• Adler, R. F., and Coauthors, 2003: The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–present). J. Hydrometeor., 4 , 11471167.

    • Search Google Scholar
    • Export Citation
  • Alishouse, J. C., , J. B. Snider, , E. R. Westwater, , C. T. Swift, , C. S. Ruf, , S. A. Snyder, , J. Vongsathorn, , and R. R. Ferraro, 1990: Determination of cloud liquid water content using the SSM/I. IEEE Trans. Geosci. Remote Sens., 28 , 817822.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , and P. Schluessel, 1993: Rainfall, total water, ice water, and water vapor over sea from polarized microwave simulations and special sensor microwave/imager data. J. Geophys. Res., 98 , D11. 2073720778.

    • Search Google Scholar
    • Export Citation
  • Bennartz, R., 2007: Global assessment of marine boundary layer cloud droplet number concentration from satellite. J. Geophys. Res., 112 .D02201, doi:10.1029/2006JD007547.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , C. Kummerow, , and C. A. Morales, 2002: Differences between East and West Pacific rainfall systems. J. Climate, 15 , 36593672.

    • Search Google Scholar
    • Export Citation
  • Bony, S., , and J-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models. Geophys. Res. Lett., 32 .L20806, doi:10.1029/2005GL023851.

    • Search Google Scholar
    • Export Citation
  • Borg, L. A., , and R. Bennartz, 2007: Vertical structure of stratiform marine boundary layer clouds and its impact on cloud albedo. Geophys. Res. Lett., 34 .L05807, doi:10.1029/2006GL028713.

    • Search Google Scholar
    • Export Citation
  • Bremen, L. V., , E. Ruprecht, , and A. Macke, 2002: Errors in liquid water path retrieval arising from cloud inhomogeneities: The beam-filling effect. Meteor. Z., 11 , 1319.

    • Search Google Scholar
    • Export Citation
  • Chang, A. T. C., , L. S. Chiu, , and G. Yang, 1995: Diurnal cycle of oceanic precipitation from SSM/I data. Mon. Wea. Rev., 123 , 33713380.

    • Search Google Scholar
    • Export Citation
  • Chevallier, F., , and P. Bauer, 2003: Model rain and clouds over oceans: Comparison with SSM/I observations. Mon. Wea. Rev., 131 , 12401255.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , C. D. Ardeel, , and L. Tian, 1990: Liquid water content and precipitation characteristics of stratiform clouds as inferred from satellite microwave measurements. J. Geophys. Res., 95 , 1665916671.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , T. R. Karl, , B. Sun, , and K. E. Trenberth, 2006: Recent trends in cloudiness over the United States: A tale of monitoring inadequacies. Bull. Amer. Meteor. Soc., 87 , 597606.

    • Search Google Scholar
    • Export Citation
  • Garreaud, R. D., , and R. Muñoz, 2004: The diurnal cycle in circulation and cloudiness over the subtropical southeast Pacific: A modeling study. J. Climate, 17 , 16991710.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , G. L. Stephens, , T. H. Vonder Haar, , and D. L. Jackson, 1993: A physical retrieval of cloud liquid water over the global oceans using special sensor microwave/imager (SSM/I) observations. J. Geophys. Res., 98 , 1847118488.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , G. L. Stephens, , S. A. Christopher, , and T. H. Vonder Haar, 1995: Observations of the global characteristics and regional radiative effects of marine cloud liquid water. J. Climate, 8 , 29282946.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , S. A. Christopher, , and J. Chou, 1997: Cloud liquid water path comparisons from passive microwave and solar reflectance satellite measurements: Assessment of subfield-of-view cloud effects in microwave retrievals. J. Geophys. Res., 102 , 1958519596.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , T. S. L’Ecuyer, , and S. A. Christopher, 2007: Evaluating specific error characteristics of microwave-derived cloud liquid water products. Geophys. Res. Lett., 34 .L22807, doi:10.1029/2007GL031180.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., , and K. Woodberry, 1993: The diurnal cycle of tropical convection. J. Geophys. Res., 98 , 1662316638.

  • Hilburn, K. A., , and F. J. Wentz, 2008: Intercalibrated passive microwave rain products from the Unified Microwave Ocean Retrieval Algorithm (UMORA). J. Appl. Meteor. Climatol., 47 , 778794.

    • Search Google Scholar
    • Export Citation
  • Horváth, A., , and R. Davies, 2007: Comparison of microwave and optical cloud water path estimates from TMI, MODIS, and MISR. J. Geophys. Res., 112 .D01202, doi:10.1029/2006JD007101.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., , P. A. Arkin, , and M. Morrissey, 1994: An examination of the diurnal cycle in oceanic tropical rainfall using satellite and in situ data. Mon. Wea. Rev., 122 , 22962311.

    • Search Google Scholar
    • Export Citation
  • Jung, T., , E. Ruprecht, , and F. Wagner, 1998: Determination of cloud liquid water path over the oceans from Special Sensor Microwave/Imager (SSM/I) data using neural networks. J. Appl. Meteor., 37 , 832844.

    • Search Google Scholar
    • Export Citation
  • Lietzke, C. E., , C. Deser, , and T. H. Vonder Haar, 2001: Evolutionary structure of the eastern Pacific double ITCZ based on satellite moisture profile retrievals. J. Climate, 14 , 743751.

    • Search Google Scholar
    • Export Citation
  • Lin, B., , and W. B. Rossow, 1994: Observations of cloud liquid water path over oceans: Optical and microwave remote sensing methods. J. Geophys. Res., 99 , D10. 2090720928.

    • Search Google Scholar
    • Export Citation
  • Liu, G., , and J. A. Curry, 1993: Determination of characteristic features of cloud liquid water from satellite microwave measurements. J. Geophys. Res., 98 , D3. 50695092.

    • Search Google Scholar
    • Export Citation
  • Mears, C. A., , D. K. Smith, , and F. J. Wentz, 2001: Comparison of Special Sensor Microwave Imager and buoy-measured wind speeds from 1987 to 1997. J. Geophys. Res., 106 , 1171911730.

    • Search Google Scholar
    • Export Citation
  • Mears, C. A., , M. C. Schabel, , and F. J. Wentz, 2003: A reanalysis of the MSU channel 2 tropospheric temperature record. J. Climate, 16 , 36503664.

    • Search Google Scholar
    • Export Citation
  • Meissner, T., , D. Smith, , and F. Wentz, 2001: A 10-year intercomparison between collocated Special Sensor Microwave Imager oceanic surface wind speed retrievals and global analyses. J. Geophys. Res., 106 , 1173111742.

    • Search Google Scholar
    • Export Citation
  • Norris, J. R., 1999: On trends and possible artifacts in global ocean cloud cover between 1952 and 1995. J. Climate, 12 , 18641870.

  • Petty, G. W., 1990: On the response of the special sensor microwave/imager to the marine environment—Implications for atmospheric parameter retrievals. Ph.D. thesis, University of Washington, 118 pp.

  • Petty, G. W., 1994: Physical retrievals of over-ocean rain rate from multichannel microwave imagery. Part II: Algorithm implementation. Meteor. Atmos. Phys., 54 , 101121.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., , N. A. Rayner, , T. M. Smith, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80 , 22612287.

  • Rozendaal, M. A., , C. B. Leovy, , and S. A. Klein, 1995: An observational study of diurnal variations of marine stratiform cloud. J. Climate, 8 , 17951809.

    • Search Google Scholar
    • Export Citation
  • Sohn, B-J., , and E. A. Smith, 2003: Explaining sources of discrepancy in SSM/I water vapor algorithms. J. Climate, 16 , 32293255.

  • Staelin, D. H., , K. F. Kunzi, , R. L. Pettyjohn, , R. K. L. Poon, , R. W. Wilcox, , and J. W. Waters, 1976: Remote sensing of atmospheric water vapor and liquid water with the Nimbus 5 microwave spectrometer. J. Appl. Meteor., 15 , 12041214.

    • Search Google Scholar
    • Export Citation
  • Thomas, S. M., , A. K. Heidinger, , and M. J. Pavolonis, 2004: Comparison of NOAA’s operational AVHRR-derived cloud amount to other satellite-derived cloud climatologies. J. Climate, 17 , 48054822.

    • Search Google Scholar
    • Export Citation
  • Tiedke, M., 1993: Representation of clouds in large-scale models. Mon. Wea. Rev., 121 , 30403061.

  • Trenberth, K. E., , J. Fasullo, , and L. Smith, 2005: Trends and variability in column-integrated atmospheric water vapor. Climate Dyn., 24 , 741758.

    • Search Google Scholar
    • Export Citation
  • Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

  • Vinnikov, K. Y., , and N. C. Grody, 2003: Global warming trend of mean tropospheric temperature observed by satellites. Science, 302 , 269272.

    • Search Google Scholar
    • Export Citation
  • Weng, F., , and N. C. Grody, 1994: Retrieval of cloud liquid water using the special sensor microwave imager (SSM/I). J. Geophys. Res., 99 , 2553525552.

    • Search Google Scholar
    • Export Citation
  • Weng, F., , N. C. Grody, , R. Ferraro, , A. Basist, , and D. Forsyth, 1997: Cloud liquid water climatology from the Special Sensor Microwave/Imager. J. Climate, 10 , 10861098.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 1997: A well-calibrated ocean algorithm for special sensor microwave imager. J. Geophys. Res., 102 , 87038718.

  • Wentz, F. J., , and R. W. Spencer, 1998: SSM/I rain retrievals within a unified all-weather ocean algorithm. J. Atmos. Sci., 55 , 16131627.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , and T. F. Meissner, 2000: Algorithm Theoretical Basis Document (ATBD) version 2, AMSR ocean algorithm. RSS Tech. Proposal 121599A-1, 59 pp.

  • Wentz, F. J., , and M. Schabel, 2000: Precise climate monitoring using complementary satellite data sets. Nature, 403 , 414416.

  • Wentz, F. J., , L. Ricciardulli, , K. A. Hilburn, , and C. Mears, 2007: How much more rain will global warming bring? Science, 317 , 233235.

