• Backus, G., , and F. Gilbert, 1970: Uniqueness in the inversion of inaccurate gross earth data. Philos. Trans. Roy. Soc. London, A266 , 123192.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , T. L’Ecuyer, , and C. Kummerow, 2006: Rainfall climate regimes: The relationship of regional TRMM rainfall biases to the environment. J. Appl. Meteor. Climatol., 45 , 434454.

    • Search Google Scholar
    • Export Citation
  • Bladé, I., , and D. L. Hartmann, 1993: Tropical intraseasonal oscillations in a simple nonlinear model. J. Atmos. Sci., 50 , 29222939.

  • Deblonde, G., , and S. J. English, 2001: Evaluation of the FASTEM-2 fast microwave ocean surface emissivity model. Tech. Proc. 11th Int. TOVS Study Conf., Budapest, Hungary, 67–78.

    • Search Google Scholar
    • Export Citation
  • Del Genio, A. D., , W. Kovari, , M. S. Yao, , and J. Jonas, 2005: Cumulus microphysics and climate sensitivity. J. Climate, 18 , 23762387.

  • Elsaesser, G. S., , and C. D. Kummerow, 2008: Toward a fully parametric retrieval of the nonraining parameters over the global oceans. J. Appl. Meteor. Climatol., 47 , 15991618.

    • Search Google Scholar
    • Export Citation
  • Farrar, M. R., , and E. A. Smith, 1992: Spatial-resolution enhancement of terrestrial features using deconvolved SSM/I microwave brightness temperatures. IEEE Trans. Geosci. Remote Sens., 30 , 349355.

    • Search Google Scholar
    • Export Citation
  • Han, Q., , W. B. Rossow, , and A. A. Lacis, 1994: Near-global survey of effective droplet radii in liquid water clouds using ISCCP data. J. Climate, 7 , 465497.

    • Search Google Scholar
    • Export Citation
  • Hartmann, D. L., , and K. Larson, 2002: An important constraint on tropical cloud–climate feedback. Geophys. Res. Lett., 29 , 1951. doi:10.1029/2002GL015835.

    • Search Google Scholar
    • Export Citation
  • Haynes, J., 2008: The near-global distribution of light rain from CloudSat. Ph.D. dissertation, Dept. of Atmospheric Science, Colorado State University, 229 pp.

  • Hu, Q., , and D. A. Randall, 1994: Low-frequency oscillations in radiative–convective systems. J. Atmos. Sci., 51 , 10891099.

  • Iguchi, T., , T. Kozu, , R. Meneghini, , J. Awaka, , and K. Okamoto, 2000: Rain-profiling algorithm for the TRMM precipitation radar. J. Appl. Meteor., 39 , 20382052.

    • Search Google Scholar
    • Export Citation
  • Johnson, R. H., , T. M. Rickenbach, , S. A. Rutledge, , P. E. Ciesielski, , and W. H. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12 , 23972418.

    • Search Google Scholar
    • Export Citation
  • Johnson, R. H., , P. E. Ciesielski, , and J. A. Cotturone, 2001: Multiscale variability of the atmospheric mixed layer over the western Pacific warm pool. J. Atmos. Sci., 58 , 27292750.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kemball-Cooke, S. R., , and B. C. Weare, 2001: The onset of convection in the Madden–Julian oscillation. J. Climate, 14 , 780793.

  • Khairoutdinov, M., , and Y. Kogan, 2000: A new cloud physics parameterization in a large-eddy simulation model of marine stratocumulus. Mon. Wea. Rev., 128 , 229243.

    • Search Google Scholar
    • Export Citation
  • King, M. D., , S-C. Tsay, , S. E. Platnick, , M. Wang, , and K-N. Liou, 1997: Cloud retrieval algorithms for MODIS: Optical thickness, effective particle radius, and thermodynamic phase. MODIS Algorithm Theoretical Basis Doc. ATBD-MOD-05, MOD06-Cloud product, 83 pp.

