• Adler, R. F., A. J. Negri, P. R. Keehn, and I. M. Hakkarinen, 1993: Estimation of monthly rainfall over Japan and surrounding waters from a combination of low orbit microwave and geosynchronous IR data. J. Appl. Meteor., 32 , 335356.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., G. J. Huffman, and P. R. Keehn, 1994: Global rain estimates from microwave adjusted geosynchronous IR data. Remote Sens. Rev., 11 , 125152.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., C. Kidd, G. Petty, M. Morrissey, and H. M. Goodman, 2001: Intercomparison of global precipitation products: The third Precipitation Intercomparison Project (PIP-3). Bull. Amer. Meteor. Soc., 82 , 13771396.

    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and B. N. Meisner, 1987: The relationship between large-scale convective rainfall and cold cloud over the Western Hemisphere during 1982–1984. Mon. Wea. Rev., 115 , 5174.

    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and P. Xie, 1994: The Global Precipitation Climatology Project: First Algorithm Intercomparison Project. Bull. Amer. Meteor. Soc., 75 , 401419.

    • Search Google Scholar
    • Export Citation
  • Chen, M., P. Xie, J. E. Janowiak, and P. A. Arkin, 2002: Global land precipitation: A 50-year monthly analysis based on gauge observations. J. Hydrometeor., 3 , 249266.

    • Search Google Scholar
    • Export Citation
  • Curtis, S., and R. F. Adler, 2000: ENSO indices based on patterns of satellite derived precipitation. J. Climate, 13 , 27862793.

  • Dai, A., T. M. L. Wigley, B. A. Boville, J. T. Kiehl, and L. E. Buja, 2001: Climates of the twentieth and twenty-first centuries simulated by the NCAR climate system model. J. Climate, 14 , 485519.

    • Search Google Scholar
    • Export Citation
  • Ebert, E. E., and M. J. Manton, 1998: Performance of satellite rainfall estimation algorithms during TOGA COARE. J. Atmos. Sci., 55 , 15371557.

    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., M. Kanamitsu, J. Potter, and M. Fiorino, 1998: An overview of Reanalysis-2. Preprints, 23d Annual Climate Diagnostics Workshop, Miami, FL, Climate Prediction Center, 119–120.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., 1997: Special Sensor Microwave Imager derived global rainfall estimates for climatological applications. J. Geophys. Res., 102 , 1671516735.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., and A. F. Krueger, 1984: The status of the NOAA outgoing longwave radiation data set. Bull. Amer. Meteor. Soc., 65 , 958962.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., X. Su, M. Kanamitsu, and J. Schemm, 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., W. Shi, E. Yarosh, and R. Joyce, 2000: Improved United States precipitation quality control system and analysis. Climate Prediction Center Atlas, No. 7, National Oceanic and Atmospheric Administration, National Weather Service. [Available from NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, MD 20746.].

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., R. F. Adler, B. R. Rudolf, U. Schneider, and P. R. Keehn, 1995: Global precipitation estimates based on a technique for combining satellite-based estimates, rain gauge analysis, and NWP model precipitation information. J. Climate, 8 , 12841295.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and Coauthors. 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc., 78 , 520.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observations. J. Hydrometeor., 2 , 3650.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., 1992: Tropical rainfall: A comparison of satellite derived rainfall estimates with model precipitation forecasts, climatologies, and observations. Mon. Wea. Rev., 120 , 448462.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., P. A. Arkin, P. Xie, M. L. Morrissey, and D. R. Legates, 1995: An examination of the east Pacific ITCZ rainfall distribution. J. Climate, 8 , 28102823.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., A. Gruber, C. R. Kondragunta, R. E. Livezey, and G. J. Huffman, 1998: A comparison of the NCEP–NCAR reanalysis precipitation and the GPCP rain gauge–satellite combined dataset with observational error considerations. J. Climate, 11 , 29602979.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and P. H. Chan, 1986: Aspects of the 40–50-day oscillation during the northern summer as inferred from outgoing longwave radiation. Mon. Wea. Rev., 114 , 13541367.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and H. T. Wu, 2001: Principal modes of rainfall–SST variability of the Asian summer monsoon: A reassessment of the monsoon–ENSO relationship. J. Climate, 14 , 28802895.

    • Search Google Scholar
    • Export Citation
  • Morrissey, M. L., M. A. Shafer, S. E. Postawko, and B. Gibson, 1995: Pacific raingauge data. Water Resour. Res., 31 , 21112113.

  • Qian, W., and S. Yang, 2000: Onset of the regional monsoon over Southeast Asia. Meteor. Atmos. Phys., 75 , 2938.

  • Reynolds, R. W., 1988: A real-time global sea surface temperature analysis. J. Climate, 1 , 7586.

  • Roads, J. O., S. C. Chen, and F. Fujioka, 2001: ECPC's weekly to seasonal global forecasts. Bull. Amer. Meteor. Soc., 82 , 639658.

  • Schneider, U., 1993: The GPCC quality-control system for gauge-measured precipitation data. GEWEX Workshop on Analysis Methods of Precipitation on Global Scale, Rep. WCRP-81, WMP/TD-588, Koblenz, Germany, WMO, A5–A9.

    • Search Google Scholar
    • Export Citation
  • Shepard, D., 1968: A two dimensional interpolation function for regularly spaced data. Proc. 23d National Conf. of American Computing Machinery, Princeton, NJ, Association for Computing Machinery, 517–524.

    • Search Google Scholar
    • Export Citation
  • Shi, W., R. W. Higgins, E. Yarosh, and V. E. Kousky, cited. 2001: The annual cycle and variability of precipitation in Brazil. NCEP/Climate Prediction Center Atlas, No. 9, National Oceanic and Atmospheric Administration. National Weather Service. [Available online at http://www.cpc.noaa.gov/research_papers/ncep_cpc_atlas/9/index.html.].

    • Search Google Scholar
    • Export Citation
  • Spencer, R. W., 1993: Global oceanic precipitation from MSU during 1979–91 and comparisons to other climatologies. J. Climate, 6 , 13011326.

    • Search Google Scholar
    • Export Citation
  • Sperber, K. R., J. M. Slingo, P. M. Inness, and W. K-M. Lau, 1997: On the maintenance and initiation of the intraseasonal oscillation in the NCEP/NCAR reanalysis and the GLA and UKMO AMIP simulations. Climate Dyn., 13 , 769795.

    • Search Google Scholar
    • Export Citation
  • Stephenson, D. B., F. Chauvin, and J-F. Royer, 1998: Simulation of the Asian summer monsoon and its dependence on model horizontal resolution. J. Meteor. Soc. Japan, 76 , 237265.

    • Search Google Scholar
    • Export Citation
  • Susskind, J., P. Piraino, L. Rokkle, L. Iredell, and A. Mehta, 1997: Characteristics of the TOVS Pathfinder Path A dataset. Bull. Amer. Meteor. Soc., 78 , 14491472.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and C. J. Guillemott, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP reanalysis. Climate Dyn., 14 , 213231.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and J. M. Caron, 2000: The Southern Oscillation revisited: Sea level pressures, surface temperatures, and precipitation. J. Climate, 13 , 43584365.

    • Search Google Scholar
    • Export Citation
  • Waliser, D. E., C. Jones, J. K. Schemm, and N. E. Graham, 1999: A statistical extended-range tropical forecast model based on the slow evolution of the Madden–Julian Oscillation. J. Climate, 12 , 19181939.

    • Search Google Scholar
    • Export Citation
  • Weickmann, K. M., G. R. Lussky, and J. E. Kutzbach, 1985: Intraseasonal (30–60 day) fluctuations of outgoing longwave radiation and 250-mb stream function during northern winter. Mon. Wea. Rev., 113 , 941961.

    • Search Google Scholar
    • Export Citation
  • Weng, F-Z., and N. C. Grody, 1994: Retrieval of cloud liquid water using the special sensor microwave imager (SSM/I). J. Geophys. Res., 99 , 2553525551.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. J., A. T. C. Chang, and L. S. Chiu, 1991: Retrieval of the monthly rainfall indices from microwave radiometric measurements using probability distribution functions. J. Atmos. Oceanic Technol., 8 , 118136.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1995: An intercomparison of gauge observations and satellite estimates of monthly precipitation. J. Appl. Meteor., 34 , 11431160.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840858.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1997a: Global Precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1997b: Global pentad precipitation analysis based on gauge observations, satellite estimates and model outputs. Extended Abstracts, Amer. Geophys. Union 1997 Fall Meeting, San Francisco, CA, Amer. Geophys. Union.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1998: Global monthly precipitation estimates from satellite-observed outgoing longwave radiation. J. Climate, 11 , 137164.

    • Search Google Scholar
    • Export Citation
  • Xie, P., B. Rudolf, U. Schneider, and P. A. Arkin, 1996: Gauge-based monthly analysis of global land precipitation from 1971–1994. J. Geophys. Res., 101 (D14) 1902319034.

    • Search Google Scholar
    • Export Citation
  • Yang, S., K-M. Lau, and P. S. Schopf, 1999: Sensitivity of the tropical Pacific Ocean to precipitation-induced freshwater flux. Climate Dyn., 15 , 737750.