    • Search Google Scholar
    • Export Citation
  • Wood, R., , C. S. Bretherton, , and D. L. Hartmann, 2002: Diurnal cycle of liquid water path over the subtropical and tropical oceans. Geophys. Res. Lett., 29 .2092, doi:10.1029/2002GL015371.

    • Search Google Scholar
    • Export Citation
  • Wylie, D. P., , and H. M. Woolf, 2002: The diurnal cycle of upper-tropospheric clouds measured by GOES-VAS and the ISCCP. Mon. Wea. Rev., 130 , 171179.

    • Search Google Scholar
    • Export Citation
  • Yang, G-Y., , and J. Slingo, 2001: The diurnal cycle in the tropics. Mon. Wea. Rev., 129 , 784801.

  • View in gallery

    Monthly mean LWP for the six SSM/Is, AMSR-E, and TMI, averaged over (a) the Niño-3.4 region (10°S–10°N, 170°–120°W) and (b) the tropics (30°S–30°N). Note the different vertical scales on the y axes. Gaps in the data (especially for the F08 SSM/I) occurred when the data were too sparse within a particular month to form a representative average.

  • View in gallery

    Estimated intramonthly LWP standard deviation for January. These data were used as input to the algorithm fitting for the mean diurnal cycle and mean yearly LWP; its value only affects the estimated fitting errors. Black pixels denote land while gray pixels denote missing data, either from the presence of sea ice or the close proximity of land.

  • View in gallery

    Example of the diurnal cycle fitting results for a 1° grid box near the northern coast of the Indonesian island of Borneo (3.5°S, 111.5°E) for the month of January, using different combinations of sensors. LWP observations taken at similar local times from the same sensor have been averaged for visibility, with corresponding one-sigma error bars shown. “Amplitude” represents 50% of the diurnal range of LWP, while “Phase” represents the local time of maximum LWP; each has been displayed along with the corresponding one-sigma error. Diurnal cycle results using (a) only SSM/I data, (b) both SSM/I and AMSR-E data, (c) only TMI data, and (d) all available data, (e) same fit as (d) but, for display purposes, the data points have been corrected for the difference in the mean LWP for the year in which they were taken and (f) mean LWP for each year when there was data along with the one-sigma error bars, corresponding to fit from (e).

  • View in gallery

    (a) Relationship between the fraction of LWP contained in clouds ( fL) and the TWP in the ECMWF forecast model vs the RSS retrieval assumption (black line), as well as two alternate assumptions: ALT (gray line) and PETTY (black dot–dashed line); see text for details. (b) Zonal mean LWP from standard RSS algorithm as well as the two alternative assumptions; (c) as in (b) but for the zonal mean rain rate.

  • View in gallery

    (top) Estimate of the systematic errors due to the cloud–rain separation assumption, and (bottom) average monthly statistical errors on LWP from UWisc climatology.

  • View in gallery

    Mean values of LWP averaged over the years 1988–2005 for (a) January, (b) April, (c) July, and (d) October. Black pixels denote land, while gray pixels denote missing data, either from the presence of sea ice or the close proximity of land.

  • View in gallery

    Mean zonal cloud liquid water path over ocean for (a) DJF and (b) JJA from three different observational climatologies and one meteorological reanalysis. The G95 climatology contains data from F08 from July 1987 through December 1991, while both the UWisc climatology and the W97 dataset range from 1988 to 2005. The displayed ERA-40 data range from 1988 to 2002. The gray shading denotes the estimated range of uncertainty in the UWisc climatology due to the cloud–rain partitioning assumption discussed in section 4e.

  • View in gallery

    Zonally averaged correlation in the (a) mean monthly and (b) deseasonalized mean monthly LWP time series between various climatologies. ERA refers to the ERA-40 and UWisc to the present work. All comparisons were done at 1° resolution. Note that the G95 climatology only includes data through 1991.

  • View in gallery

    (a) Correlation coefficient in the time series of mean monthly LWP between the ERA-40 reanalysis and the UWisc climatology, from January 1988 through August 2002; (b) as in (a) but the time series have been deseasonalized before the calculation of the correlation coefficients. Note the different scales between (a) and (b).

  • View in gallery

    Relative amplitude and phase of the diurnal cycle in the months of January and July. The diurnal cycle phase is the local time of maximum LWP. Black pixels denote land, while gray pixels denote missing data, from either the presence of sea ice or the close proximity of land; (b), (d) gray pixels also indicate locations without a well-defined diurnal maximum in LWP.

  • View in gallery

    Map of the amplitude of the second harmonic of the diurnal cycle for a region off the coast of South America for the months of April and October. The plotted region extends from the equator to 40°S, 70°–110°W.

  • View in gallery

    Diurnal cycle fits for the two small regions off the west coast of South America, depicted by filled black circles in Fig. 11 for the months of (top) April and (bottom) October. The more northern location has a single-peaked diurnal cycle in April but a strong double-peaked cycle in October; the more southern location has a double peak in both months, and in fact throughout the year (not shown).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 356 356 56
PDF Downloads 240 240 43

Cloud Liquid Water Path from Satellite-Based Passive Microwave Observations: A New Climatology over the Global Oceans

View More View Less
  • 1 Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado
  • | 2 Remote Sensing Systems, Santa Rosa, California
  • | 3 Department of Atmospheric and Oceanic Sciences, University of Wisconsin—Madison, Madison, Wisconsin
© Get Permissions
Full access

Abstract

This work describes a new climatology of cloud liquid water path (LWP), termed the University of Wisconsin (UWisc) climatology, derived from 18 yr of satellite-based passive microwave observations over the global oceans. The climatology is based on a modern retrieval methodology applied consistently to the Special Sensor Microwave Imager (SSM/I), the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and the Advanced Microwave Scanning Radiometer (AMSR) for Earth Observing System (EOS) (AMSR-E) microwave sensors on eight different satellite platforms, beginning in 1988 and continuing through 2005. It goes beyond previously published climatologies by explicitly solving for the diurnal cycle of cloud liquid water by providing statistical error estimates, and includes a detailed discussion of possible systematic errors.

A novel methodology for constructing the climatology is used in which a mean monthly diurnal cycle as well as monthly means of the liquid water path are derived simultaneously from the data on a 1° grid; the methodology also produces statistical errors for these quantities, which decrease toward the end of the time record as the number of observations increases. The derived diurnal cycles are consistent with previous findings in the tropics, but are also derived for higher latitudes and contain more information than in previous studies. The new climatology exhibits differences with previous observationally based climatologies and is found to be more consistent with the 40-yr ECMWF Re-Analysis (ERA-40) than are the previous climatologies.

Potential systematic errors of the order of 15%–30% or higher exist in the LWP climatology. A previously unexplored source of systematic error is caused by the assumption that all microwave-based retrievals of LWP must make regarding the partitioning of cloud water and rainwater, which cannot be determined using microwave observations alone. The potentially large systematic errors that result may hamper the usefulness of microwave-based climatologies of both cloud liquid water and especially rain rate, particularly in certain regions of the tropics and midlatitudes where the separation of rain from liquid cloud water is most critical.

Corresponding author address: Christopher W. O’Dell, Department of Atmospheric Science, Colorado State University, 200 West Lake St., Fort Collins, CO 80523. Email: odell@atmos.colostate.edu

Abstract

This work describes a new climatology of cloud liquid water path (LWP), termed the University of Wisconsin (UWisc) climatology, derived from 18 yr of satellite-based passive microwave observations over the global oceans. The climatology is based on a modern retrieval methodology applied consistently to the Special Sensor Microwave Imager (SSM/I), the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and the Advanced Microwave Scanning Radiometer (AMSR) for Earth Observing System (EOS) (AMSR-E) microwave sensors on eight different satellite platforms, beginning in 1988 and continuing through 2005. It goes beyond previously published climatologies by explicitly solving for the diurnal cycle of cloud liquid water by providing statistical error estimates, and includes a detailed discussion of possible systematic errors.

A novel methodology for constructing the climatology is used in which a mean monthly diurnal cycle as well as monthly means of the liquid water path are derived simultaneously from the data on a 1° grid; the methodology also produces statistical errors for these quantities, which decrease toward the end of the time record as the number of observations increases. The derived diurnal cycles are consistent with previous findings in the tropics, but are also derived for higher latitudes and contain more information than in previous studies. The new climatology exhibits differences with previous observationally based climatologies and is found to be more consistent with the 40-yr ECMWF Re-Analysis (ERA-40) than are the previous climatologies.

Potential systematic errors of the order of 15%–30% or higher exist in the LWP climatology. A previously unexplored source of systematic error is caused by the assumption that all microwave-based retrievals of LWP must make regarding the partitioning of cloud water and rainwater, which cannot be determined using microwave observations alone. The potentially large systematic errors that result may hamper the usefulness of microwave-based climatologies of both cloud liquid water and especially rain rate, particularly in certain regions of the tropics and midlatitudes where the separation of rain from liquid cloud water is most critical.

Corresponding author address: Christopher W. O’Dell, Department of Atmospheric Science, Colorado State University, 200 West Lake St., Fort Collins, CO 80523. Email: odell@atmos.colostate.edu

1. Introduction

Relatively few long-term studies are available that monitor global changes in cloud properties. Some long-term surface observations exist, but they are widely believed to be fraught with errors and biases (e.g., Norris 1999; Dai et al. 2006). Although satellite observations of cloud properties have existed since the 1980s, constructing an accurate long-term satellite cloud record is hampered due to slow drifts in satellite parameters, such as the onboard calibration, and the necessity of merging data from multiple satellites (as each satellite lasts only a few years). Observations in the visible and thermal infrared employed by the International Satellite Cloud Climatology Project (ISCCP) seem to indicate long-term variability in the global cloudiness (Rossow and Schiffer 1999), while observations of clouds from the Advanced Very High Resolution Radiometer indicate little long-term change in global cloudiness (Thomas et al. 2004). It is not yet known if these climatologies or others have the fidelity to accurately characterize variability in cloud properties over decadal time scales.

