priori databases of hydrometeor profiles and their computed brightness temperatures over the oceans. Using a Bayesian inversion technique, the TMI observed brightness temperatures are inverted to the 2A12 rainfalls with a spatial resolution of about 12 km. Over land, the 2A12 rainfalls are derived from the empirical relationships with the high frequency brightness temperatures [McCollum and Ferraro, 2003]. The TPR rainfall algorithm (2A25) uses attenuation corrected radar reflectivity at 13.8 GHz for rainfall estimates. The details of the TPR algorithm are described by Iguchi et al. [2000]. The TRMM combined algorithm (TCA or 2B31) provides the vertical structure of rainfall (rates and drop size distribution parameters) based upon the TMI and TPR within the TPR swath. The radar measurements for every likely value of the drop size distribution (DSD) shape parameters are first inverted. The resulting rainfall estimates are used to produce the corresponding expected brightness temperatures, which are then compared to the actual passive measurements to select the most probable DSD shape parameter value [Haddad et al., 1997]. [5] To facilitate analysis, these level 2 rainfall data are binned at 24 hourly grids at latitude longitude grids. The product is called 3G68 in TRMM data classification. Because the 3G68 data include the number of total pixels (N) and the number of raining pixels as well as unconditional hourly rain rates (R), average rain rates for a certain period and grids can be calculated with the resolution of the multiple of 0.5 as follows, hri ¼ P K P M i¼1 j¼1 P K P M N ij i¼1 j¼1 R ij N ij where the angular bracket denotes a temporal and spatial average. K indicates the number of hours for a certain period and M implies the number of subgrids (0.5 ) within a grid with the resolution of the multiple of 0.5. Monthly areaaveraged rain rates of three TRMM level 2 data for latitude longitude grid boxes are computed from the resolution 3G68 data. [6] The National Centers for Environmental Prediction/ National Center for Atmospheric Research (NCEP/NCAR) reanalysis data (R1) [Kalnay et al., 1996] are also used in this study. The NCEP reanalysis data provides a long term global model analysis of basic atmospheric quantities. The data are based on the NCEP global operational model (T62 with 28 vertical levels) and additional observations including the global rawinsonde data, Comprehensive Ocean Atmosphere Data Set (COADS) surface marine data, aircraft data, Global Telecommunications System (GTS) data, satellite sounder data, Special Sensor Microwave Imager (SSM/I) surface wind speeds, and satellite cloud drift winds. The NCEP reanalysis data are available at 17 pressure levels (1000, 925, 850, 700, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 20, and 10 hpa) at 6 hourly temporal resolution. ð1þ Surface flux parameters including precipitation rate are available at a Gaussian grid of (192 94). [7] This study also compares monthly mean precipitation over latitude longitude grids from the Global Precipitation Climatology Product (GPCP) and Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data which are estimated by merging satellitederived and in situ global precipitation data. A distinct difference of the two data sets can be found in the use of the satellite derived precipitation components. Over the ocean, the GPCP uses the Special Sensor Microwave Imager (SSM/I) emission based precipitation estimates [Wilheit et al., 1991; Chiu and Chokngamwong, 2010], Geostationary Operational Environmental Satellite (GOES) Precipitation Index (GPI) [Janowiak and Arkin, 1991], Television Infrared Observation Sounder (TIROS) Operational Vertical Sounder (TOVS) [Susskind et al., 1997] and Outgoing Longwave Radiation (OLR) Precipitation Index (OPI) [Xie and Arkin, 1998]. The GPI and SSM/I emission based precipitation estimates are also used in CMAP, but TOVS is not incorporated. Instead, CMAP includes SSM/I scattering based precipitation estimates [Ferraro and Marks, 1995] and the Microwave Sounding Unit (MSU) [Spencer, 1993]. The merging techniques of GPCP and CMAP are also different. The GPCP derived microwave/ir calibration ratios and applied the ratios to GPI estimates. The multisatellite estimates are combined with rain gauge data over land by inverse variance weighting [Huffman et al., 1997]. The CMAP merges satellite estimates using a maximum