    • Search Google Scholar
    • Export Citation
  • Zhou, J., and W. K-M. Lau, 1999: Summertime intraseasonal variability over South America. Preprints, 24th Annual Climate Diagnostics Workshop, Tucson, AZ, Climate Prediction Center, 299–302.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Precipitation (mm day−1) for pentad 41 (20–24 Jul) of 1988 as observed in the satellite estimates of GPI, SSM/I scattering (SCT), SSM/I emission (EMS), OPI, and MSU; the gauge-based analyses; the merged analyses of pentad CMAP (observation-only version); and the pentad GPCP

  • View in gallery

    Distribution of mean precipitation (mm day−1) for a 20-yr period from 1979 to 1998 as defined from (top) the monthly GPCP analyses version 2 dataset, (middle) that from the observation-only version of pentad CMAP, and (bottom) the difference between the two

  • View in gallery

    Correlation between the monthly GPCP analyses and the monthly accumulation of pentad analyses defined by adjusting the original pentad CMAP/O by a ratio calculated over various time–space-averaging domains for a 20-yr period from 1979 to 1998.

  • View in gallery

    Same as in Fig. 3, except for bias (%) relative to the mean value of the monthly GPCP

  • View in gallery

    (top) Correlation, (middle) relative bias (%), and (bottom) relative rms error (%) between the total precipitation in the pentad GPCP and that in the guage-based analyses of Higgins et al. (2000) over the United States for an 18-yr period from 1979 to 1996

  • View in gallery

    Same as in Fig. 5, except for comparison with Shi et al. (2001) over Brazil

  • View in gallery

    Longitudinal profiles of mean annual cycle of precipitation (mm day−1) averaged over (top) the ocean, (middle) land, and (bottom) the entire globe

  • View in gallery

    Time–longitude sections of (left) 20–100-day bandpass-filtered precipitation (mm day−1) and (middle) OLR (W m−2) averaged over 10°S–10°N for the period from Oct 1996 to May 1997. (right) The time series of an index associated with the TISO defined as bandpass-filtered velocity potential averaged over 10°S–10°N, 100°–140°E

  • View in gallery

    Composites of (left) 20–100-day bandpass-filtered precipitation (mm day−1) and (right) OLR (W m−2) defined by dividing a cycle of intraseasonal oscillation into four phases based on the TISO index. Phases 1 and 3 denote time when the TISO index reaches max and min, respectively, while phases 2 and 4 are periods in between. Only cases with max/min index values larger/smaller than 0.75/−0.75 std dev are included in defining the composites

  • View in gallery

    GPCP pentad analysis global distribution of (top left) seasonal mean precipitation (mm day−1) and std dev of precipitation (mm day−1) for (top right) the mean annual cycle, and components associated with (bottom left) interannual and (bottom right) intraseasonal components for the DJF season

  • View in gallery

    Same as in Fig. 10, except for the OLR (W m−2) observed by the NOAA satellites

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 412 381 31
PDF Downloads 227 200 19

GPCP Pentad Precipitation Analyses: An Experimental Dataset Based on Gauge Observations and Satellite Estimates

View More View Less
  • * NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, Maryland
  • | + NOAA/OAR Office of Global Programs, Silver Spring, Maryland
  • | # Laboratory for Atmospheres, NASA Goddard Space Flight Center, Greenbelt, Maryland
  • | @ NOAA/NESDIS Office of Research and Application, Camp Springs, Maryland
  • | & Science Systems and Applications, Inc., Lanham, Maryland
  • | ** Joint Center for Earth Systems Technology, University of Maryland, Baltimore County, Baltimore, Maryland
© Get Permissions
Full access

Abstract

As part of the Global Precipitation Climatology Project (GPCP), analyses of pentad precipitation have been constructed on a 2.5° latitude–longitude grid over the globe for a 23-yr period from 1979 to 2001 by adjusting the pentad Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) against the monthly GPCP-merged analyses. This adjustment is essential because the precipitation magnitude in the pentad CMAP is not consistent with that in the monthly CMAP or monthly GPCP datasets primarily due to the differences in the input data sources and merging algorithms, causing problems in applications where joint use of the pentad and monthly datasets is necessary. First, pentad CMAP-merged analyses are created by merging several kinds of individual data sources including gauge-based analyses of pentad precipitation, and estimates inferred from satellite observations. The pentad CMAP dataset is then adjusted by the monthly GPCP-merged analyses so that the adjusted pentad analyses match the monthly GPCP in magnitude while the high-frequency components in the pentad CMAP are retained. The adjusted analyses, called the GPCP-merged analyses of pentad precipitation, are compared to several gauge-based datasets. The results show that the pentad GPCP analyses reproduced spatial distribution patterns of total precipitation and temporal variations of submonthly scales with relatively high quality especially over land. Simple applications of the 23-yr dataset demonstrate that it is useful in monitoring and diagnosing intraseasonal variability. The Pentad GPCP has been accepted by the GPCP as one of its official products and is being updated on a quasi-real-time basis.

Corresponding author address: Dr. Pingping Xie, NOAA/NWS/NCEP Climate Prediction Center, 5200 Auth Rd., #605, Camp Springs, MD 20746. Email: Pingping.Xie@noaa.gov

Abstract

As part of the Global Precipitation Climatology Project (GPCP), analyses of pentad precipitation have been constructed on a 2.5° latitude–longitude grid over the globe for a 23-yr period from 1979 to 2001 by adjusting the pentad Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) against the monthly GPCP-merged analyses. This adjustment is essential because the precipitation magnitude in the pentad CMAP is not consistent with that in the monthly CMAP or monthly GPCP datasets primarily due to the differences in the input data sources and merging algorithms, causing problems in applications where joint use of the pentad and monthly datasets is necessary. First, pentad CMAP-merged analyses are created by merging several kinds of individual data sources including gauge-based analyses of pentad precipitation, and estimates inferred from satellite observations. The pentad CMAP dataset is then adjusted by the monthly GPCP-merged analyses so that the adjusted pentad analyses match the monthly GPCP in magnitude while the high-frequency components in the pentad CMAP are retained. The adjusted analyses, called the GPCP-merged analyses of pentad precipitation, are compared to several gauge-based datasets. The results show that the pentad GPCP analyses reproduced spatial distribution patterns of total precipitation and temporal variations of submonthly scales with relatively high quality especially over land. Simple applications of the 23-yr dataset demonstrate that it is useful in monitoring and diagnosing intraseasonal variability. The Pentad GPCP has been accepted by the GPCP as one of its official products and is being updated on a quasi-real-time basis.

Corresponding author address: Dr. Pingping Xie, NOAA/NWS/NCEP Climate Prediction Center, 5200 Auth Rd., #605, Camp Springs, MD 20746. Email: Pingping.Xie@noaa.gov

1. Introduction

Significant progress has been made in the last two decades in quantitatively documenting global precipitation variations, thanks to the advent and continuous operation of satellite observations with advanced infrared (IR) and microwave (MW) instruments. Various algorithms have been developed and applied to derive precipitation estimates over both land and ocean from these observations. Among many other products, precipitation estimates have been produced on an operational and experimental basis by the IR-based Geostationary Operational Environmental Satellite (GOES) Precipitation Index (GPI; Arkin and Meisner 1987), the Special Sensor Microwave Imager (SSM/I) scattering-based algorithm of Ferraro (1997), the SSM/I emission-based technique of Wilheit et al. (1991), the Microwave Sounding Unit (MSU) emission-based method of Spencer (1993), the outgoing longwave radiation (OLR)-based Precipitation Index (OPI; Xie and Arkin 1998), and the Television Infrared Observational Satellite (TIROS) Operational Vertical Sounder (TOVS) based approach of Susskind et al. (1997). Together with gauge observations and precipitation fields produced by various numerical models, these satellite estimates provide important information about precipitation especially over the global oceanic areas. Several intercomparisons have been conducted among various individual data sources of precipitation, including gauge observations, satellite estimates, and model outputs. The results show that all individual data sources present similar distribution patterns of overall structures of global precipitation including rainbands associated with the ITCZ, the South Pacific convergenze zone (SPCZ), and major convection centers over the Tropics and storm tracks over the extratropics. But differences exist in smaller-scale features and in magnitude. At least three major deficiencies exist in the individual data sources: 1) incomplete global coverage, 2) significant random error, and 3) non-negligible bias (Janowiak 1992; Arkin and Xie 1994, Xie and Arkin 1995; Ebert and Manton 1998; Adler et al. 2001).