Another potential avenue for studying long-term cloud variability lies in the satellite-based passive microwave record, which has been successfully applied to other climate studies. For instance, since 1979 the Microwave Sounding Unit (MSU) aboard multiple polar-orbiting satellites has detected a warming of the lower troposphere (Mears et al. 2003; Vinnikov and Grody 2003). Observations from the Special Sensor Microwave Imager (SSM/I) sensor aboard multiple polar-orbiting platforms indicate increases in global water vapor from the record’s inception in 1987 (Wentz and Schabel 2000; Trenberth et al. 2005). Both sets of observations are consistent with predictions of anthropogenic climate change. It appears that issues associated with calibration, orbital drift, and other effects have been addressed to the degree necessary to reveal long-term global trends. New results based on SSM/I also suggest a modest decrease in global surface winds (Wentz et al. 2007). In addition to water vapor and surface wind speed, the SSM/I is also sensitive to column-integrated cloud liquid water path (LWP) and precipitation. Because the low-frequency microwave channels fully penetrate virtually all clouds, they can provide a direct measurement of the total liquid (but not solid) cloud condensate amount, though there is some dependency on the cloud temperature (Petty 1990; Liu and Curry 1993; Lin and Rossow 1994; Greenwald et al. 1995, hereafter G95). In contrast, near-infrared sensors can also retrieve LWP, but they provide a relatively indirect measurement as the near-infrared radiation primarily samples the cloud top, and the associated retrievals depend on how the cloud drop size distribution is vertically stratified within the cloud (Bennartz 2007; Borg and Bennartz 2007).

A great variety of LWP retrieval algorithms have been introduced since the first SSM/I was launched (e.g., Petty 1990; Alishouse et al. 1990; Liu and Curry 1993; Bauer and Schluessel 1993; Greenwald et al. 1993; Lin and Rossow 1994; Weng and Grody 1994; G95; Jung et al. 1998). Two of these algorithms were applied to multiyear, global-scale SSM/I data to generate over-ocean LWP climatologies (G95; Weng et al. 1997, hereafter W97). However, these climatologies are not generally applicable in the presence of rainfall and are based on relatively simple retrieval methodologies. More significantly, they arrive at fairly different results, for instance, in terms of the zonal distribution of LWP, and this disagreement gives climate modelers a large degree of latitude when testing their simulated cloud products for consistency with observations. There are now nearly two decades of SSM/I data, and retrieval techniques have become somewhat more sophisticated since the previous studies.

In this work, we describe the construction of a long-term climatology of LWP retrieved from the SSM/I, the Advanced Microwave Scanning Radiometer for Earth Observing System (EOS) (AMSR-E) aboard Aqua, and the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) instruments for all observations through the year 2005. The retrieval algorithm is that of Remote Sensing Systems (RSS), to be discussed in more detail below. The LWP climatology has estimates of both statistical errors as well as systematic errors for some sources. In constructing a climatology of LWP, it is important to take into account the diurnal cycle since drifts in local satellite overpass time can lead to artifacts in the climatology such as spurious trends. Thus, we also fit for the diurnal cycle and remove it using an accurate and self-consistent methodology.

The rest of this paper is organized as follows: Section 2 gives an overview of the LWP data and retrieval algorithm used as the basis for the climatology. Section 3 describes the methodology used to simultaneously determine both the diurnal cycle and monthly mean of LWP, while section 4 addresses potential systematic errors in the LWP retrieval that may affect the climatology, with a particular focus on the separation of cloud water from precipitation. In section 5, the features of the mean LWP climatology are explored, and the differences between this climatology as compared to both previous observational climatologies as well as the 40-yr ECMWF Re-Analysis (ERA-40) are discussed. Section 6 discusses some of the features of the resulting climatology of the diurnal cycle of LWP, and finally section 7 offers a general discussion and summary of the main results of this work.

2. Data

The data used in this study are the over-ocean microwave retrieval products calculated at Remote Sensing Systems, specifically for LWP and gridded daily data for both descending and ascending overpasses at 0.25°. A total of 56.6 satellite-years of data are used, from January 1988 through December 2005. The characteristics of the instruments utilized are summarized in Table 1. The satellite crossing time at the equator corresponding to the ascending overpass is referred to as equator-crossing time (ECT). For most of the satellite platforms employed, the ECT drifted in time throughout their lifespan. For the F10, F11, and F13 spacecraft, the ECT drifted toward later times, while for F14 and F15 it initially drifted toward later times but has since drifted more significantly toward earlier times. F08 had and AMSR-E has a very stable ECT. The displayed footprints represent the 3-dB size of the 37-GHz footprint, which corresponds to the resolution of the LWP product.

The LWP retrieval algorithm used in this study is part of a unified, over-ocean retrieval algorithm from RSS. The latest version of the RSS algorithm, called the Unified Microwave Ocean Retrieval Algorithm, has been described at length in a series of publications and on the RSS Web site (http://www.remss.com); we give only a brief overview of the algorithm here. The original algorithm for nonraining pixels was introduced in Wentz (1997) for SSM/I; the retrieval was extended to precipitating regions in Wentz and Spencer (1998), and further refinements to the algorithm are discussed in Wentz and Meissner (2000) and Hilburn and Wentz (2008). The retrieval products used in this study are the latest available for each microwave sensor, specifically the version 6 (V6) SSM/I algorithm, the version 4 TMI algorithm, and the version 5 AMSR-E algorithm. The retrieval algorithm is essentially identical for each sensor, only varying to take into account instrumental differences such as resolution, calibration coefficients, and viewing geometry. In addition to changes in ECT, all satellite platforms occasionally underwent orbital maneuvers that caused small changes in altitude. This has the effect of changing the viewing incidence angle but, as viewing angle is an input parameter to the RSS algorithm (unlike some algorithms), the retrieval results should be insensitive to the altitude changes.

Earlier versions of the RSS retrievals were sufficiently accurate for monitoring water vapor, which requires calibration accuracies of 0.2 K. The V6 SSM/I were better calibrated for climate-quality retrievals for more sensitive parameters, such as wind. The V6 calibration method is similar to the MSU target-factor approach (Mears et al. 2003). The V6-intercalibrated brightness temperatures are in agreement to 0.2 K or better (Hilburn and Wentz 2008). While the SSM/I version 5 calibration also used the target-factor method, there were problems owing to the thermal environment of F14 and attitude anomalies for the F10 spacecraft. These issues have been resolved in V6 and are further discussed in Wentz et al. (2007).

Only brightness temperatures passing a set of quality control measures are used in the retrieval; quality controls include tests for land, sea ice, sun glint, calibration problems, and other issues that could lead to poor over ocean retrievals. For valid pixels, the retrieval uses five channels at frequencies of 19, 22, and 37 GHz to initially retrieve three basic parameters simultaneously over ocean: surface wind speed, column-integrated water vapor path (WVP), and column-integrated total liquid water path (TWP), which includes both falling liquid precipitation as well as suspended cloud water droplets. Note that the algorithm specifically does not use the 85-GHz channel since the retrieval model is emission based and there is significant scattering at this frequency that cannot be neglected. For this reason also, the retrieval is mostly insensitive to cloud ice, though larger frozen hydrometeors may impact the 37-GHz radiances in some instances (Curry et al. 1990). Required auxiliary information consists of the sea surface temperature (SST) and salinity. Salinity is based on a standard climatology. For AMSR-E and TMI, the SST is retrieved directly using lower-frequency channels, while for the SSM/I, which lacks these low frequencies, the SST is drawn from the Reynolds optimally interpolated product (Reynolds et al. 2002). The wind speeds produced by this algorithm have been extensively validated (Mears et al. 2001; Meissner et al. 2001), and the retrieved water vapor paths are also generally consistent with observations (Sohn and Smith 2003; Trenberth et al. 2005).

In nonprecipitating cases, which account for roughly 90% of all footprints, the cloud liquid water path is equal to the TWP. However, the presence of rainfall presents a serious complication as cloud and rainwater are very similar in terms of their radiative effects at these microwave frequencies. Therefore, Wentz and Spencer (1998) adopted a heuristic approach to disentangle the two quantities. In the presence of rain, the cloud liquid water path is assumed to be related to the average rain rate as
i1520-0442-21-8-1721-e1
where R is the average columnar rain rate (mm h−1) and H is the height (km) of the raining column, which is parameterized as a simple function of SST; C is the threshold value of LWP below which rain is assumed not to occur and is assigned a value of 0.18 kg m−2. The first parameterization used for H led to significant errors in the derived climatology of rain rate. A new parameterization based on the freezing level from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis is used in the version 6 SSM/I data and is simply linear in SST, varying roughly from 0.5 to 5 km between 0° and 30°C; this parameterization yields a more accurate climatology of rainfall (Hilburn and Wentz 2008). Potential systematic errors related to the cloud/rain partitioning are explored in section 4.

The LWP retrievals at the footprint sizes shown in Table 1 are then binned into 0.25° × 0.25° grids for each satellite and for both ascending and descending overpasses, forming the basic level-3 product. At higher latitudes, consecutive orbits for the polar-orbiting satellites overlap and, hence, more than one ascending or descending observation may occur for a given location each day; in such instances, only the most recent observation was retained in the daily gridded data. For TMI in its near-equatorial orbit, this effect is strongest from 30° to 40° in absolute latitude, and at these locations all data were averaged together in the level-3 algorithm. Because all diurnal information is lost in the averaging process, no TMI data were used for absolute latitudes higher than 30°.