likelihood estimation with weights which are inversely proportional to the squares of the individual errors. The land and atoll gauge data adjustment is then performed [Xie and Arkin, 1997; Yin et al., 2004]. [8] This study also includes the TRMM Multisatellite Precipitation Analysis (TMPA) data set. The TMPA is a combined product of multiple satellites and rain gauge data sets. It provides fine scales estimates, latitudelongitude grids and 3 hourly [Huffman et al., 2007]. The TMPA uses passive microwave data from TMI, SSM/I, Advanced Microwave Scanning Radiometer Earth Observing System (AMSR E) and Advanced Microwave Sounding Unit B (AMSU B). The CPC merged infrared (IR) data and GPCP IR histogram, TCA estimate, GPCP monthly rain gauge analysis and Climate Assessment and Monitoring System (CAMS) monthly rain gauge analysis are also used as data sources. 3. Agreement Analyses 3.1. Domain Means and Anomalies [9] The time series of domain mean precipitation obtained from grid mean precipitation data over the ocean and land for the period between 1998 and 2008 are shown in Figure 1. Since the TRMM observation data are collected within the region from 40 S to 40 N, all analyses in this study are limited to the TRMM observation area. The TRMM domain means of precipitation from TPR, TMI, TCA, R1, GPCP, CMAP, and TMPA are 2.44, 2.67, 2.70, 3.36, 3.02, 3.38, and 2.80 mm d 1 over the ocean. The three satellite derived precipitation data sets tend to have lower rain rates than the multisatellite and reanalysis data products. Estimates of domain land precipitation for the seven data sets are 1.98, 2of8

3 Figure 1. Time series of monthly domain mean rain rates for three TRMM rainfall data sets (TPR, TMI, and TCA), NCEP reanalysis data (R1), GPCP, CMAP, and TMPA products over (a) the ocean and (b) land. 2.59, 2.42, 2.83, 2.69, 2.43, and 2.44 mm d 1. The strength of the relationship between the domain precipitation means over the ocean and land is illustrated by the correlation coefficients (Table 1). Over the ocean, the relatively strong correlations are found between TPR and TCA (0.86) and between TPR and TMPA (0.65), but correlations between TPR and the other four data sets are much weaker. There are also noticeable correlations between TMI and TCA (0.63), between TMI and GPCP (0.83), between TMI and CMAP (0.70), and between GPCP and CMAP (0.78). The correlation in mean precipitation is typically much greater in the matched data sets from over land than in those from over the ocean, except for the much weaker correlations between TMI and GPCP as well as TMI and CMAP. Chiu et al. [2006] also reported the TRMM domain averages from TPR, TMI, TCA and TMPA for are 2.61 (2.07), 2.63 (2.64), 2.84 (2.49) and 2.86 (2.45) mm d 1 over the ocean (land), respectively. The average precipitation intensities are slightly different from those of this study due to the different averaging periods. The relative differences in precipitation intensity of the data sets obtained from this study, however, are consistent with those reported by Chiu et al. [2006]. [10] The anomalies relative to the averages of the period between 1998 and 2008 over the ocean and land are shown in Figure 2. Figure 2 shows that the anomalies from all the precipitation data sets agree generally well in their variability patterns. It appears that the magnitudes of the anomaly variations are significantly lower over the ocean than over land (Table 2). The larger fluctuations may be associated Table 1. Correlation Matrices for the Domain Monthly Mean Data Sets Over the Ocean and Land a Ocean (Land) TPR TMI TCA R1 GPCP CMAP TMPA TPR 1.00 (1.00) 0.49 (0.76) 0.86 (0.96) 0.38 (0.66) 0.29 (0.64) 0.32 (0.58) 0.65 (0.56) TMI 1.00 (1.00) 0.63 (0.82) 0.36 (0.37) 0.83 (0.34) 0.70 (0.32) 0.25 (0.32) TCA 1.00 (1.00) 0.38 (0.62) 0.46 (0.59) 0.50 (0.53) 0.49 (0.48) R (1.00) 0.39 (0.63) 0.52 (0.58) 0.20 (0.48) GPCP 1.00 (1.00) 0.78 (0.92) 0.15 (0.86) CMAP 1.00 (1.00) 0.11 (0.86) TMPA 1.00 (1.00) a Land values are in parentheses. 3of8

4 Figure 2. Time series of rainfall anomalies relative to the period between 1998 and 2008 for seven precipitation data sets. 4of8