Acknowledgment of the limitations inherent in the individual datasets has led to the development of algorithms to merge them so as to take advantage of the strengths of each to produce the best possible analyses of global precipitation. One such algorithm was developed by a group at the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC) by combining gauge observations with estimates derived from IR, OLR, SSM/I, and TOVS (Adler et al. 1993, 1994; Huffman et al. 1995). It has been applied successfully to construct the global monthly precipitation analyses for the Global Precipitation Climatology Project (GPCP) for the period from 1979 to the present (Huffman et al. 1997; Adler et al. 2002, manuscript submitted to J. Hydrometer., hereafter ADL). Another merging algorithm was developed by Xie and Arkin (1996) that takes the gauge observations, satellite estimates derived from IR, OLR, MSU, and SSM/I, and the precipitation distributions from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis as inputs. Using this algorithm, a global monthly precipitation dataset, called the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP; Xie and Arkin 1997a), has been created for the same period as the GPCP product. Both the GPCP and the CMAP monthly precipitation datasets have been applied widely in climate analysis (Curtis and Adler 2000; Trenberth and Caron 2000; Lau and Wu 2001), numerical model verification (Stephenson et al. 1998; Janowiak et al. 1998; Dai et al. 2001), hydrological studies (Trenberth and Guillemott 1998), and other investigations (Yang et al. 1999).

The GPCP and the CMAP analyses described above are constructed for monthly precipitation with spatial resolution of 2.5° latitude–longitude. Many applications, including climate diagnosis of intraseasonal variability, surface water budget analysis, and verification of regional and mesoscale models, however, require a time series of precipitation on finer temporal and/or spatial resolution. To meet this requirement, Huffman et al. (2001) developed a satellite-based technique and applied it successfully to construct analyses of daily precipitation on a 1° latitude–longitude grid over the globe. Called the One-Degree Daily (1DD) technique, it first defines an all-satellite product of daily precipitation by combined use of IR, MW, and TOVS satellite observations. It then adjusts the daily values month by month so that the local monthly accumulation of the 1DD analyses matches the local monthly GPCP analysis value. The 1DD precipitation analyses have been produced for the period from 1997 to the present and have been approved by the GPCP as one of its official products. Rain gauge information is not used in the 1DD analyses primarily because the 24-h period over which rain gauge data are collected varies widely among countries, and the availability of subdaily reports is very limited.

Another approach toward submonthly temporal resolution is that of Xie and Arkin (1997b). Adopting the algorithm used to create the monthly CMAP, analyses of pentad precipitation are defined on a 2.5° latitude–longitude grid over the globe by merging gauge observations, satellite estimates, and, optionally, precipitation fields from the NCEP–NCAR reanalysis. As of April 2002, the pentad CMAP analyses have been constructed for the 23-yr period from 1979 to 2001. The dataset has been used by many scientists as a useful information source for applications associated with weather, climate, and hydrological variations on submonthly scales (Ebisuzaki et al. 1998; Zhou and Lau 1999; Qian and Yang 2000; Roads et al. 2001). The precipitation magnitude in the pentad CMAP, however, is not consistent with that in the monthly CMAP or GPCP datasets, primarily due to the differences in the input data sources and the merging algorithms. This inconsistency may cause problems in applications where joint use of the pentad and monthly datasets is necessary. For example, a thorough verification of a climate model would require datasets of monthly and pentad precipitation with consistent magnitude to examine its ability to represent the variability of different temporal scales (decadal, interannual, and intraseasonal).

The objective of this work is to create analyses of pentad precipitation that are consistent with the monthly GPCP product. The basic notion of the work is to adjust the original pentad CMAP analyses by the monthly GPCP so that the magnitude of the adjusted pentad analyses are close to that of the monthly GPCP while the high-frequency components in the pentad CMAP are retained. This work is done as part of the GPCP and the adjusted pentad analyses, called the pentad GPCP analyses, have been accepted by the project as its official product for pentad precipitation. The pentad CMAP is adjusted to the monthly GPCP instead of the monthly CMAP because the adjusted pentad analyses are going to be used as part of the GPCP product suite that include monthly GPCP analyses of ADL and the GPCP 1DD dataset of Huffman et al. (2001).

Section 2 of this paper gives a brief description of the merging algorithm and input data sources used to define the pentad CMAP dataset; section 3 presents procedures to define the pentad GPCP analyses by adjusting the pentad CMAP; section 4 shows validation results for the pentad GPCP, section 5 illustrates some of its applications in analysis of intraseasonal variations; and a summary is given in section 6.

2. Defining the pentad CMAP by merging individual data sources

The pentad CMAP analyses are created by merging several kinds of individual data sources of precipitation, including gauge observations, satellite estimates, and, optionally, precipitation fields from the NCEP–NCAR reanalysis. The algorithm used to define the pentad analyses is a modification of that for the monthly CMAP (Xie and Arkin 1996, 1997a), while the inputs are the same products as or similar replacements to those used in the monthly CMAP. Two versions of the pentad CMAP datasets are created. In the first version (CMAP/A), all available input datasets are used to define the merged analyses to ensure complete spatial coverage with reasonable quality over the entire globe. In the second version (CMAP/O), only observation-based input datasets are used so that the resulting merged analyses are clear of any influences from numerical models used in creating the reanalysis precipitation fields. In this work, the observation-only version of the pentad CMAP (CMAP/O) is utilized to create the adjusted analyses, in accordance with the GPCP policy of avoiding numerical model influence in its product suite. Brief descriptions of the merging algorithm and the various inputs to define the CMAP/O are given below.

a. Merging algorithm

The algorithm used to define the pentad CMAP/O dataset is basically the same as that used for the monthly CMAP/O (Xie and Arkin 1996, 1997a). The merging of the individual input data sources is conducted in two steps. First, to reduce the random error, the satellite estimates are combined linearly through the maximum likelihood estimation method, in which the linear combination coefficients are inversely proportional to the squares of local random error of the individual data sources. Over the global land areas, the individual random error is defined for each grid and for each pentad by comparing the data sources with the concurrent gauge-based analysis over the surrounding areas. Over global oceanic areas, it is defined by comparison with atoll gauge data (Morrissey et al. 1995) over the Tropics and by subjective assumptions regarding the error structures over the extratropics (see Xie and Arkin 1997a for details).

Since the output of the first step contains a bias that is passed through from the individual input data sources, a second step is included to remove it. For that purpose, the gauge-based analyses are combined with the output of the first step. Over land areas, the gauge data and the output of the first step are blended through the method of Reynolds (1988), in which the first-step output and the gauge data are used to define the relative distribution (or “shape”) and the magnitude of the precipitation fields, respectively. Over the oceans, the bias in the first-step output is removed by comparison with atoll gauge data over the Tropics and by subjective assumptions regarding the bias over the extratropics. In the process of defining the pentad CMAP, the gauge data are used twice, first as “ground truth” to define the random error for each satellite estimates and then as “anchors” to determine the magnitude of the precipitation. By doing this, the algorithm is able to better take advantage of the quantitative accuracy of the gauge observations.

b. Input data sources

In creating the monthly CMAP/O dataset, six kinds of individual data sources are used as inputs to the merging process. These are the gauge data (the gauge-based analyses over land and the atoll gauge observations over ocean) and five sets of satellite estimates of monthly precipitation derived from 1) the IR-based GPI (Arkin and Meisner 1987), 2) the SSM/I scattering-based ALG85 (Ferraro 1997), 3) the SSM/I emission-based algorithm (Wilheit et al. 1991), 4) the MSU-based method (Spencer 1993), and 5) the OLR-based OPI (Xie and Arkin 1998). Among these individual datasets, the gauge data, the satellite estimates of the GPI, ALG85, MSU, and OPI are available for pentad temporal resolution and are utilized as inputs to define the pentad CMAP/O dataset. Pentad precipitation estimates from the SSM/I emission-based algorithm of Wilheit et al. (1991), however, are not available as of April 2002 and a replacement has to be used to fill in the gap. A description of this replacement product is presented later in this section.

The gauge-based analyses of pentad precipitation are constructed by interpolating gauge observations from over 6000 Global Telecommunication System (GTS) stations over the global land areas. First, station observations of pentad precipitation are defined for each GTS station by accumulating daily reports for the corresponding period. Analyses of pentad precipitation are then created on a 2.5° latitude–longitude over global land areas by interpolating the station observations using the algorithm of Shepard (1968).

The atoll gauge rainfall data of Morrissey et al. (1995) are used to define the error structure of the individual input precipitation fields over tropical oceanic areas. The atoll gauge dataset used here consists of station observations of daily precipitation from over 100 gauges located on atolls and small islands without high terrain. These atoll gauges are located mainly in the central and western tropical Pacific Ocean along a northwest to southeast axis extending from 10°N and 140°E to 20°S and 140°W (see Fig. 1 of Morrissey et al. 1995). In this study, pentad precipitation is first defined for each atoll gauge station by accumulating corresponding daily observations. Areal mean precipitation is then calculated for 2.5° latitude–longitude grid boxes with one or more atoll gauges by taking the arithmetic mean of the corresponding stations.

The GPI technique estimates area mean precipitation from fractional coverage of clouds colder than 235 K in IR images using an empirical linear equation. Covering 40°S–40°N over both land and ocean, the GPI estimates are available in pentad and monthly accumulations for a period from 1986 to the present. Here, the pentad GPI estimates are used as input to create the merged analyses.