From the gridded 0.25° daily data, there are a total of roughly 13.2 billion valid LWP observations of ocean grid boxes from 1988 to 2005. In this study, these data were further averaged to 1° resolution to reduce the overall data volume while still maintaining a relatively high resolution. As an example of the fidelity of the satellite intercalibration, Fig. 1 shows the monthly mean LWP from each satellite sensor averaged over two different regions: (i) the Niño-3.4 region (10°S–10°N, 170°–120°W) and (ii) the tropics (30°S–30°N). These averages were constructed directly from the 1° data without any correction for the diurnal cycle although, as both ascending and descending overpasses have been used, most diurnal effects should cancel out. For each region the intercalibration differences appear quite small: the rms differences in the overlap regions are less than 1.5 g m−2 and 95% of the overlap differences are less than 3.3 and 2.1 g m−2 for Figs. 1a,b, respectively. In Fig. 1a, the long-term variability is seen to be dominated by large positive departures in the LWP for the years around 1992 and 1998, likely an effect of El Niño. The long-term variability in Fig. 1b is more complicated, though a seasonal cycle is clearly present and there is no obvious long-term trend.

3. Methodology

The LWP climatology to be constructed is the monthly mean LWP and associated statistical, or random, errors at 1° resolution for the years 1988–2005, corrected for the diurnal cycle of LWP. However, in constructing this climatology, the diurnal cycle of LWP is itself derived to an unprecedented accuracy. The technique is to fit data simultaneously for both the monthly mean and the mean monthly diurnal cycle of LWP (i.e., averaged over all years). This approach thus produces monthly means that are automatically corrected for the average diurnal cycle of LWP. Other models of the diurnal cycle were also tried, including scaling the diurnal cycle amplitude with the mean LWP, or allowing the diurnal cycle to vary freely from year to year. The former approach did not produce similarly good fits as did the adopted method, while the latter was found to be too noisy to be feasible. The diurnal cycle may well exhibit interannual variability in certain locations, but accurately characterizing these variations likely requires a dataset with better temporal resolution.

Previous work by Wood et al. (2002, hereafter W02) determined the diurnal cycle of LWP in the tropics using TMI data from 1999 to 2000. W02 used a simple sinusoid with two parameters representing amplitude and phase to characterize the diurnal cycle in 2.5° grid boxes on seasonal time scales. With the much larger dataset consisting of all years of SSM/I, AMSR-E, and TMI data, this analysis can be significantly extended, both to high latitudes and to higher spatial and temporal resolution (monthly instead of seasonal).

The basic fitting procedure is as follows: For each 1° grid box and month, the measured LWP is modeled according to
i1520-0442-21-8-1721-e2
where Yi is the year corresponding to the ith measurement, ti is the local time (in hours) of this measurement (h), and ω is the radial frequency that corresponds to a 24-h period; (Y) is the mean LWP in year Y, and A1 (T1) and A2 (T2) represent the amplitudes (phases) of the first and second harmonics of the diurnal cycle, respectively. The term n(t) in Eq. (2) represents all sources of variability not captured by the other terms and may, in general, be quite large. Its inclusion in the formalism will enable the calculation of statistical errors upon the derived LWP climatological products. Alhough it could vary with month and grid box, n(t) was assumed to be independent of year, local time, and sensor. Because of this, the magnitude of n(t) has no effect on the values of fitted parameters; it only affects the estimated errors of those parameters. Because this variability does not include the effects of interannual or diurnal variability, we denote it as “intramonthly variability,” which tacitly includes radiometer noise as well as sources of natural LWP variability such as synoptic disturbances. The standard deviation of the intramonthly variability was estimated directly from the data for each month and grid box, smoothed with a 3° full-width-at-half-maximum Gaussian kernel in order to eliminate some of the spatial noise, which is mainly due to the fact that the amount of data for any 1° × 1° grid box is too limited. Figure 2 shows the estimated standard deviation of n(t) for the month of January. The regions of highest mean LWP generally also have the highest variability in LWP.

To reduce the amount of data, a temporal binning was performed before the fitting procedure in which all data from a given satellite within the same 1° grid box, from the same month and year, and taken at the same local time of day were averaged. For all polar-orbiting sensors, this yields at most two values, one from the ascending overpass and one from the descending overpass, since satellite drift over a single month was considered negligible. For TMI, the data were averaged into eight equally spaced local time bins of 3 h each. For a given month and grid box, this yielded, in general, N monthly average LWP measurements, denoted by y1, y2, . . . yN. Here N ranged from a value of 2 early in the record, when only a single polar-orbiting satellite was available, to as high as 16 later in the record (representing measurements from F13, F14, F15, AMSR-E, and TMI). Each of these monthly averaged measurements was assigned an uncertainty equal to σ2/ni, where σ2 is the previously determined variance of the intramonthly variability for this grid box and month, for instance, as shown in Fig. 2, and ni was the number of observations contributing to yi. The total number of measurements for a certain month and grid box is therefore given by ΣNi=1ni, and was generally between 50 and 150.

The procedure used to determine the best-fit values of the parameters (Y = 1988, · · · , 2005) and {A1, T1, A2, T2} was to minimize the following quantity:
i1520-0442-21-8-1721-e3
where yi represents the observed LWP and LWP(Yi, ti) is the modeled LWP from Eq. (2).

The minimization was conducted after transforming Eq. (2) to one that is linear in its parameters and assuming the noise n(t) to be Gaussian; the solution is then explicit and may be obtained by standard matrix techniques. The fitting process also yielded Gaussian error bars of the fit parameters. The astute reader will have noticed that there is a slight “chicken and egg” problem here, since the standard deviation of the intramonthly variability given by σ2 can only be calculated by subtracting the diurnal and interannual variability from the observed LWP, but these terms are only known once the fit is calculated. This problem was solved by first assuming a fixed value for σ, calculating the fit parameters, and then using the results to calculate the true σ2. The fit was then repeated with the new value of σ2; as stated previously, this did not change the values of the fit parameters, but only their estimated statistical uncertainties.

Nearly all grid boxes and months were successfully fit to Eq. (2), but some were excluded. First, fits were restricted to grid boxes and months with at least 1000 total satellite overpasses over the 18 available years of observations. Next, for the few ocean grid boxes at high latitudes that experience no sunlight for an entire month, the diurnal cycle was simply set to zero. This action is a drawback of using the level-3 gridded data, which only keeps the latest overpass for a given daily ascending or descending dataset; the resulting time gaps are generally too large to enable a good fit on the diurnal cycle. Some grid boxes, particularly on certain coastal regions, still had poor coverage in the 0–24-h range and were also excluded from the climatology.

An example of the fitting procedure is shown in Fig. 3, in the month of January, for a grid box off the north coast of Borneo in Indonesia. Figure 3a shows the results of the fit if only SSM/I data are used. Each point represents all observations from the same satellite and local overpass time. The error bar for a point representing n observations (each from a single satellite overpass) is determined from σ2/n, as described previously, and is inversely proportional to the weight given to that point in the fitting procedure. Because the local overpass times of the various SSM/I instruments are not that different, there are large gaps in the data for certain times of day; as a result, a simple sinusoid is fit to the diurnal cycle of the data instead of the full four-parameter diurnal cycle from Eq. (2). In Fig. 3b, AMSR-E data are used in addition to the SSM/I data. Now the full four-parameter diurnal cycle from Eq. (2) has been employed. This leads to a slight narrowing of the peak and widening of the trough, and causes a slight increase in the error bars on the amplitude and phase owing to the extra free parameters. Figure 3c shows the fit results when only TMI data are used; the higher error bars are caused by the smaller amount of data being used. Fits from TMI were generally consistent with those using SSM/I and AMSR-E only, within the specified error ranges; this was nearly always the case in the tropics and is evidence both that the method works and that the error estimates are roughly accurate. Figure 3d shows the result when all SSM/I, TMI, and AMSR-E data are used in the fit. Recall, however, that the assumption of Eq. (2) is that each year has a different mean LWP. Figure 3e shows the same fit as for Fig. 3d but plots the data points such that differences among the mean yearly LWP values have been removed. This visibly reduces the scatter and illustrates the utility of fitting to the form given by Eq. (2).

Finally, Fig. 3f shows the mean LWP for each year. Note that the error bars are significantly smaller toward the end of the data record; this is a direct consequence of the increased amount of data available in later years. The seemingly large interannual variability is real, and is not particular to this chosen location; several regions exhibited the behavior (mainly in the tropics and some coastal areas), though many as well had a small variation in mean LWP, such as the stratocumulus regions and extratropics (not shown). The fitting algorithm was nonetheless found to give robust estimates of all fitted parameters, even in instances where a few years contained large outliers of mean LWP.

4. Retrieval errors

Before describing the resulting LWP climatology, it is worth discussing possible sources of error that could affect the climatology. Assuming that the methodology described in section 3 is accurate, the primary sources of error can be expected to arise from systematic errors in the retrieval algorithm itself. In this section, we give a brief overview of some commonly discussed error sources and go into more detail for a source of error, not commonly discussed in the literature, related to the separation of rainwater from cloud water. Rigorous quantification of all possible error sources is an extensive task and is beyond the scope of the present work. For a more general discussion of LWP retrieval errors and the difficulty of LWP validation, see, for instance, Lin and Rossow (1994).

a. Cross-talk errors

Because the LWP retrieval depends in part on the retrievals of WVP and surface wind speed, as well as surface temperature (either retrieved or specified from a climatology), errors in these parameters can cause errors in the retrieved LWP. These effects were studied in some detail by Lin and Rossow (1994). They found that large (5 K) errors in SST introduce small errors in LWP (less than 10 g m−2). Water vapor and wind speed errors can both lead to errors in LWP, which they showed to be on the order of 30 g m−2 per kg m−2 of water vapor error, and 10 g m−2 per m s−1 of wind speed error. The wind speed errors were found to cause worse relative LWP errors at wind speeds less than 10 m s−1. If these sensitivities are representative for the RSS retrieval, we can combine them with the expected rms errors in these quantities to predict the resulting rms errors in LWP. Using the rms errors given by Wentz and Meissner (2000) of 0.6 kg m−2 for water vapor and 0.9 m s−1 for wind speed, the resulting rms errors in LWP are 18 and 9 g m−2, respectively.