5 Table 2. Variances of the Anomaly Data Sets Over the Ocean and Land Ocean Land TPR TMI TCA R GPCP CMAP TMPA correlated. The correlation constants between the SOI and TRMM precipitation products, TPR, TMI, and TCA, are 0.75, 0.71, and 0.76, respectively. The PCTSs of GPCP, CMAP, and TMPA data sets show relatively high correlations of 0.78, 0.80, and 0.77, respectively. The lowest cor- with great variability in precipitation pattern over land illustrating the difficulty of the retrievals of land precipitation [Shin et al., 2001]. We can also note that the variance of the TPR anomaly ( mm 2 d 2 ) is significantly lower than those of the other data sets ( mm 2 d 2 ), especially over land. The different variability may be due to differences in sensor characteristics. That is, the rain intensity measured by the PR is proportional to the pathintegrated attenuation (PIA) which is mainly affected by the raindrop size distribution along the radar beam [Iguchi et al., 2000].Meanwhile,infraredandmicrowaveradiometric signatures used in the merged rainfall products tend to be affected by highly variable land emissivity EOF and PCTS [11] The principal components or empirical orthogonal functions (EOFs) can be obtained by diagonalizing the covariance matrix calculated by subtracting out the mean state from a multivariate data set. The resulting eigenvectors are orthogonal to each other and each eigenvector is associated with a positive eigenvalue which indicates the amount of variance explained by the eigenvector field. For this reason, EOF or principal component analysis has been widely used to investigate climate variability and related physical modes [e.g., North, 1984; Dommenget and Latif, 2002]. [12] We compare the EOF patterns of seven precipitation data sets. The maps associated with the first EOF of each data set are shown in Figure 3. The variances explained by the first EOFs are 8.3%, 10.5%, 8.8%, 8.5%, 10.5%, 12.7%, and 10.7% for TPR, TMI, TCA, R1, GPCP, CMAP, and TMPA, respectively. The common pattern is characterized by the opposition of precipitation variability between the region extending from the western and south Pacific and the area in the equatorial Pacific with a branch extending into the south Pacific. [13] This dipole pattern in the Pacific typically matches the precipitation variability associated with the El Niño Southern Oscillation (ENSO) event [e.g., Janowiak and Arkin, 1991;Arkin et al., 1994; Huffman et al., 1997;Dai and Wigley, 2000; Adler et al., 2003]. Thus, the principal component time series (PCTSs) of the first eigenvectors and the southern oscillation index (SOI) are compared in Figure 4. The SOI, which is defined by the difference in sea level pressure between Darwin and Tahiti, is usually used as an indicator of the onset of El Niño and La Niña episodes. Each PCTS is statistically independent of the others and the sum of variances of all PCTSs is identical to the total variance. The seven PCTSs show similar fluctuating patterns. It appears that the SOI and each PCTS are well Figure 3. The first EOFs of seven precipitation data sets. Numbers in parentheses indicate the variances explained by the first EOFs and the correlations with TPR. 5of8

6 Figure 4. The principal component time series (PCTS) of the first EOF mode with the SOI time series. Numbers in parentheses indicate the correlations between SOI and each data set. relation coefficient (0.68) is found between the SOI and the PCTS of the R1 data set Maps of Agreement [14] The regional differences between the precipitation data sets are investigated in this section. The significance of the difference between two data sets is tested using a paired t test [Chang et al., 1999; Shin et al., 2001]. The t statistic for the paired rain rates are defined as t ¼ x y. xy p ffiffiffiffi N where x and y are the ensemble averages of the monthly mean data from two different data sets, s xy is the standard deviation of x y and N is the number of pairs. [15] Figure 5 shows the distribution of the t statistic for the pairs of TPR and the other five data sets (TPR TMI, TPR TCA, TPR GPCP, TPR CMAP, and TPR TMPA). The R1 data set is not included in the paired t test as the spatial resolution of its data ( ) differs from that of the other data sets. For a two tailed t test using the 132 pairs of the monthly mean rain rates, t values larger than 1.98 are significant at the 95% level and the null hypothesis that the two means are equal is rejected at the p = 0.05 level of significance. Figure 5 illustrates that TPR TMI pairs (Figure 5a) seem to show significant negative t values (TMI is larger) in high rain regions. Significant