The SSM/I scattering-based precipitation estimates used here are those produced by the ALG85 algorithm (Ferraro 1997). First, rain rates are calculated from MW-scattering signals of ice particles and large water droplets using an empirical relation derived by comparison with radar observations. An additional rainfall retrieval is then made over oceanic areas using the 19 and 39-GHz components of the liquid water emission technique (Weng and Grody 1994) to pick up rainfall unidentified in the first step. The pentad precipitation estimates of ALG85 are available from 60°S to 60°N over both land and ocean and cover a period from July 1987 to the present with data for December 1987 missing.

Since the SSM/I emission-based precipitation estimates of Wilheit et al. (1991) are not available at pentad resolution, the oceanic components of ALG37 (Ferraro 1997) are used as an alternative product. The ALG37 retrieves oceanic precipitation from brightness temperatures observed from the 19- and 37-GHz channels of the SSM/I using a liquid water emission technique developed by Weng and Grody (1994). A simple adjustment was conducted for the original ALG37 estimates to ensure that the monthly climatology of the ALG37 matches that of the Wilheit et al. (1991). This was done by comparing the original pentad ALG37 and the Wilheit et al. (1991) estimates for a 8-yr period from July 1987 to June 1995. First, monthly fields of ALG37 were computed for the 8-yr period by accumulating the corresponding pentad estimates, and monthly climatologies were calculated for both the ALG37 and the Wilheit et al. (1991) estimates. The adjustment factor was then defined for each grid box and for each calendar month as the ratio between the local mean value of the estimates of Wilheit et al. (1991) to that of the original ALG37 over a 9 × 9 array of grid boxes centered at the target. Finally, this adjustment factor was interpolated back to pentad intervals and used to modify the original ALG37 estimates for the period from July 1987 to the present.

The MSU-based precipitation data used here are defined from daily estimates of Spencer (1993), which cover the global ocean from 60°S to 60°N and extend from January 1979 to May 1994. Following the procedures in producing the monthly CMAP (Xie and Arkin 1997a), the original MSU estimates are adjusted by a “base product” of pentad precipitation defined by merging gauge observations and satellite estimates of GPI, SSM/I scattering, and SSM/I emission (see section 2 of Xie and Arkin 1997a for details). The objective of this adjustment procedure is to reduce the systematic differences in spatial distribution observed between the MSU and the base product (Janowiak et al. 1995).

The OPI technique derives pentad precipitation in three steps. First, the mean annual cycle of pentad precipitation is defined by averaging the pentad base product for an 8-yr period from July 1987 to June 1995. The pentad anomaly of precipitation is then calculated from the pentad OLR anomaly using proportional constants that are linear functions of local pentad climatology. The OPI estimates of total precipitation are finally obtained by adding the anomaly to the pentad climatology (Xie and Arkin 1998). The pentad OPI estimates are available over most of the globe and for a period from January 1979 to the present.

c. Pentad CMAP merged analyses

The CMAP/O analyses of pentad precipitation are constructed on a 2.5° latitude–longitude over the globe for the 23-yr period from 1979 to 2001 by merging the six kinds of individual data sources, whenever available, using the algorithm described in section 2a. Figure 1 shows an example of precipitation distribution for pentad 41 (20–24 July) of 1988 as obtained from the individual inputs and the merged analyses of CMAP/O. In general, all of the satellite estimates present similar large-scale distribution patterns, characterized by rain bands associated with the ITCZ, the SPCZ over the Tropics, and storm tracks extending from the Tropics to the midlatitudes. The GPI and the OPI exhibit broader and smoother distributions of raining areas compared to those in the SSM/I-based products. Over land, the GPI shows overestimates compared to the gauge-based analyses, especially over the extratropics. The merged analyses of CMAP/O present spatial distribution patterns similar to those in the individual satellite estimates, while their magnitude over land is close to that of the gauge-based analyses, indicating that the bias present in the individual satellite estimates has been reduced substantially.

3. Defining the Pentad GPCP by adjusting the pentad CMAP

a. Discrepancies between the pentad and monthly merged analyses

The pentad CMAP/O analyses contain useful information of precipitation variations with submonthly timescales. Their magnitude, however, is inconsistent with that in the merged analyses of monthly precipitation. Figure 2 shows the distribution of annual mean precipitation for a 20-yr period from 1979 to 1998 as defined from the version 2 dataset of the monthly GPCP-merged analyses (Fig. 2, top; ADL 3), the observation-only version of the pentad CMAP (CMAP/O, Fig. 2, middle), and the differences between them (Fig. 2, bottom). The two datasets present very similar spatial distribution patterns in annual mean precipitation, characterized by rain bands associated with the ITCZ and SPCZ in the Tropics and storm tracks in the extratropics. The temporal correlation between the monthly GPCP and the monthly precipitation accumulated from the pentad CMAP/O is very high over most of the globe (figures not shown here), indicating good agreements in temporal variation patterns of monthly and longer timescales between the two datasets.

Systematic differences, however, are observed in the overall magnitude of the precipitation (Fig. 2, bottom). Over the ocean, the pentad CMAP/O is wetter over the Tropics and drier over the mid- and high latitudes compared to the monthly GPCP analyses. In addition, the pentad CMAP/O tends to have smaller amounts of precipitation over some of the global land areas. These differences are caused primarily by the differences in the algorithms used to define the merged analyses and in the input data sources (Gruber et al. 2000). While the GPCP merged analyses of monthly precipitation derive their magnitude from the SSM/I-based precipitation estimates over the oceans, the pentad CMAP, as described in section 2a, determines the oceanic precipitation magnitude by comparison with observations made by gauges located over atolls and small islands over the western Pacific Ocean. Differences over land are caused primarily by discrepancies between gauge observations from monthly reports and those from the accumulation of daily reports. The accumulation of daily reports tends to underestimate the precipitation over some of the global land areas, especially over North and South Americas. In general, the quality of monthly reports is better because higher-quality observations are available at monthly time resolution (“CLIMAT” reports) and because daily errors tend to average out over time. The comparison results between the monthly GPCP and the monthly accumulation of pentad CMAP/O show that while the two sets of data are in good agreement in temporal variation patterns, systematic differences do exist and may cause problems in some applications. While there are uncertainties in the magnitude of the analyses over oceanic areas, the differences over land are caused mostly by some gauge observations with the less desirable quality in the pentad dataset. Adjusting the pentad CMAP/O against the monthly GPCP will not only create a pentad precipitation analysis with a magnitude consistent with that of the monthly GPCP but also will improve the quantitative accuracy of the adjusted product, at least over land.

b. Selection of the adjustment method

To create a pentad precipitation dataset consistent with the monthly-merged analyses, we decided to adjust the original pentad CMAP by the monthly GPCP-merged analyses so that the magnitude of the adjusted analyses is close to that of the monthly GPCP while components of high-frequency variations in the original pentad CMAP are retained. In accordance with the GPCP policy of avoiding numerical model influence in its product suite, the observation-only version of the pentad CMAP (CMAP/O) is used to define the adjusted analyses. The resulting pentad analyses therefore will have missing values over high-latitude oceanic areas. The version 2 dataset of the GPCP-merged analyses of precipitation is used here as the reference to determine the magnitude in the adjustment. As described in detail in ADL, the monthly GPCP version 2 dataset is constructed by combining gauge observations and satellite estimates of GPI, SSM/I, OPI, and TOVS and covers the period from 1979 to the present.

While other methods might have been used, a simple and straightforward approach was adopted here; namely, to adjust the pentad CMAP/O at each grid box and for each pentad by multiplying the original analysis values by an adjustment factor. The adjustment factor is defined as the ratio between the local mean value of the monthly GPCP and that of the pentad CMAP/O averaged over a time–space domain centered at the target grid box and the target pentad. Experiments were conducted to determine the best combination of the time- and space-averaging scales over which the mean values of the monthly GPCP and pentad CMAP/O are used to define the adjustment factor. In the following discussions, time–space averaging means the process used to calculate the local mean values of the monthly GPCP and the monthly accumulations of the pentad CMAP/O. No smoothing is applied to the ratio between the local means.

First, monthly values for the CMAP/O were calculated for a 20-yr period from 1979 to 1998 by accumulating corresponding pentad analyses. Ratios between the mean value of the monthly GPCP and that of the accumulated monthly CMAP/O were then computed for each grid box and for each pentad over 16 combinations of time–space-averaging domains. These domains include time averaging of 1, 3, 5, and 7 months around the target pentad and space averaging of 1, 3, 5, and 7 grid boxes of 2.5° latitude–longitude in both the north–south and east–west directions centered at the target grid box. We denote the averaging scale of 1 monthly–1 grid box as no temporal–spatial averaging, meaning that only the monthly GPCP and monthly accumulation of CMAP/O for the month including the target pentad over the target grid box is included in the calculation. The ratio is limited to a range of 0.2–5.0 to avoid unrealistic adjustments.

As expected, noticeable discontinuities are observed in the ratios over the monthly boundaries when no temporal averaging is applied in calculating the local means for the monthly GPCP and monthly accumulations of pentad CMAP/O (not shown here). Adding temporal averaging, meanwhile, results in smooth variations in the time series. This implies that adjustment based on only the spatial averaging may alias the high-frequency temporal variation components in the pentad precipitation analyses.