b. Cloud temperature and height errors

The microwave absorption of cloud liquid water has a significant dependence on temperature, with colder liquid clouds absorbing more strongly than warmer clouds. Errors in the assumed cloud temperature TL will therefore lead to errors in LWP (Staelin et al. 1976; Petty 1990; Liu and Curry 1993; Lin and Rossow 1994). Staelin et al. (1976) found this error to be approximately 3% per kelvin of TL error for nonraining scenes. For the RSS retrieval we find the error is 2.7% K−1 for nonraining scenes, and is approximately constant at 3–5 g m−2 K−1 of error for raining scenes. However, what kind of errors in TL can we expect? In the RSS scheme, TL is parameterized as a function of SST and WVP (Hilburn and Wentz 2008). We have estimated the errors in TL by using approximately five million profiles from the widely used ECMWF global forecast model; these profiles were drawn from all times of year and all locations to be globally and annually representative. Clouds are generated in the model using the parameterization of Tiedke (1993). The true TL for each profile is the cloud temperature necessary to retrieve the correct LWP, and can be roughly calculated according to
i1520-0442-21-8-1721-e4
where T(z) is the temperature profile and A(z) is the profile of absorption at 37 GHz from Hilburn and Wentz (2008); because the temperature dependence of liquid water absorption is nearly the same at 19 and 37 GHz, it matters little which frequency is used. When compared with the RSS parameterization, there is a rms error of about 5 K with a small negative bias around −1 K; 95% of the differences were less than 10 K. If these profiles are indicative of nature, we can then expect a corresponding rms error in LWP of about 13% for nonraining scenes and between 5% and 10% for raining scenes (with an error that decreases with increasing rain). Note that retrieved LWP is also dependent on the assumed rain-column height, as given in Eq. (1), for footprints found to contain rain (defined in the RSS retrieval as LWP > 180 g m−2). However, this dependence is quite weak and yields maximum errors of 3–5 g m−2.

c. Clear-sky bias and beam-filling effect

Ideally, the RSS model would retrieve on average zero LWP in clear-sky scenes, but in reality this is not the case. Using concurrent visible and microwave observations, Bennartz (2007) found a positive bias of roughly 20 g m−2 in AMSR-E retrievals under clear-sky conditions in marine stratocumulus regions. Using a similar methodology applied to a larger, more global dataset, Greenwald et al. (2007, hereafter G07) found a slightly lower clear-sky bias for AMSR-E LWP of roughly 7 g m−2 with 70% of the errors occurring between −11 and +25 g m−2. This bias was found to be independent of SST, but depends positively on water vapor, with clear-sky biases as high as 60 g m−2 for WVP values above 60 kg m−2. G07 also find a troublingly strong negative dependence on wind speed, with clear-sky biases of +14 g m−2 near zero wind speed, decreasing linearly to −16 g m−2 for wind speeds above 15 m s−1. A similar dependence on wind speed was found to occur for overcast warm clouds as well. The cause of this bias is currently unclear, as are the resulting regional and global biases, but the preliminary work of G07 found mean global biases of approximately 20% for nonprecipitating clouds.

A related error source is the so-called beam-filling effect (e.g., Greenwald et al. 1997; Bremen et al. 2002) in which clouds only partially field the sensor field of view (FOV). This bias is a function of both cloud fraction and the FOV-averaged LWP. For retrieval algorithms that ignore this effect, the error generally increases with decreasing cloud fraction and increasing LWP. However, a correction is made in the RSS algorithm to account for the beam-filling effect, but only in raining pixels (Wentz and Spencer 1998). Recent refinements to the beam-filling correction were made that account for the differing beam sizes of the various microwave instruments, as well as saturation effects in cases of heavy rain (Hilburn and Wentz 2008); these refinements apply to the products used in this work.

It is not yet clear how well the beam-filling correction works in practice. Horváth and Davies (2007) compared microwave-retrieved LWP from TMI using the RSS algorithm with those from the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on Terra and found a positive bias of the microwave retrieval relative to the visible retrieval that worsened with decreasing cloud fraction, consistent with the clear-sky results of G07. Horváth and Davies acknowledge that this bias could be due to problems with the visible retrieval, the microwave retrieval, or both. Thus, the errors due to beam-filling effects in the RSS retrieval are currently not clear and warrant future work.

d. Scattering by frozen particles

Scattering by cloud ice can be neglected for particles smaller than about 150 μm (Lin and Rossow 1994). For particles larger than this, scattering will cause a brightness temperature depression that is stronger at higher frequencies. However, the RSS algorithm neglects the effects of frozen precipitation because they are difficult to model and most strongly affect the 85-GHz channel of SSM/I, which is not used in the retrieval. It is a small effect at 19 GHz, but may have an effect on TBs in the 37-GHz channel for heavily precipitating systems with large amounts of bigger ice particles aloft. Because scattering by ice particles causes a brightness temperature depression, this effect will cause the retrieved LWP to be an underestimate of the true value (Liu and Curry 1993). Resulting errors will undoubtedly depend upon the particular retrieval algorithm, and unfortunately have not yet been studied for the RSS retrieval.

e. Cloud–rain partitioning

A particularly worrisome source of error for all passive microwave-based LWP climatologies is the treatment of LWP in the presence of rain. Because of the similar radiative effects of cloud and rainwater at microwave frequencies, their separation cannot be accomplished based on low-frequency microwave observations alone. In the case of the RSS algorithm, this partitioning relies on the assumption presented in Eq. (1), which as previously stated yields good agreement in the distribution of global rainfall coverage as compared with other established rainfall climatologies. However, this by no means is a guarantee of its accuracy. In general, large systematic errors can potentially result from the assumption. It is also important to stress that all microwave-based retrievals of LWP must make an assumption regarding this partitioning of cloud water and rainwater, and are therefore subject to errors resulting from it.

Let us now explore the potential errors from the rain/cloud separation assumption. Because the true distribution of cloud water in the presence of precipitation is not yet known, we explore the differences in the LWP climatology when different assumptions for the LWP/rain separation are used. We make two simple alternative assumptions in this regard. Assumption ALT assumes a separation of the form of Eq. (1) but with the parameter C = 0.12 (instead of the RSS value of 0.18). However, this suffers a slight deficiency in that it will produce an occurrence of rainfall that is too high relative to observations. Thus we also choose a simple second assumption, assumption PETTY, which states that the RSS assumption is valid for LWP less than 0.5 kg m−2; for higher values, the LWP is fixed at 0.5 kg m−2 and any additional liquid water exists as rain. This assumption was made in the rainfall retrieval algorithm of Petty (1994).

Figure 4a illustrates these three rain–cloud separation assumptions, in which the fraction of the total liquid water existing as cloud, fL, is plotted versus the vertical integral of the cloud plus rainwater [referred to as total water path (TWP)]. All assumptions exhibit decreasing fL with TWP and are nearly independent of H (not shown). Generally speaking, the ALT assumption has 10%–15% more water classified as rain (and hence less cloud) as compared with the RSS assumption. The PETTY assumption is the same as the RSS assumption until roughly a value of 0.75 kg m−2 in TWP; then fL rapidly decreases as all additional water is categorized as rain. For comparison, Fig. 4a also shows the same relationship as produced by a numerical weather model. Specifically, the light-gray shading denotes the 68% confidence limit of the ECMWF model profile dataset described in section 4b. For TWP values less than 1–2 kg m−2, the mean model relationship exhibits a qualitatively similar partitioning as compared with the various assumptions, though quantitatively the model produces significantly less cloud for a given amount of total liquid water than any of the observations for a large range in TWP. This is perhaps to be expected since this model is known to generally overestimate precipitation frequency (Chevallier and Bauer 2003).

We next explore the LWP results produced by the alternative assumptions. Figure 4b shows the zonal mean LWP produced by the RSS assumption as well as the two alternative assumptions. As expected, assumption ALT systematically finds 10%–15% less LWP than the RSS assumption. The PETTY assumption produces even less cloud in the ITCZ peak than the ALT assumption, but at higher latitudes it approaches the RSS result; this is simply because higher values of TWP tend to occur in the tropics, and it is only for these high values of TWP that the PETTY assumption differs from the control assumption. The overall conclusion to be drawn here is that whatever partitioning assumption is made may lead to systematic errors in LWP on the order of tens of percent; specifically how much is unknown because of the lack of observations to constrain this relationship. For instance, if the mean partitioning relationship from the ECMWF model is used, calculations indicate that the errors can approach 50%. The zonal situation for rainfall is shown in Fig. 4c, where it is evident that rain rate is even more sensitive to the partitioning assumption than is cloud water, especially outside of the tropics. Indeed, the zonal mean rain rate at higher latitudes varies by some 50%–100% between the three assumptions!

It is tempting to attempt to address the cloud–rain partitioning assumption by comparing to a standard climatology of rainfall rate, such as the commonly used Global Precipitation Climatology Project (Adler et al. 2003). However, over oceans this product relies primarily on SSM/I-based retrievals of rain rate for its calibration and, as such, faces the same fundamental limitations as described above.

How is this systematic error in LWP due to the cloud–rain partitioning assumption expected to be geographically distributed? Figure 5 compares an estimate of this systematic LWP error (upper panel) to the mean annual statistical (random) error in LWP (lower panel). The quantity in the upper panel is the percent difference on the mean annual LWP between either the PETTY or the ALT assumption, whichever is worse, as compared with the climatology constructed using the standard RSS assumption. As expected, the error tends to be larger in the regions of higher mean rainfall. The quantity in the lower panel was constructed by taking the statistical error on the mean LWP for a given month and year, as produced by the methodology described in section 3, and averaging it over all months and years to give an estimate of the average statistical error on a monthly mean LWP value. The faint horizontal lines at ±30° latitude are due to TMI. Certain areas in the ITCZ, as well as some coastal areas, show larger fractional statistical errors; this latter effect is primarily due to lower sampling in those regions. These mean statistical errors range from about 5% to 30% with a mean statistical error of about 15%. Note that, while the systematic error will be relatively constant over time, the statistical error on any LWP quantity will “integrate down” as longer time periods are averaged; the statistical error for a single month was selected arbitrarily for comparison. It is thus clear that the systematic and statistical errors are roughly comparable on the monthly time scale, but have somewhat different regional distributions. On longer time scales, systematic errors such as due to the cloud/rain partitioning assumption can be expected to dominate the overall error on LWP.