positive t values (TPR is larger) are found over some of the oceanic dry regions, such as the southeastern Pacific. The differences may be explained by earlier results [Berg et al., 2010; Stephens et al., 2010]. Berg et al. [2010], based on the CloudSat Profiling Radar (CPR) and PR rainfall comparisons, reported that the amount of light rain missed by the PR is highest over the convergence zones, but it accounts for a far larger fraction of the total rainfall over the subsidence zones. Both the studies showed that the PR based products are biased low in estimation of precipitation by about 10% compared to the CPR measured rainfalls due to the lower sensitivity of the PR to the light rain. ð2þ [16] The t values between TPR and TCA are generally small and statistically insignificant except over small oceanic dry regions. The distribution of t values between TPR and TCA is similar to that of TPR and TCA with positive t values over small coastal regions. The TPR GPCP and TPR CMAP pairs show patterns of positive t values similar to that of the TPR TMI pair. However, unlike the TPR TMI pair, the pairs tend to show significant negative t values over most oceanic regions. This is consistent with the GPCP and CMAP rainfalls being significantly higher than what is reported by TPR in oceanic regions. 4. Summary and Conclusions [17] This study examined differences in the mean and anomaly fields of seven monthly precipitation data sets derived from the TRMM satellite (TPR, TMI, and TCA), NCEP reanalysis (R1), and merged analysis of gauge satellite (GPCP, CMAP, and TMPA) for the 11 year period from 1998 to The domain means of oceanic precipitation from the TRMM satellite (TPR, TMI, and TCA) range from 2.44 to 2.70 mm d 1. The estimates from reanalysis and merged gauge satellite analysis (R1, GPCP, CMAP, and TMPA) tend to be higher than those from the TRMM satellite, ranging from 2.80 to 3.38 mm d 1. Most disparately, the TPR provides the lowest domain precipitation estimate (1.98 mm d 1 ) over land, but differences among the other six data sets are less significant. The precipitation anomaly plots show that all the data sets agree relatively well in their variability patterns. This study also performed an EOF analysis for the seven precipitation data sets. It appears that the dominant mode of interannual variability indicated by the first EOF and PCTS of each data set is clearly associated with the ENSO pattern with relatively high correlations ranging from 0.68 (R1) to 0.80 (CMAP). According to a paired t test, major differences in monthly precipitation estimates exist for TPR GPCP and TPR CMAP in most of the oceanic regions and some land areas such as over the North American and African continents. It is also found that monthly precipitation estimates from the TRMM satellite, particularly TPR and TMI, produce statistically sig- 6of8

7 Figure 5. Maps of t statistics from a paired t test. nificant differences in the oceanic dry regions, northern African, and southern Australian continents. [18] The agreement analyses of precipitation intensities in this study indicate that there remains a challenge in making consistent precipitation estimates from different sensors and merging methods due to the uncertainties introduced by employing various sensor assumptions and different data merging methodologies. However, similar precipitation 7of8

8 anomalies on annual to interannual timescales suggest that certain climate variability analyses on the time scales may not be greatly influenced by a specific data set. [19] Acknowledgment. This work was funded by the Research Agency for Climate Studies under grant RACS References Adler, R. F., et al. (2003), The version 2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979 present), J. Hydrometeorol., 4(6), , doi: / (2003)004<1147: TVGPCP>2.0.CO;2. Arkin, P. A., R. Joyce, and J. E. Janowiak (1994), IR techniques: GOES precipitation index, Remote Sens. Rev., 11, Berg, W., T. L Ecuyer, and J. M. Haynes (2010), The distribution of rainfall over oceans from space borne radars, J. Appl. Meteorol. Climatol., 49(3), , doi: /2009jamc Chang, A. T. C., L. S. Chiu, C. Kummerow, J. Meng, and T. T. Wilheit (1999), First results of the TRMM microwave imager (TMI) monthly oceanic rain rate: Comparison with SSM/I, Geophys. Res. Lett., 26(15), , doi: /1998gl Chiu, L. S., and R. Chokngamwong (2010), Microwave emission brightness