The ratios calculated over various time–space-averaging domains were applied to the original CMAP/O, creating 16 sets of adjusted pentad precipitation analyses for the 20-yr period from 1979 to 1998. These adjusted analyses were then compared to the monthly GPCP and the original pentad CMAP/O to examine their performance.

Two sets of comparisons were conducted for the 16 sets of the adjusted pentad analyses to determine the best combination of time–space-averaging domain. In the first set, the monthly accumulations of the various adjusted pentad analyses were compared to the monthly GPCP to examine how well their magnitudes match. Table 1 presents the comparison results over the entire globe from 60°S to 60°N and for the 20-yr period from 1979 to 1998. Overall, all of the 16 sets of the adjusted pentad analyses based on various combinations of the time–space-averaging domains yield good agreement in both the magnitude and variation patterns. The correlation is higher than 0.9 and the bias is only −0.7% to −1.4% for the various adjusted analyses. The best agreement with the monthly GPCP is observed for the adjusted analyses with no temporal and spatial averaging, for which the correlation reaches 0.997, the bias is as low as −0.7%, and the random error is only 7.4% (Table 1). Averaging in both space and time degrades the agreements between the resulting adjusted analyses and the monthly GPCP. The correlation, bias, and random error are 0.908, −1.3%, and 41.0%, respectively, for time–space averaging of 7 months and 7 grid boxes. While good agreement is expected for the adjusted analysis with no temporal and spatial averaging, high correlation for the pentad analyses adjusted with various averaging scales implies a reasonable match in the temporal–spatial patterns in the monthly GPCP and the original pentad CMAP/O over the averaging domains.

Figures 3 and 4 show the spatial distribution of the temporal correlation and bias for eight selected sets of adjusted pentad analyses. Almost perfect agreement is observed between the monthly GPCP and the adjusted analyses based on ratios calculated with no time–space averaging. The correlation is close to 1.0 (Fig. 3), the bias is nearly 0 (Fig. 4), and the random error is very small (not shown) over most of the global areas. Although the agreement becomes worse as the averaging scale increases, the correlation is higher than 0.7 over most of the globe for adjusted analyses based on various combination of averaging scales (Fig. 3). Particularly noticeable is the spatial distribution of the bias for the various adjusted pentad analyses (Fig. 4). Bands of bias with alternating signs are observed around major precipitation systems, indicating that systematic under- and overestimation of precipitation occur in the adjusted analyses if spatial averaging is included in calculating the adjustment factors.

The following two things are clear from the comparison results described above: 1) the adjustment based on less averaging yields better agreement with the monthly GPCP; and 2) spatial averaging is not desirable in calculating the adjustment factor.

The procedures described above examined the best averaging scales to ensure magnitude agreement with the monthly GPCP. However, it is equally important to ensure that the high-frequency variations inherent in the pentad CMAP/O are retained. To this end, 20–100-day bandpass filtering was performed for the time series of the original pentad CMAP/O and for the 16 sets of the adjusted pentad analyses. Comparisons were then conducted between the bandpass-filtered components in the original pentad CMAP/O and those in the 16 sets of the adjusted analyses. Table 2 presents the correlation coefficients between the bandpass-filtered components of the original pentad CMAP/O and those of the various adjusted analyses calculated over the global areas from 60°S to 60°N and for a time period from 1979 to 1998.

In general, the agreement in high-frequency components is very good for all of the 16 adjusted analyses (Table 2). The correlation is the lowest for the analyses adjusted with no time and space averaging. The correlation improves with increasing averaging scale in both time and space and reaches the highest for space averaging of seven grid boxes and time averaging of seven months. Applying averaging in either the space or time direction results in substantial improvements in the correlation compared to that for the adjustment with no averaging at all. The correlation jumps from 0.914 for the nonaveraging option to 0.950/0.953 with a one-step averaging in time–space.

Overall, in considering the agreement in components of high-frequency temporal variations, it is desirable to adjust the original CMAP/O by a ratio between the local mean values of the monthly GPCP and that of the monthly accumulation of the pentad CMAP/O averaged over a larger time–space domain. Especially, temporal averaging is necessary to avoid aliasing of the temporal variability.

The following four things are clear from examination of both the magnitude agreement with the monthly GPCP and the agreement in high-frequency components with the original pentad CMAP/O:

  1. Averaging on a smaller time–space domain results in better magnitude agreement with the monthly GPCP;

  2. Improved agreement in high-frequency components is achieved when the adjustment factor is calculated on a larger/longer averaging domain;

  3. Spatial averaging yields undesirable artificial bias patterns; and

  4. Temporal averaging is necessary to avoid aliasing in high-frequency variability.

These four conclusions point to an option to define the adjustment factor with temporal averaging of some extent but with no space averaging. Since defining the adjustment factor over time-averaging domains of 3 or 7 months would yield only minor differences in the performance (Table 2), we chose the time averaging of 3 months, or ±1 month around the target pentad, as the best averaging domain for calculation of the adjustment factor.

c. Construction of the adjusted pentad analyses

We applied the adjustment factor calculated in this manner to adjust the original pentad CMAP/O for a 23-yr period from 1979 to 2001. We call this adjusted pentad analyses the pentad GPCP merged analyses. An example of the adjusted pentad analyses for pentad 41 (20–24 July) of 1988 is shown in Fig. 1. As expected, the pentad GPCP exhibits very similar spatial distribution patterns with that of the original pentad CMAP/O while differences in magnitude are observed.

4. Validation of the pentad GPCP analysis

The pentad GPCP merged analyses of precipitation were compared to three gauge-based datasets to examine their ability to represent temporal and spatial variations in several regions over the globe. The three gauge-based datasets used here are those of Higgins et al. (2000) over the United States, Shi et al. (2001) over Brazil, and the atoll gauge data of Morrissey et al. (1995) over the central and western Pacific Ocean.

The dataset of Higgins et al. (2000) consists of analyses of daily precipitation on a 0.25° latitude–longitude grid over the continental United States covering a 51-yr period from 1948 to 1998. The analyses are defined by interpolating quality-controlled gauge observations at over 8000 stations collected from multiple sources. Pentad accumulation of precipitation over 2.5° latitude–longitude grid boxes was calculated from the daily analyses and compared to our pentad GPCP-merged analyses. Although gauge observations at many GTS stations used as part of the inputs to the pentad CMAP/O and therefore the pentad GPCP-merged analyses are also included in creating the daily analyses of Higgins et al. (2000), the latter contains at least 10 times more gauges than those in the GTS over the United States. The two datasets therefore can be considered largely independent.

Table 3 (top) presents the comparison results with the gauge-based analysis of Higgins et al. (2000) over the entire continental United States and for the 18-yr period from 1979 to 1996. Overall, the pentad GPCP merged analyses compare very well with the gauge-based analyses of Higgins et al. (2000). The correlation for the total precipitation, pentad anomaly, and intraseasonal components (defined as the 20–100-day bandpass-filtered components) are 0.873, 0.842, and 0.865, respectively, indicating that the pentad GPCP analyses are capable of representing the spatial distribution and temporal variation of total precipitation as well as its components of submonthly timescales with good accuracy. A negative bias of −6.7% is reported over the combined time–space domain, suggesting a slight underestimation of the pentad product compared to Higgins et al. (2000).

Figure 5 shows the spatial distribution of the temporal correlation, bias, and random error between the total precipitation of the pentad GPCP and that of the Higgins et al. (2000) for the 18-yr period from 1979 to 1996. The correlation for the total precipitation (Fig. 5, top) is higher than 0.8 and the random error is less than 60% over most of the United States. The best agreement is observed over the central United States where precipitation with less spatial variation is observed by a relatively dense network of GTS gauges. The worst performance of the pentad GPCP, meanwhile, is seen over the western mountainous areas where the GTS gauge network is sparse and the satellite estimates are less accurate. Negative bias in the pentad GPCP is observed over most of the United States, especially over the western mountainous and the coastal regions. This negative bias is caused primarily by the underestimation of the gauge-based analyses of the Global Precipitation Climatology Centre (GPCC; Schneider 1993) and Xie at al. (1996), which dominate the land portion of the monthly GPCP merged analyses that in turn control the magnitude of precipitation in our pentad GPCP analyses. A preliminary examination showed that the gauge-based analyses of GPCC and Xie at al. (1996), defined by interpolating station observations of total precipitation, may contain bias over areas where systematic differences exist between the total precipitation over the target grid points and that over the reporting gauge stations (Chen et al. 2002). Over the United States, the GTS stations tend to be located over flat areas with less precipitation and the resulting gauge-based analyses therefore may underestimate precipitation. The dataset of Higgins et al. (2000), meanwhile, is created using station observations from many more gauges that are better representative of precipitation distribution over the region.