5. Comparison of mean LWP to previous climatologies

Keeping both the statistical and potential systematic error sources in mind, let us now examine the resulting LWP climatology itself. The basic features of our LWP climatology, hereafter referred to as the University of Wisconsin (UWisc) climatology, are consistent with previous observations. The mean LWP maps for January, April, July, and October are shown in Fig. 6. Mean values range from less than 10 g m−2 in the driest parts of the world, such as the Mediterranean Sea in July and the waters near West Africa in January, to greater than 300 g m−2 in areas of frequent precipitation, in particular the ITCZ and Indian monsoon regions. Most years, especially La Niña years, exhibit the well-known double ITCZ in the eastern Pacific during March and April, as explained for instance by Lietzke et al. (2001); this is faintly visible in Fig. 6b.

Though the basic features of the mean LWP are as expected, there are notable differences with previous climatologies. Figure 7 shows the UWisc zonal mean LWP compared against the climatologies of G95 and W97, seasonally averaged for both boreal winter [December–February (DJF)] and summer [June–August (JJA)]. The gray shading denotes an estimated error range on the UWisc result owing to the cloud–rain partitioning assumption discussed in section 4e. The UWisc climatology is found to exhibit substantially higher LWP in the tropics from roughly −20° to + 10° in latitude, as well as at high latitudes in boreal winter. The former discrepancy is most likely due to precipitation screening in the previous climatologies causing a low bias in LWP (especially in the G95 retrieval, which screens out all precipitation); the higher LWP in the polar latitudes has no obvious explanation but could be due to differences in sea ice screening, or more likely basic algorithmic differences. Generically, the UWisc climatology agrees better with G95 in the midlatitudes and better with W97 in the tropics.

It is also instructive to compare our results with those from reanalysis products in order to identify potential model deficiencies or issues with the derived LWP observations. In particular, we compare our results with those from ERA-40 (Uppala et al. 2005); the NCEP–NCAR reanalysis is left out of the comparison since it does not contain the vertically integrated LWP. Note that the LWP product in the ECMWF reanalysis is not driven directly by observations of clouds; rather, it is observations of temperature and moisture profiles that drive the moist physics parameterizations in the model in order to produce clouds. Thus, differences between the reanalysis and observations in terms of LWP can be due to either deficiencies in the observations or in the moist physics model. The zonal mean LWP for ERA-40 is shown as a dotted line in Fig. 7. It is noteworthy that all of the observational climatologies are significantly lower than that predicted from ERA40, even taking into account potential systematic errors. In particular, the UWisc climatology is an average of 40–50 g m−2 lower than the ERA-40 reanalysis; this difference was found to be much more pronounced in regions of strong convective activity, specifically the ITCZ and midlatitude storm track regions. This is likely explained from the fact that the moist physics parameterization in the model tends to produce too much rain and cloud water nearly everywhere, though it can fail to capture the moist, intense deep convective events (Chevallier and Bauer 2003). It seems unlikely that errors in the observations could be of a large enough magnitude to account for the discrepancy with ERA-40. For example, both the cloud–rain partitioning effect and the clear-sky bias likely make the observations overestimates of LWP. However, as stated in section 4, the neglect of ice scattering in the RSS retrieval will generally bias the retrieved LWP low. Could this account for the discrepancy? This also seems unlikely, as there are some regions that exhibit a large LWP difference between UWisc and ERA-40, such as the tropical eastern Pacific, but contain only small amounts of ice on average (Berg et al. 2002). While it is possible that there is some unknown bias in all microwave estimates of LWP that results in a significant underestimation of LWP, the available evidence indicates that the ERA-40 LWP is simply too high.

In terms of climate, it is useful to compare the temporal variability in monthly mean LWP captured by the various datasets. Figure 8a shows the correlation coefficient of the monthly mean LWP between all possible pairings of the climatologies. The UWisc climatology exhibits higher correlations with W97 in the tropics and with G95 in the extratropics, consistent with the previous findings of Fig. 7. Of the three observational climatologies, the UWisc climatology has the highest correlations with the ERA-40 data; this is true for all latitudes. The correlations between any pair of observational datasets is higher than the correlation of any observational dataset with the reanalysis; this is not surprising since they all stem from ultimately the same data source. Because the seasonal cycle is often the dominant temporal pattern in the data for most locations, the above graphic was repeated but with the seasonal cycle removed from all datasets and the correlations recalculated. It is clear that the same relationships among the various correlations are unchanged, and thus the common temporal patterns are not simply due to the seasonal cycle alone.

The correlation between the UWisc LWP and that of the ERA-40 reanalysis is further explored in Fig. 9, which shows maps of both the correlation coefficient of the monthly mean LWP values (Fig. 9a) and the deseasonalized values (Fig. 9b) between these two products. Though the correlations are generally between 60% and 90%, a striking feature is the low, even negative, correlation in the subtropical stratocumulus (Sc) regions. The correlation is worst in the Sc regions in the Southern Hemisphere off the west coasts of Africa and South America, but low correlation also is evident in the stratocumulus area to the west of California. This is perhaps not surprising, given that most models tend to have a great deal of trouble simulating clouds in these particular regions (e.g., see Bony and Dufresne 2005). The Sc observations are expected to be quite accurate due to the fact that there is rarely precipitation or cloud ice present, further indicating that the difference is most likely due to the model.

A study of the time series of mean monthly LWP in the Sc regions indicates that the model incorrectly predicts the seasonal cycle of LWP there. Observations show that the LWP experiences a maximum near September and a minimum near February in the southern Sc regions. In contrast, the reanalysis finds that the LWP peaks around March and experiences a minimum around September, almost completely out of phase with the observations. Upon removal of the seasonal cycle from the LWP mean monthly time series, the correlations as shown in Fig. 9b between the new climatology and the observations do not change a great deal. The correlations are still relatively high in the same locations, and relatively low in the Sc regions, especially in the Southern Hemisphere. The general conclusion remains that stratocumulus regions are poorly represented in ERA-40.

6. LWP diurnal cycle features

Many authors have previously investigated the diurnal cycles of cloud cover and precipitation. The strongest diurnal cycles in clouds and precipitation are generally found over land where solar-driven diurnal heating leads to a maximum in convection in the late afternoon (e.g., Hendon and Woodberry 1993; Yang and Slingo 2001). A weak morning maximum in convection over the tropical oceans has been identified by several authors (Wylie and Woolf 2002; Janowiak et al. 1994; Chang et al. 1995). Gravity waves and other mechanisms can link the diurnal cycles between land and ocean, for instance, as reported by Yang and Slingo (2001). Several studies have noted a relatively large diurnal cycle in cloudiness in the marine stratocumulus regions (Rozendaal et al. 1995; W02; Garreaud and Muñoz 2004).

The cloud fraction and precipitation diurnal studies still leave open the issue of the diurnal cycle of LWP, which is a direct measurement of the hydrological cycle and thus can potentially provide an important constraint on global models. The diurnal cycle of LWP described in W02 was somewhat noise limited, as it used only two years of data from a single satellite.

Quite a large number of interesting features become apparent with the 56 satellite-years of data in the present dataset. Figure 10 shows the relative amplitude and phase of the diurnal cycle for the months of January and July. Generally speaking, the diurnal cycle is strongest near coasts, in the subtropical stratocumulus regions, and in the monsoon regions. These regions can experience a diurnal cycle amplitude that is 50% or more of their mean LWP; this is much larger than is associated with the diurnal cycle in cloud fraction, which is rarely higher than 10% (e.g., Rozendaal et al. 1995; Wylie and Woolf 2002). Note also the evident asymmetry between the Northern and Southern Hemispheres; the stratocumulus regions experience weaker and less extended diurnal cycles during boreal summer than do the stratocumulus regions in austral summer. However, there is also a diurnal cycle associated with the Northern Hemisphere storm tracks in the boreal summer and with the Southern Hemisphere storm tracks during austral summer.

The local time of maximum LWP, referred to as the phase in Fig. 10, is relatively uniform over the oceans, typically peaking in the morning between 0400 and 0800 LT. In the higher latitudes, the maximum occurs nearer to 0200 LT, while several coastal regions, such as in the northern Gulf of Mexico and along the western coast of Central America, experience maximum LWP nearer to midday. This is also true for an extended region in the southeastern Atlantic Ocean to the east of Brazil in austral summer and in the tropical eastern Pacific in boreal summer.

The LWP diurnal cycle associated with the Indian monsoon in austral summer is complex and exhibits a diurnal cycle that is a strong function of location, as noted for instance by Yang and Slingo (2001). It is strongest in the western side of the Bay of Bengal, though the monsoon itself is stronger on the eastern side. Also, the time of peak LWP steadily increases as one moves eastward from the eastern coast of India, that is, beginning near 0500 LT at the coast, and, as one progresses eastward, to as late as 2300 LT in the center of the bay.