temperature histogram (METH) rain rates for climate studies: Remote Sensing Systems SSM/I version 6 results,j. Appl. Meteorol. Climatol., 49(1), , doi: /2009jamc Chiu, L., D. B. Shin, and J. Kwaitkowski (2006), Surface rain rates from tropical rainfall measuring mission satellite algorithms, in Earth Science Satellite Remote Sensing, vol. 1, edited by J. J. Qu et al., pp , Springer, Berlin, doi: / _17. Dai, A., and T. M. L. Wigley (2000), Global patterns of ENSO induced precipitation, Geophys. Res. Lett., 27, , doi: / 1999GL Dommenget, D., and M. Latif (2002), A cautionary note on the interpretation of EOFs, J. Clim., 15(2), , doi: / (2002) 015<0216:ACNOTI>2.0.CO;2. Ferraro, R. R., and G. F. Marks (1995), The development of SSM/I rain rate retrieval algorithms using ground based radar measurements, J. Atmos. Oceanic Technol., 12, , doi: / (1995)012<0755: TDOSRR>2.0.CO;2. Haddad, Z. S., D. A. Short, S. L. Durden, E. Im, S. Hensley, M. B. Grable, and R. A. Black (1997), A new parameterization of the rain drop size distribution, IEEE Trans. Geosci. Remote Sens., 35(3), , doi: / Huffman, G. J., et al. (1997), The Global Precipitation Climatology Project (GPCP) combined precipitation dataset, Bull. Am. Meteorol. Soc., 78, 5 20, doi: / (1997)078<0005:tgpcpg>2.0.co;2. Huffman, G. J., et al. (2007), The TRMM multisatellite precipitation analysis (TMPA): Quasi global, multiyear, combined sensor precipitation estimates at fine scales, J. Hydrometeorol., 8, 38 55, doi: / JHM Iguchi, T., T. Kozu, R. Meneghini, J. Awaka, and K. Okamoto (2000), Rain profiling algorithm for the TRMM precipitation radar, J. Appl. Meteorol., 39(12), , doi: / (2001)040<2038: RPAFTT>2.0.CO;2. Janowiak, J. E., and P. A. Arkin (1991), Rainfall variations in the tropics during as estimated from observations of cloud top temperature, J. Geophys. Res., 96, Kalnay, E., et al. (1996), The NCEP/NCAR 40 year reanalysis project, Bull. Am. Meteorol. Soc., 77(3), , doi: / (1996)077<0437:TNYRP>2.0.CO;2. Kummerow, C., et al. (2001), The evolution of the Goddard profiling algorithm (GPROF) for rainfall estimation from passive microwave sensors, J. Appl. Meteorol., 40(11), , doi: / (2001) 040<1801:TEOTGP>2.0.CO;2. McCollum, J. R., and R. R. Ferraro (2003), Next generation of NOAA/ NESDIS TMI, SSM/I, and AMSR E microwave land rainfall algorithms, J. Geophys. Res., 108(D8), 8382, doi: /2001jd North, G. R. (1984), Empirical orthogonal functions and normal modes, J. Atmos. Sci., 41, , doi: / (1984)041<0879: EOFANM>2.0.CO;2. Shin, D. B., L. S. Chiu, and M. Kafatos (2001), Comparison of the monthly precipitation derived from the TRMM satellite, Geophys. Res. Lett., 28(5), , doi: /2000gl Spencer, R. W. (1993), Global oceanic precipitation from the MSU during and comparisons to other climatologies, J. Clim., 6, , doi: / (1993)006<1301:gopftm>2.0.co;2. Stephens, G. L., T. L Ecuyer, R. Forbes, A. Gettlemen, J. C. Golaz, A. Bodas Salcedo, K. Suzuki, P. Gabriel, and J. Haynes (2010), Dreary state of precipitation in global models, J. Geophys. Res., 115, D24211, doi: /2010jd Susskind, J., P. Piraino, L. Rokke, L. Iredell, and A. Mehta (1997), Characteristics of the TOVS Pathfinder Path A dataset, Bull. Am. Meteorol. Soc., 78(7), , doi: / (1997)078<1449: COTTPP>2.0.CO;2. Trenberth, K. E., L. Smith, T. Qian, A. Dai, and J. Fasullo (2007), Estimates of the global water budget and its annual cycle using observational and model data, J. Hydrometeorol., 8(4), , doi: / JHM Wilheit, T. T., A. T. C. Chang, and L. S. Chiu (1991), Retrieval of monthly rainfall indices from microwave radiometric measurements using probability distribution functions, J. Atmos. Oceanic Technol., 8, , doi: / (1991)008<0118:romrif>2.0.co;2. Xie, P. P., and P. A. Arkin (1997), Global precipitation: A 17 year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs, Bull. Am. Meteorol. Soc., 78(11), , doi: / (1997)078<2539:GPAYMA>2.0.CO;2. Xie, P. P., and P. A. Arkin (1998), Global monthly precipitation estimates from satellite observed outgoing longwave radiation, J. Clim., 11(2), , doi: / (1998)011<0137:gmpefs>2.0.co;2. Yin, X. G., A. Gruber, and P. Arkin (2004), Comparison of the GPCP and CMAP merged gauge satellite monthly precipitation products for the period , J. Hydrometeorol., 5(6), , doi: / JHM J. H. Kim, H. J. Park, and D. B. Shin, Department of Atmospheric Sciences, Yonsei University, Seoul , South Korea. (dbshin@yonsei. ac.kr) 8of8