Similar comparisons of the pentad GPCP analyses were conducted with the gauge-based analyses of Shi et al. (2001) over Brazil for the 18-yr period from 1979 to 1996. Like Higgins et al. (2000), the dataset of Shi et al. (2001) also comprises analyses of daily precipitation created by interpolating gauge observations from up to 1000 stations over the nation. The original daily analyses were created on a 1.0° latitude–longitude grid over the domain and cover a 38-yr period from 1960 to 1997. Pentad accumulation of precipitation was computed on a 2.5° latitude–longitude grid over the domain for the 18-yr period from 1979 to 1996. They were then compared to our pentad GPCP dataset. Since the number of gauges used to define the daily analyses of Shi et al. (2001) is at least an order of magnitude more than that from the GTS for the period of comparison, the two datasets are largely independent.

Table 3 (middle) presents the comparison results between the pentad GPCP analyses and the pentad accumulation of Shi et al. (2001) over the entire region of Brazil and for the entire 18-yr period from 1979 to 1996. Good agreement is observed between the pentad GPCP and the gauge-based dataset of Shi et al. (2001) over the combined space–time domain. The correlation is 0.776, 0.660, and 0.688, respectively, for the total value, anomaly, and intraseasonal components of the pentad precipitation. The bias is only 0.7% and the random error is 70.7% relative to the gauge-based analyses over the combined space–time domain. The spatial distribution of the comparison statistics for total precipitation (Fig. 6), however, exhibits regional differences in the performance of the pentad GPCP analyses. While excellent agreement is observed over the eastern half of the Brazil where reasonable GTS coverage is available to define the pentad GPCP, the agreement is degraded over the central portion of the Amazon basin where heavy rainfall is observed by relatively sparsely distributed GTS networks. The correlation is over 0.8 over the eastern portion, while it is around 0.6–0.7 over the central Amazon basin. Particularly interesting is the spatial distribution patterns of the bias in the pentad GPCP merged analyses. While no significant bias exists over the combined space–time domain, over-and underestimation of precipitation were reported over the northern coastal regions and the central Amazon basin, respectively. As discussed for the comparison over the United States, most of this bias is likely attributed to the systematic differences in the gauge-based analyses of monthly precipitation over the region.

Since no independent gauge observations of pentad precipitation are available over an extended area over the global oceanic areas and for an extended time period, here we tried to use the atoll rain gauge observations to examine the performance of the pentad GPCP in representing oceanic precipitation. Although, as described in section 2, the atoll gauge data are used to determine the error structure of the individual input data sources in constructing the pentad CMAP and therefore the comparison of the pentad GPCP with the atoll data is not truly independent, we hope that this comparison is still be able to provide us with some information about the performance of the pentad GPCP merged analyses.

The correlation is 0.667, 0.637, and 0.661 for the total value, anomaly, and intraseasonal components of the pentad precipitation, respectively, over the combined space–time domain compared to the atoll gauge data (Table 3c), indicating that the pentad GPCP analyses represent precipitation variations reasonably well over the oceanic areas examined here. Part of the degradation of the agreement is due to the limited number of atoll gauges available to define the grid box mean of precipitation. Previous work by Xie and Arkin (1995) revealed that the correlation between the satellite estimates and atoll gauge data improves for grid boxes with more atoll gauges. It is not surprising that the pentad GPCP merged analyses exhibit a negative bias of 21.6% compared to the atoll gauge observations. As described in section 3, the magnitude of the pentad GPCP analyses over the global ocean is adjusted against the monthly GPCP whose magnitude is dominated by the satellite estimates of Wilheit et al. (1991). Negative bias of the Wilheit et al. (1991) against the atoll gauge data has been reported by several intercomparison projects (e.g., Adler et al. 2001). The real magnitude accuracy of Wilheit et al. (1991), however, is unknown due to the lack of reliable in situ observations over the open ocean. Although only gauge observations over stations located over atolls and small islands are included in the comparison, they may not be representative of precipitation over surrounding open oceans due to local circulations induced by the topography.

It is clear from the comparisons with the three gauge-based datasets that the pentad GPCP analyses are able to represent precipitation variations of intraseasonal and longer timescales with good accuracy over both land and oceanic areas examined here. Biases, however, exist in the magnitude of precipitation over some of the land areas and the quantitative accuracy of the pentad GPCP is uncertain over oceanic areas.

5. Applications of the Pentad GPCP analyses

The annual, interannual, and intraseasonal variability of global precipitation, as observed in the Pentad GPCP dataset, is examined for the 22-yr period from 1979 to 2000 and compared with that in the NOAA pentad OLR dataset (Gruber and Krueger 1984).

First, components associated with the mean annual cycle, interannual and intraseasonal variability are defined for the precipitation and OLR, respectively. For this purpose, the mean values of precipitation and OLR are calculated for each of the 73 pentads and for each 2.5° latitude–longitude grid box over the globe from the 22-yr pentad datasets of precipitation and OLR. Harmonic analysis is then applied to the time series of 73 mean values and the accumulation of the first 6 harmonics is used to approximate the annual cycle of the precipitation and OLR. The components associated with interannual and intraseasonal variability, meanwhile, are defined by applying bandpass filtering to the time series of pentad anomaly defined by subtracting the annual cycle from the original total precipitation and OLR. The bandwidth used to extract the interannual and intraseasonal components is 73–365 pentads (1–5 yr) and 4–20 pentads (20–100 days), respectively.

Shown in Fig. 7 are latitudinal profiles of the annual cycle of mean precipitation averaged over ocean (top), land (middle), and the entire globe (bottom). The evolution of mean precipitation is dominated by the migration of rainbands associated with the ITCZ, SPCZ, convergence zones over South America and Africa, and storm tracks over the extratropics. The center of heavy rain is located south of the equator during boreal winter, moves northward during spring and reaches ∼10°N in boreal summer. Over ocean, the magnitude of precipitation is the maximum during boreal summer when an intensified ITCZ extends across the Pacific basin. The precipitation over land, however, is stronger from December to May when enhanced convection is present over tropical Africa, the Maritime Continent, and the Amazon basin. A band of precipitation is observed over the midlatitude on both hemispheres throughout the year. Over land, this midlatitude rainband reaches maximum intensity in summer, while over ocean, the maximum appears in winter when the storm tracks are strong.

One of the major components of the global climate, the intraseasonal variability has long been examined using pentad averages of the OLR observed by the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites (e.g., Weickmann et al. 1985; Lau and Chan 1986; Waliser et al. 1999). While the OLR data are thought to be a good index of tropical convection and a reasonable proxy for latent heat release over the Tropics, a dataset of pentad precipitation is preferable for its direct and quantitative relation to latent heating.

Fig. 8 shows the time–longitude sections of the 20–100-day bandpass-filtered precipitation (mm day−1, left) and OLR (W m−2, middle) averaged over 10°S–10°N for a period from October 1996 to May 1997. Also plotted in Fig. 8 (right) is the time series of an index for tropical intraseasonal oscillation (TISO) defined as the 20–100-day bandpass-filtered velocity potential in the NCEP–NCAR reanalysis averaged over a domain of 10°S–10°N, 100°–140°E (Sperber et al. 1997). Eastward propagation of the anomaly fields is apparent in both the precipitation and OLR with a period of about 40 days. The correspondence between anomaly in precipitation and that in OLR (convection) is generally very good and both of them are in phase with changes in the TISO index. In general, a positive TISO index (convergence at 200 mb) is accompanied by depressed convection (enhanced OLR) and weakened precipitation over the Maritime Continent area. Compared to that of the OLR, the bandpass-filtered anomaly of precipitation contains many small-scale features and some breaks over the Maritime Continent area.

To further understand the behavior of the precipitation and OLR in response to TISO, precipitation and OLR composites were assembled for different phases of the TISO evolution during the December–January–February (DJF) seasons. The entire life cycle of a TISO is divided into four phases based on the bandpass-filtered TISO index. Phases 1 and 3 are assigned to the pentad periods when the TISO index reaches maximum and minimum, respectively, while phases 2 and 4 are labeled to the periods in between. To make the resulting composites typical of the TISO evolution, only cases with maximum/minimum index values greater/less than 0.75/−0.75 standard deviation were included in defining the composites.

As shown in Fig. 9, during phase I, a weak positive precipitation anomaly appears over the central and western Indian Ocean, while suppressed precipitation is observed over the Maritime Continent area and its vicinity. As it propagates eastward, the positive precipitation anomaly intensifies and its extent widens (phase 2). It then reaches its maximum in phase 3 when the enhanced precipitation is over the Maritime Continent. Upon passing the landmass, the anomaly weakens as it moves toward the southeast. In general, the OLR (right panels) shows similar evolution processes to those for the precipitation, with negative–positive anomaly in OLR corresponding generally to enhanced–depressed precipitation. Mismatches between the precipitation and OLR anomalies, however, exist especially over some of the land areas. The negative OLR anomalies over tropical Africa in phase 1, over tropical Africa and south Australia in phase 2, and over the Sahara Desert in phase 4 are not accompanied by enhanced precipitation. A brief examination of both the precipitation and the OLR fields showed that there is no precipitation observed in the DJF seasons over these land areas. The negative OLR anomalies are therefore most likely attributed to changes in surface features and clouds not associated with precipitation.