A basic sinusoidal shape with a period of 24 h is sufficient to account for the diurnal behavior of LWP for the majority of locations studied. One location stands out, however, in that it is distinctly not a simple sinusoid with one daily peak. This region is off the west coast of South America, and the amplitude of the second harmonic of the diurnal cycle fit is plotted in Fig. 11. This region exhibits a diurnal cycle with a strong second harmonic; in fact, it shows a distinctly double-peaked behavior. No other region around the globe stands out so strongly in this regard (although an area off the southwestern coast of Africa also exhibits similar behavior but to a lesser extent). The region of this interesting diurnal cycle off the west coast of South America is present throughout the year, but is roughly weakest in April and strongest in October. The diurnal fits for two specific locations within this region, as indicated by the filled black circles, are shown in Fig. 12. The more northerly location exhibits a “normal” diurnal cycle in April but a doubly peaked diurnal cycle in October, with peaks at 0700 and 1900 LT. The more southerly location exhibits a doubly peaked cycle all year long, with LWP peaks occurring at similar times. This region was examined in detail in the modeling study of Garreaud and Muñoz (2004). This study showed that the overall subtropical southeast Pacific is dominated by subsidence, but that a wave of positive vertical velocity (denoted an “upsidence wave”), driven partially by the presence of the high Andes mountains, often forms off the coast of Peru in the early evening, detaches from the coast, and moves out over the ocean in a southeasterly direction. This causes a strong diurnal cycle in vertical velocity in addition to temperature. It is possible that the double-peaked behavior of the LWP diurnal cycle is caused by the superposition of the diurnal cycles in solar heating and vertical velocity. This hypothesis is supported by the qualitative agreement of the diurnal cycle of the southernmost point (right-hand panels of Fig. 12) with the results of Garreaud and Muñoz (2004) from a single-column model study at 21°S, 76°W; in that study, the diurnal cycle in LWP exhibits two broad peaks: one in the morning and one in the evening. Further research is clearly required to fully understand the specific causes of this behavior.

7. Discussion

In this work we have presented a new climatology of column-integrated LWP over the global oceans, termed the UWisc climatology, based upon 18 years of passive microwave observations from eight satellite platforms. The climatology presented contains both a mean monthly diurnal cycle and monthly mean values of LWP.1

Estimates of the statistical errors on these values are included within the climatology. However, systematic errors are likely present in the record from a number of different sources, such as the assumption regarding cloud temperature, the beam-filling effect, and the clear-sky bias. We have also attempted to quantify the error owing to a previously unexplored source, the cloud–rain partitioning assumption. This error, endemic to microwave retrievals of LWP, is typically O(5%–30%) depending on the region and may dominate all other error sources, especially in regions of high average rainfall. In order for microwave-based cloud and rainfall climatologies to improve, the relationship between cloud and rainwater must be more thoroughly explored, ideally through separate observations of the cloud water and rain content of precipitating clouds. Additionally, a potentially promising avenue of research would be the retrieval of total liquid water (rain liquid water plus cloud liquid water), thus circumventing the large potential systematic error source in the separation of the two.

The resulting monthly mean LWP was next compared to previous observational climatologies, as well as ERA-40. It was found that the new climatology agrees better with ERA-40 than the previous climatologies, but all observations find significantly less LWP than the reanalysis. This is especially true in regions of higher convective activity.

The opposite was found to be true in the subtropical stratocumulus regions where, not only does the reanalysis make too little cloud, but the modeled seasonal cycle tends to completely disagree with observations. This is unsurprising given that most climate models have significant trouble in predicting clouds in marine stratocumulus regions (e.g., Bony and Dufresne 2005); the new climatology may therefore serve as a useful point of comparison for global models.

The diurnal cycle of LWP was found to be strong in both coastal regions and the stratocumulus regions, consistent with previous findings. These diurnal cycles tend to be stronger than the corresponding cycle in cloud fraction. The midlatitudes also exhibit moderate diurnal cycles, with amplitudes of 10%–20% relative to the mean LWP during local summer. Most diurnal cycles of LWP were well characterized by a single sine wave with a period of 24 h, but the region just west of South America, and to a lesser extent the region off the southwestern coast of Africa, exhibits a distinctly double-peaked behavior; further research into this interesting phenomenon is warranted.

With 18 years of data now available, it is tempting to investigate the possibility of long-term trends in the LWP data. Clouds serve as an important feedback in the climate system, and several mechanisms are believed to exist that could cause cloud parameters to change on regional and perhaps even global scales. If our understanding of the systematic error sources in the retrieval of LWP is sufficiently enhanced, it is possible that long-term regional and potentially even global trends in the cloud system may be investigated.

Acknowledgments

All SSM/I, TMI, and AMSR-E data used in the climatology construction were produced by Remote Sensing Systems (available online at http://www.remss.com), with partial sponsorship from the NASA AMSR-E science team, as well as the NASA Earth Science REASoN project (Research, Education and Applications Solution Network), a distributed network of data and information providers for NASA’s Earth Science Enterprise (ESE) Science, Applications and Education Programs. Liquid water data using the Greenwald et al. (1993) algorithm were produced at the NASA Langley Research Center; liquid water paths using the W97 algorithm were produced at the National Climatic Data Center, a division of NOAA/National Environmental Satellite, Data, and Information Service (NESDIS). The ERA-40 data were obtained through the Computational Information and Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR). Special thanks to Ralph Ferraro for help with accessing and interpreting the NOAA/NESDIS data.

REFERENCES

  • Adler, R. F., and Coauthors, 2003: The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–present). J. Hydrometeor., 4 , 11471167.

    • Search Google Scholar
    • Export Citation
  • Alishouse, J. C., , J. B. Snider, , E. R. Westwater, , C. T. Swift, , C. S. Ruf, , S. A. Snyder, , J. Vongsathorn, , and R. R. Ferraro, 1990: Determination of cloud liquid water content using the SSM/I. IEEE Trans. Geosci. Remote Sens., 28 , 817822.

    • Search Google Scholar
    • Export Citation
  • Bauer, P., , and P. Schluessel, 1993: Rainfall, total water, ice water, and water vapor over sea from polarized microwave simulations and special sensor microwave/imager data. J. Geophys. Res., 98 , D11. 2073720778.

    • Search Google Scholar
    • Export Citation
  • Bennartz, R., 2007: Global assessment of marine boundary layer cloud droplet number concentration from satellite. J. Geophys. Res., 112 .D02201, doi:10.1029/2006JD007547.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , C. Kummerow, , and C. A. Morales, 2002: Differences between East and West Pacific rainfall systems. J. Climate, 15 , 36593672.

    • Search Google Scholar
    • Export Citation
  • Bony, S., , and J-L. Dufresne, 2005: Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models. Geophys. Res. Lett., 32 .L20806, doi:10.1029/2005GL023851.

    • Search Google Scholar
    • Export Citation
  • Borg, L. A., , and R. Bennartz, 2007: Vertical structure of stratiform marine boundary layer clouds and its impact on cloud albedo. Geophys. Res. Lett., 34 .L05807, doi:10.1029/2006GL028713.

    • Search Google Scholar
    • Export Citation
  • Bremen, L. V., , E. Ruprecht, , and A. Macke, 2002: Errors in liquid water path retrieval arising from cloud inhomogeneities: The beam-filling effect. Meteor. Z., 11 , 1319.

    • Search Google Scholar
    • Export Citation
  • Chang, A. T. C., , L. S. Chiu, , and G. Yang, 1995: Diurnal cycle of oceanic precipitation from SSM/I data. Mon. Wea. Rev., 123 , 33713380.

    • Search Google Scholar
    • Export Citation
  • Chevallier, F., , and P. Bauer, 2003: Model rain and clouds over oceans: Comparison with SSM/I observations. Mon. Wea. Rev., 131 , 12401255.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , C. D. Ardeel, , and L. Tian, 1990: Liquid water content and precipitation characteristics of stratiform clouds as inferred from satellite microwave measurements. J. Geophys. Res., 95 , 1665916671.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , T. R. Karl, , B. Sun, , and K. E. Trenberth, 2006: Recent trends in cloudiness over the United States: A tale of monitoring inadequacies. Bull. Amer. Meteor. Soc., 87 , 597606.

    • Search Google Scholar
    • Export Citation
  • Garreaud, R. D., , and R. Muñoz, 2004: The diurnal cycle in circulation and cloudiness over the subtropical southeast Pacific: A modeling study. J. Climate, 17 , 16991710.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , G. L. Stephens, , T. H. Vonder Haar, , and D. L. Jackson, 1993: A physical retrieval of cloud liquid water over the global oceans using special sensor microwave/imager (SSM/I) observations. J. Geophys. Res., 98 , 1847118488.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , G. L. Stephens, , S. A. Christopher, , and T. H. Vonder Haar, 1995: Observations of the global characteristics and regional radiative effects of marine cloud liquid water. J. Climate, 8 , 29282946.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , S. A. Christopher, , and J. Chou, 1997: Cloud liquid water path comparisons from passive microwave and solar reflectance satellite measurements: Assessment of subfield-of-view cloud effects in microwave retrievals. J. Geophys. Res., 102 , 1958519596.

    • Search Google Scholar
    • Export Citation
  • Greenwald, T. J., , T. S. L’Ecuyer, , and S. A. Christopher, 2007: Evaluating specific error characteristics of microwave-derived cloud liquid water products. Geophys. Res. Lett., 34 .L22807, doi:10.1029/2007GL031180.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., , and K. Woodberry, 1993: The diurnal cycle of tropical convection. J. Geophys. Res., 98 , 1662316638.

  • Hilburn, K. A., , and F. J. Wentz, 2008: Intercalibrated passive microwave rain products from the Unified Microwave Ocean Retrieval Algorithm (UMORA). J. Appl. Meteor. Climatol., 47 , 778794.

    • Search Google Scholar
    • Export Citation
  • Horváth, A., , and R. Davies, 2007: Comparison of microwave and optical cloud water path estimates from TMI, MODIS, and MISR. J. Geophys. Res., 112 .D01202, doi:10.1029/2006JD007101.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., , P. A. Arkin, , and M. Morrissey, 1994: An examination of the diurnal cycle in oceanic tropical rainfall using satellite and in situ data. Mon. Wea. Rev., 122 , 22962311.

    • Search Google Scholar
    • Export Citation
  • Jung, T., , E. Ruprecht, , and F. Wagner, 1998: Determination of cloud liquid water path over the oceans from Special Sensor Microwave/Imager (SSM/I) data using neural networks. J. Appl. Meteor., 37 , 832844.

    • Search Google Scholar
    • Export Citation
  • Lietzke, C. E., , C. Deser, , and T. H. Vonder Haar, 2001: Evolutionary structure of the eastern Pacific double ITCZ based on satellite moisture profile retrievals. J. Climate, 14 , 743751.

    • Search Google Scholar
    • Export Citation
  • Lin, B., , and W. B. Rossow, 1994: Observations of cloud liquid water path over oceans: Optical and microwave remote sensing methods. J. Geophys. Res., 99 , D10. 2090720928.

    • Search Google Scholar
    • Export Citation
  • Liu, G., , and J. A. Curry, 1993: Determination of characteristic features of cloud liquid water from satellite microwave measurements. J. Geophys. Res., 98 , D3. 50695092.

    • Search Google Scholar
    • Export Citation
  • Mears, C. A., , D. K. Smith, , and F. J. Wentz, 2001: Comparison of Special Sensor Microwave Imager and buoy-measured wind speeds from 1987 to 1997. J. Geophys. Res., 106 , 1171911730.

    • Search Google Scholar
    • Export Citation
  • Mears, C. A., , M. C. Schabel, , and F. J. Wentz, 2003: A reanalysis of the MSU channel 2 tropospheric temperature record. J. Climate, 16 , 36503664.

    • Search Google Scholar
    • Export Citation
  • Meissner, T., , D. Smith, , and F. Wentz, 2001: A 10-year intercomparison between collocated Special Sensor Microwave Imager oceanic surface wind speed retrievals and global analyses. J. Geophys. Res., 106 , 1173111742.

    • Search Google Scholar
    • Export Citation
  • Norris, J. R., 1999: On trends and possible artifacts in global ocean cloud cover between 1952 and 1995. J. Climate, 12 , 18641870.

  • Petty, G. W., 1990: On the response of the special sensor microwave/imager to the marine environment—Implications for atmospheric parameter retrievals. Ph.D. thesis, University of Washington, 118 pp.

  • Petty, G. W., 1994: Physical retrievals of over-ocean rain rate from multichannel microwave imagery. Part II: Algorithm implementation. Meteor. Atmos. Phys., 54 , 101121.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., , N. A. Rayner, , T. M. Smith, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80 , 22612287.

  • Rozendaal, M. A., , C. B. Leovy, , and S. A. Klein, 1995: An observational study of diurnal variations of marine stratiform cloud. J. Climate, 8 , 17951809.

    • Search Google Scholar
    • Export Citation
  • Sohn, B-J., , and E. A. Smith, 2003: Explaining sources of discrepancy in SSM/I water vapor algorithms. J. Climate, 16 , 32293255.

  • Staelin, D. H., , K. F. Kunzi, , R. L. Pettyjohn, , R. K. L. Poon, , R. W. Wilcox, , and J. W. Waters, 1976: Remote sensing of atmospheric water vapor and liquid water with the Nimbus 5 microwave spectrometer. J. Appl. Meteor., 15 , 12041214.

    • Search Google Scholar
    • Export Citation
  • Thomas, S. M., , A. K. Heidinger, , and M. J. Pavolonis, 2004: Comparison of NOAA’s operational AVHRR-derived cloud amount to other satellite-derived cloud climatologies. J. Climate, 17 , 48054822.

    • Search Google Scholar
    • Export Citation
  • Tiedke, M., 1993: Representation of clouds in large-scale models. Mon. Wea. Rev., 121 , 30403061.

  • Trenberth, K. E., , J. Fasullo, , and L. Smith, 2005: Trends and variability in column-integrated atmospheric water vapor. Climate Dyn., 24 , 741758.

    • Search Google Scholar
    • Export Citation
  • Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

  • Vinnikov, K. Y., , and N. C. Grody, 2003: Global warming trend of mean tropospheric temperature observed by satellites. Science, 302 , 269272.

    • Search Google Scholar
    • Export Citation
  • Weng, F., , and N. C. Grody, 1994: Retrieval of cloud liquid water using the special sensor microwave imager (SSM/I). J. Geophys. Res., 99 , 2553525552.

    • Search Google Scholar
    • Export Citation
  • Weng, F., , N. C. Grody, , R. Ferraro, , A. Basist, , and D. Forsyth, 1997: Cloud liquid water climatology from the Special Sensor Microwave/Imager. J. Climate, 10 , 10861098.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., 1997: A well-calibrated ocean algorithm for special sensor microwave imager. J. Geophys. Res., 102 , 87038718.

  • Wentz, F. J., , and R. W. Spencer, 1998: SSM/I rain retrievals within a unified all-weather ocean algorithm. J. Atmos. Sci., 55 , 16131627.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , and T. F. Meissner, 2000: Algorithm Theoretical Basis Document (ATBD) version 2, AMSR ocean algorithm. RSS Tech. Proposal 121599A-1, 59 pp.

  • Wentz, F. J., , and M. Schabel, 2000: Precise climate monitoring using complementary satellite data sets. Nature, 403 , 414416.

  • Wentz, F. J., , L. Ricciardulli, , K. A. Hilburn, , and C. Mears, 2007: How much more rain will global warming bring? Science, 317 , 233235.

    • Search Google Scholar
    • Export Citation
  • Wood, R., , C. S. Bretherton, , and D. L. Hartmann, 2002: Diurnal cycle of liquid water path over the subtropical and tropical oceans. Geophys. Res. Lett., 29 .2092, doi:10.1029/2002GL015371.

    • Search Google Scholar
    • Export Citation
  • Wylie, D. P., , and H. M. Woolf, 2002: The diurnal cycle of upper-tropospheric clouds measured by GOES-VAS and the ISCCP. Mon. Wea. Rev., 130 , 171179.

    • Search Google Scholar
    • Export Citation
  • Yang, G-Y., , and J. Slingo, 2001: The diurnal cycle in the tropics. Mon. Wea. Rev., 129 , 784801.

Fig. 1.
Fig. 1.

Monthly mean LWP for the six SSM/Is, AMSR-E, and TMI, averaged over (a) the Niño-3.4 region (10°S–10°N, 170°–120°W) and (b) the tropics (30°S–30°N). Note the different vertical scales on the y axes. Gaps in the data (especially for the F08 SSM/I) occurred when the data were too sparse within a particular month to form a representative average.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 2.
Fig. 2.

Estimated intramonthly LWP standard deviation for January. These data were used as input to the algorithm fitting for the mean diurnal cycle and mean yearly LWP; its value only affects the estimated fitting errors. Black pixels denote land while gray pixels denote missing data, either from the presence of sea ice or the close proximity of land.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 3.
Fig. 3.

Example of the diurnal cycle fitting results for a 1° grid box near the northern coast of the Indonesian island of Borneo (3.5°S, 111.5°E) for the month of January, using different combinations of sensors. LWP observations taken at similar local times from the same sensor have been averaged for visibility, with corresponding one-sigma error bars shown. “Amplitude” represents 50% of the diurnal range of LWP, while “Phase” represents the local time of maximum LWP; each has been displayed along with the corresponding one-sigma error. Diurnal cycle results using (a) only SSM/I data, (b) both SSM/I and AMSR-E data, (c) only TMI data, and (d) all available data, (e) same fit as (d) but, for display purposes, the data points have been corrected for the difference in the mean LWP for the year in which they were taken and (f) mean LWP for each year when there was data along with the one-sigma error bars, corresponding to fit from (e).

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 4.
Fig. 4.

(a) Relationship between the fraction of LWP contained in clouds ( fL) and the TWP in the ECMWF forecast model vs the RSS retrieval assumption (black line), as well as two alternate assumptions: ALT (gray line) and PETTY (black dot–dashed line); see text for details. (b) Zonal mean LWP from standard RSS algorithm as well as the two alternative assumptions; (c) as in (b) but for the zonal mean rain rate.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 5.
Fig. 5.

(top) Estimate of the systematic errors due to the cloud–rain separation assumption, and (bottom) average monthly statistical errors on LWP from UWisc climatology.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 6.
Fig. 6.

Mean values of LWP averaged over the years 1988–2005 for (a) January, (b) April, (c) July, and (d) October. Black pixels denote land, while gray pixels denote missing data, either from the presence of sea ice or the close proximity of land.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 7.
Fig. 7.

Mean zonal cloud liquid water path over ocean for (a) DJF and (b) JJA from three different observational climatologies and one meteorological reanalysis. The G95 climatology contains data from F08 from July 1987 through December 1991, while both the UWisc climatology and the W97 dataset range from 1988 to 2005. The displayed ERA-40 data range from 1988 to 2002. The gray shading denotes the estimated range of uncertainty in the UWisc climatology due to the cloud–rain partitioning assumption discussed in section 4e.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 8.
Fig. 8.

Zonally averaged correlation in the (a) mean monthly and (b) deseasonalized mean monthly LWP time series between various climatologies. ERA refers to the ERA-40 and UWisc to the present work. All comparisons were done at 1° resolution. Note that the G95 climatology only includes data through 1991.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 9.
Fig. 9.

(a) Correlation coefficient in the time series of mean monthly LWP between the ERA-40 reanalysis and the UWisc climatology, from January 1988 through August 2002; (b) as in (a) but the time series have been deseasonalized before the calculation of the correlation coefficients. Note the different scales between (a) and (b).

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 10.
Fig. 10.

Relative amplitude and phase of the diurnal cycle in the months of January and July. The diurnal cycle phase is the local time of maximum LWP. Black pixels denote land, while gray pixels denote missing data, from either the presence of sea ice or the close proximity of land; (b), (d) gray pixels also indicate locations without a well-defined diurnal maximum in LWP.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 11.
Fig. 11.

Map of the amplitude of the second harmonic of the diurnal cycle for a region off the coast of South America for the months of April and October. The plotted region extends from the equator to 40°S, 70°–110°W.

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Fig. 12.
Fig. 12.

Diurnal cycle fits for the two small regions off the west coast of South America, depicted by filled black circles in Fig. 11 for the months of (top) April and (bottom) October. The more northern location has a single-peaked diurnal cycle in April but a strong double-peaked cycle in October; the more southern location has a double peak in both months, and in fact throughout the year (not shown).

Citation: Journal of Climate 21, 8; 10.1175/2007JCLI1958.1

Table 1.

Properties of microwave sensors used in this work.

Table 1.

1

The full climatology may be obtained by sending an e-mail query to the corresponding author.

Save