With the 22-yr dataset of pentad precipitation, it becomes possible to compare the relative magnitude of variability with different timescales. To do this, the standard deviation is calculated for the time series of components associated with annual, interannual, and intraseasonal variability for precipitation and OLR, respectively. Presented in Figs. 10 and 11 are spatial distributions of the standard deviation of the mean annual cycle (top right), interannual (bottom left), and intraseasonal components (bottom right) for the DJF season for the 22-yr period from 1979 to 2000 for precipitation and OLR, respectively. The distribution of annual mean precipitation and OLR (top left) is also plotted for comparison purpose.

The magnitude of the annual cycle of precipitation (Fig. 10, top left) is large over the major rainbands associated with the ITCZ, SPCZ, and storm tracks, both over land and over ocean. Especially the annual cycle is stronger over the eastern Pacific compared to the western Pacific, although the opposite is observed in the seasonal mean precipitation. The interannual variability in precipitation (Fig. 10, bottom left) exhibits a large magnitude over the entire tropical Pacific basin with its maxima centered at the central Pacific near the date line. Strong interannual variability in precipitation is also observed over Brazil, the tropical Atlantic, and the regions over and south of the storm tracks over the Northern Hemisphere.

Overall, the magnitude of the standard deviation in the pentad precipitation anomaly components associated with the intraseasonal variability (Fig. 10, bottom right) is much larger than that of the interannual variability over most of the globe. During the DJF season, a large magnitude is observed over the eastern Indian and western Pacific Oceans where the maximum values are over 6 mm day−1, almost double of that for the interannual variability over the central Pacific. This intraseasonable variability is noticeable over South America where the standard deviation of the bandpass-filtered pentad precipitation anomaly is over 2 mm day−1 over most of the region.

The spatial distribution pattern of annual, interannual, and intraseasonal variations as observed in the NOAA pentad OLR data (Fig. 11) is similar to those in the precipitation over most of the Tropics as described above. Significant differences, however, exist over the extratropics and over some of the tropical areas. As shown in Figs. 10 and 11, most of the extratropical features in the annual, interannual, and intraseasonal variability of precipitation are missing in those of the OLR, attributed mostly to the fact that the OLR is determined by the surface temperatures and clouds that may or may not be precipitating. In particular, differences in the magnitude in precipitation and OLR are noted over tropical Africa and northern Australia. While our Pentad GPCP precipitation dataset shows little variance in precipitation over these regions (Fig. 10), the OLR data present relatively large variance over the same region. Comparisons with the distribution of the annual mean precipitation and OLR show that mean precipitation is very small and the OLR values are high, indicating that the variability in the OLR is caused mostly to changes in surface temperature and nonprecipitating clouds.

6. Summary

Analyses of pentad precipitation have been constructed on a 2.5° latitude–longitude grid over the globe for the 23-yr period from 1979 to 2001 by adjusting the observation-only version of the pentad CMAP (CMAP/O) against the monthly GPCP-merged analyses. First, pentad CMAP/O-merged analyses are created by merging several kinds of individual data sources using the same algorithm as that for the monthly CMAP (Xie and Arkin 1997a). The individual data sources used as inputs to the merging process include the gridded fields of pentad precipitation derived by interpolating GTS gauge observations, and estimates inferred from satellite observations of GPI, SSM/I, MSU, and OPI. The pentad CMAP/O dataset is then adjusted by the monthly GPCP-merged analyses (ADL) so that the adjusted pentad analyses match the magnitude of the monthly GPCP while their high-frequency components are the same as those in the original pentad CMAP/O. The adjustment is done for each grid box and for each pentad time step by first calculating the ratio between the temporal mean of the monthly GPCP and that of the pentad CMAP/O over the target grid box for a 3-month period centered at the target pentad and then multiplying the ratio by the original pentad CMAP/O.

Called the GPCP-merged analyses of pentad precipitation, the adjusted analyses are compared to several gauge-based datasets of precipitation. The results showed that the Pentad GPCP analyses are capable of reproducing spatial distribution patterns of total precipitation and temporal variations of submonthly scales with relatively high quality. Bias, however, exist in the analyses over some of the land areas. In addition, the quantitative accuracy is uncertain over oceanic areas due to lack of appropriate independent observations of precipitation. Preliminary analysis of the 23-yr dataset demonstrated its potential applications in monitoring and diagnosing intraseasonal variability.

Accepted by the GPCP as one of its official products, the Pentad GPCP dataset is being updated on a quasi-real-time basis. The current version of the Pentad GPCP merged analyses described in this paper is experimental in nature. Modifications and improvements are planned for the pentad dataset once more information about its strengths and shortcomings are gathered from scientists in various fields. In particular, since gauge observations of daily and pentad precipitation from many more stations recently became available, improvements of the pentad merged analyses are expected by inclusion of those additional station data.

Acknowledgments

The authors would like to express their thanks to J. E. Schemm, W. Shi, W.-Q. Wang, Y. Xue, S. Yang, E. Yarosh, and J.-Y. Zhou for their invaluable discussions and comments on the work. They are also indebted to Y. Yarosh for her excellent work in timely updating the pentad GPCP analyses, and to G. Fullwood and S. C. Handel for providing the GTS daily reports used in this study. Comments made by two anonymous reviewers greatly improved the quality of this paper. The pentad GPCP dataset is available through anonymous ftp from the Climate Prediction Center (CPC) online at ftp.ncep.noaa.gov/pub/precip/GPCP_PEN and from the NOAA National Climatic Data Center (NCDC) of at ftp.ncdc.noaa.gov/pub/data/gpcp.

REFERENCES

  • Adler, R. F., A. J. Negri, P. R. Keehn, and I. M. Hakkarinen, 1993: Estimation of monthly rainfall over Japan and surrounding waters from a combination of low orbit microwave and geosynchronous IR data. J. Appl. Meteor., 32 , 335356.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., G. J. Huffman, and P. R. Keehn, 1994: Global rain estimates from microwave adjusted geosynchronous IR data. Remote Sens. Rev., 11 , 125152.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., C. Kidd, G. Petty, M. Morrissey, and H. M. Goodman, 2001: Intercomparison of global precipitation products: The third Precipitation Intercomparison Project (PIP-3). Bull. Amer. Meteor. Soc., 82 , 13771396.

    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and B. N. Meisner, 1987: The relationship between large-scale convective rainfall and cold cloud over the Western Hemisphere during 1982–1984. Mon. Wea. Rev., 115 , 5174.

    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and P. Xie, 1994: The Global Precipitation Climatology Project: First Algorithm Intercomparison Project. Bull. Amer. Meteor. Soc., 75 , 401419.

    • Search Google Scholar
    • Export Citation
  • Chen, M., P. Xie, J. E. Janowiak, and P. A. Arkin, 2002: Global land precipitation: A 50-year monthly analysis based on gauge observations. J. Hydrometeor., 3 , 249266.

    • Search Google Scholar
    • Export Citation
  • Curtis, S., and R. F. Adler, 2000: ENSO indices based on patterns of satellite derived precipitation. J. Climate, 13 , 27862793.

  • Dai, A., T. M. L. Wigley, B. A. Boville, J. T. Kiehl, and L. E. Buja, 2001: Climates of the twentieth and twenty-first centuries simulated by the NCAR climate system model. J. Climate, 14 , 485519.

    • Search Google Scholar
    • Export Citation
  • Ebert, E. E., and M. J. Manton, 1998: Performance of satellite rainfall estimation algorithms during TOGA COARE. J. Atmos. Sci., 55 , 15371557.

    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., M. Kanamitsu, J. Potter, and M. Fiorino, 1998: An overview of Reanalysis-2. Preprints, 23d Annual Climate Diagnostics Workshop, Miami, FL, Climate Prediction Center, 119–120.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., 1997: Special Sensor Microwave Imager derived global rainfall estimates for climatological applications. J. Geophys. Res., 102 , 1671516735.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., and A. F. Krueger, 1984: The status of the NOAA outgoing longwave radiation data set. Bull. Amer. Meteor. Soc., 65 , 958962.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., X. Su, M. Kanamitsu, and J. Schemm, 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., W. Shi, E. Yarosh, and R. Joyce, 2000: Improved United States precipitation quality control system and analysis. Climate Prediction Center Atlas, No. 7, National Oceanic and Atmospheric Administration, National Weather Service. [Available from NOAA/NWS/NCEP Climate Prediction Center, Camp Springs, MD 20746.].

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., R. F. Adler, B. R. Rudolf, U. Schneider, and P. R. Keehn, 1995: Global precipitation estimates based on a technique for combining satellite-based estimates, rain gauge analysis, and NWP model precipitation information. J. Climate, 8 , 12841295.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and Coauthors. 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc., 78 , 520.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observations. J. Hydrometeor., 2 , 3650.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., 1992: Tropical rainfall: A comparison of satellite derived rainfall estimates with model precipitation forecasts, climatologies, and observations. Mon. Wea. Rev., 120 , 448462.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., P. A. Arkin, P. Xie, M. L. Morrissey, and D. R. Legates, 1995: An examination of the east Pacific ITCZ rainfall distribution. J. Climate, 8 , 28102823.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., A. Gruber, C. R. Kondragunta, R. E. Livezey, and G. J. Huffman, 1998: A comparison of the NCEP–NCAR reanalysis precipitation and the GPCP rain gauge–satellite combined dataset with observational error considerations. J. Climate, 11 , 29602979.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and P. H. Chan, 1986: Aspects of the 40–50-day oscillation during the northern summer as inferred from outgoing longwave radiation. Mon. Wea. Rev., 114 , 13541367.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., and H. T. Wu, 2001: Principal modes of rainfall–SST variability of the Asian summer monsoon: A reassessment of the monsoon–ENSO relationship. J. Climate, 14 , 28802895.

    • Search Google Scholar
    • Export Citation
  • Morrissey, M. L., M. A. Shafer, S. E. Postawko, and B. Gibson, 1995: Pacific raingauge data. Water Resour. Res., 31 , 21112113.

  • Qian, W., and S. Yang, 2000: Onset of the regional monsoon over Southeast Asia. Meteor. Atmos. Phys., 75 , 2938.

  • Reynolds, R. W., 1988: A real-time global sea surface temperature analysis. J. Climate, 1 , 7586.

  • Roads, J. O., S. C. Chen, and F. Fujioka, 2001: ECPC's weekly to seasonal global forecasts. Bull. Amer. Meteor. Soc., 82 , 639658.

  • Schneider, U., 1993: The GPCC quality-control system for gauge-measured precipitation data. GEWEX Workshop on Analysis Methods of Precipitation on Global Scale, Rep. WCRP-81, WMP/TD-588, Koblenz, Germany, WMO, A5–A9.

    • Search Google Scholar
    • Export Citation
  • Shepard, D., 1968: A two dimensional interpolation function for regularly spaced data. Proc. 23d National Conf. of American Computing Machinery, Princeton, NJ, Association for Computing Machinery, 517–524.

    • Search Google Scholar
    • Export Citation
  • Shi, W., R. W. Higgins, E. Yarosh, and V. E. Kousky, cited. 2001: The annual cycle and variability of precipitation in Brazil. NCEP/Climate Prediction Center Atlas, No. 9, National Oceanic and Atmospheric Administration. National Weather Service. [Available online at http://www.cpc.noaa.gov/research_papers/ncep_cpc_atlas/9/index.html.].

    • Search Google Scholar
    • Export Citation
  • Spencer, R. W., 1993: Global oceanic precipitation from MSU during 1979–91 and comparisons to other climatologies. J. Climate, 6 , 13011326.

    • Search Google Scholar
    • Export Citation
  • Sperber, K. R., J. M. Slingo, P. M. Inness, and W. K-M. Lau, 1997: On the maintenance and initiation of the intraseasonal oscillation in the NCEP/NCAR reanalysis and the GLA and UKMO AMIP simulations. Climate Dyn., 13 , 769795.

    • Search Google Scholar
    • Export Citation
  • Stephenson, D. B., F. Chauvin, and J-F. Royer, 1998: Simulation of the Asian summer monsoon and its dependence on model horizontal resolution. J. Meteor. Soc. Japan, 76 , 237265.

    • Search Google Scholar
    • Export Citation
  • Susskind, J., P. Piraino, L. Rokkle, L. Iredell, and A. Mehta, 1997: Characteristics of the TOVS Pathfinder Path A dataset. Bull. Amer. Meteor. Soc., 78 , 14491472.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and C. J. Guillemott, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP reanalysis. Climate Dyn., 14 , 213231.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and J. M. Caron, 2000: The Southern Oscillation revisited: Sea level pressures, surface temperatures, and precipitation. J. Climate, 13 , 43584365.

    • Search Google Scholar
    • Export Citation
  • Waliser, D. E., C. Jones, J. K. Schemm, and N. E. Graham, 1999: A statistical extended-range tropical forecast model based on the slow evolution of the Madden–Julian Oscillation. J. Climate, 12 , 19181939.

    • Search Google Scholar
    • Export Citation
  • Weickmann, K. M., G. R. Lussky, and J. E. Kutzbach, 1985: Intraseasonal (30–60 day) fluctuations of outgoing longwave radiation and 250-mb stream function during northern winter. Mon. Wea. Rev., 113 , 941961.

    • Search Google Scholar
    • Export Citation
  • Weng, F-Z., and N. C. Grody, 1994: Retrieval of cloud liquid water using the special sensor microwave imager (SSM/I). J. Geophys. Res., 99 , 2553525551.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. J., A. T. C. Chang, and L. S. Chiu, 1991: Retrieval of the monthly rainfall indices from microwave radiometric measurements using probability distribution functions. J. Atmos. Oceanic Technol., 8 , 118136.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1995: An intercomparison of gauge observations and satellite estimates of monthly precipitation. J. Appl. Meteor., 34 , 11431160.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840858.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1997a: Global Precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1997b: Global pentad precipitation analysis based on gauge observations, satellite estimates and model outputs. Extended Abstracts, Amer. Geophys. Union 1997 Fall Meeting, San Francisco, CA, Amer. Geophys. Union.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. A. Arkin, 1998: Global monthly precipitation estimates from satellite-observed outgoing longwave radiation. J. Climate, 11 , 137164.

    • Search Google Scholar
    • Export Citation
  • Xie, P., B. Rudolf, U. Schneider, and P. A. Arkin, 1996: Gauge-based monthly analysis of global land precipitation from 1971–1994. J. Geophys. Res., 101 (D14) 1902319034.

    • Search Google Scholar
    • Export Citation
  • Yang, S., K-M. Lau, and P. S. Schopf, 1999: Sensitivity of the tropical Pacific Ocean to precipitation-induced freshwater flux. Climate Dyn., 15 , 737750.

    • Search Google Scholar
    • Export Citation
  • Zhou, J., and W. K-M. Lau, 1999: Summertime intraseasonal variability over South America. Preprints, 24th Annual Climate Diagnostics Workshop, Tucson, AZ, Climate Prediction Center, 299–302.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Precipitation (mm day−1) for pentad 41 (20–24 Jul) of 1988 as observed in the satellite estimates of GPI, SSM/I scattering (SCT), SSM/I emission (EMS), OPI, and MSU; the gauge-based analyses; the merged analyses of pentad CMAP (observation-only version); and the pentad GPCP

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 2.
Fig. 2.

Distribution of mean precipitation (mm day−1) for a 20-yr period from 1979 to 1998 as defined from (top) the monthly GPCP analyses version 2 dataset, (middle) that from the observation-only version of pentad CMAP, and (bottom) the difference between the two

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 3.
Fig. 3.

Correlation between the monthly GPCP analyses and the monthly accumulation of pentad analyses defined by adjusting the original pentad CMAP/O by a ratio calculated over various time–space-averaging domains for a 20-yr period from 1979 to 1998.

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 4.
Fig. 4.

Same as in Fig. 3, except for bias (%) relative to the mean value of the monthly GPCP

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 5.
Fig. 5.

(top) Correlation, (middle) relative bias (%), and (bottom) relative rms error (%) between the total precipitation in the pentad GPCP and that in the guage-based analyses of Higgins et al. (2000) over the United States for an 18-yr period from 1979 to 1996

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 6.
Fig. 6.

Same as in Fig. 5, except for comparison with Shi et al. (2001) over Brazil

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 7.
Fig. 7.

Longitudinal profiles of mean annual cycle of precipitation (mm day−1) averaged over (top) the ocean, (middle) land, and (bottom) the entire globe

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 8.
Fig. 8.

Time–longitude sections of (left) 20–100-day bandpass-filtered precipitation (mm day−1) and (middle) OLR (W m−2) averaged over 10°S–10°N for the period from Oct 1996 to May 1997. (right) The time series of an index associated with the TISO defined as bandpass-filtered velocity potential averaged over 10°S–10°N, 100°–140°E

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 9.
Fig. 9.

Composites of (left) 20–100-day bandpass-filtered precipitation (mm day−1) and (right) OLR (W m−2) defined by dividing a cycle of intraseasonal oscillation into four phases based on the TISO index. Phases 1 and 3 denote time when the TISO index reaches max and min, respectively, while phases 2 and 4 are periods in between. Only cases with max/min index values larger/smaller than 0.75/−0.75 std dev are included in defining the composites

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 10.
Fig. 10.

GPCP pentad analysis global distribution of (top left) seasonal mean precipitation (mm day−1) and std dev of precipitation (mm day−1) for (top right) the mean annual cycle, and components associated with (bottom left) interannual and (bottom right) intraseasonal components for the DJF season

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Fig. 11.
Fig. 11.

Same as in Fig. 10, except for the OLR (W m−2) observed by the NOAA satellites

Citation: Journal of Climate 16, 13; 10.1175/2769.1

Table 1.

Comparison between the monthly GPCP analyses and monthly accumulations of original pentad CMAP and those adjusted with factors calculated over various spatial and temporal averaging scales

Table 1.
Table 2.

Correlation between the bandpassed components of the original pentad CMAP and those adjusted by the monthly GPCP with factors calculated over various spatial and temporal averaging scales

Table 2.
Table 3.

Comparison of the pentad GPCP analyses with gauge observations

Table 3.
Save