    • Search Google Scholar
    • Export Citation
  • Kohn, D. J., 1995: Refinement of a semi-empirical model for the microwave emissivity of the sea surface as a function of wind speed. M.S. thesis, Dept. of Meteorology, Texas A&M University, 44 pp.

  • Kummerow, C., , W. Barnes, , T. Kozu, , J. Shiue, , and J. Simpson, 1998: The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Oceanic Technol., 15 , 808816.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and H. T. Wu, 2003: Warm rain processes over tropical oceans and climate implications. Geophys. Res. Lett., 30 , 2290. doi:10.1029/2003GL018567.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , H. T. Wu, , Y. C. Sud, , and G. K. Walker, 2005: Effects of cloud microphysics on tropical atmospheric hydrologic processes and intraseasonal variability. J. Climate, 18 , 47314751.

    • Search Google Scholar
    • Export Citation
  • Liebe, H. J., , G. A. Hufford, , and T. Manabe, 1991: A model for the complex permittivity of water at frequencies below 1 THz. Int. J. Infrared Millimeter Waves, 12 , 659675.

    • Search Google Scholar
    • Export Citation
  • Liebe, H. J., , G. A. Hufford, , and M. G. Cotton, 1993: Propagation modeling of moist air and suspended water particles at frequencies below 1000 GHz. Proc. Conf. on Atmospheric Propagation Effects through Natural and Man-Made Obscurants for Visible to MM-Wave Radiation, AGARD-CP-542, Neuilly sur Seine, France, Advisory Group for Aerospace Research and Development, 3-1–3-10.

    • Search Google Scholar
    • Export Citation
  • Lindzen, R. S., , M-D. Chou, , and A. Hou, 2001: Does the earth have an adaptive infrared iris? Bull. Amer. Meteor. Soc., 82 , 417432.

  • Long, D. G., , and D. L. Daum, 1998: Spatial resolution enhancement of SSM/I data. IEEE Trans. Geosci. Remote Sens., 36 , 407417.

  • Marks, C. J., , and C. D. Rodgers, 1993: Determination of characteristic features of cloud liquid water from satellite microwave measurements. J. Geophys. Res., 98 , 50695092.

    • Search Google Scholar
    • Export Citation
  • Marshall, J. S., , and W. M. Palmer, 1948: The distribution of raindrops with size. J. Meteor., 5 , 165166.

  • Nakajima, T., , and M. D. King, 1990: Determinations of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I: Theory. J. Atmos. Sci., 47 , 18781893.

    • Search Google Scholar
    • Export Citation
  • Nakajima, T. Y., , and T. Nakajima, 1995: Wide-area determination of cloud microphysical properties from NOAA AVHRR measurements for FIRE and ASTEX regions. J. Atmos. Sci., 52 , 40434059.

    • Search Google Scholar
    • Export Citation
  • O’Dell, C. W., , F. J. Wentz, , and R. Bennartz, 2008: Cloud liquid water path from satellite-based passive microwave observations: A new climatology over the global oceans. J. Climate, 21 , 17211739.

    • Search Google Scholar
    • Export Citation
  • Petty, G. W., 1999: Prevalence of precipitation from warm-topped clouds over eastern Asia and the western Pacific. J. Climate, 12 , 220229.

    • Search Google Scholar
    • Export Citation
  • Platnick, S., , and F. P. Valero, 1995: A validation of a satellite cloud retrieval during ASTEX. J. Atmos. Sci., 52 , 29853001.

  • Platnick, S., , M. D. King, , S. A. Ackerman, , W. P. Menzel, , B. A. Baum, , J. C. Riédi, , and R. A. Frey, 2003: The MODIS cloud products: Algorithms and examples from Terra. IEEE Trans. Geosci. Remote Sens., 41 , 459473.

    • Search Google Scholar
    • Export Citation
  • Ramanathan, V., , and W. Collins, 1991: Thermodynamic regulation of ocean warming by cirrus clouds deduced from observations of the 1987 El Niño. Nature, 351 , 2732.

    • Search Google Scholar
    • Export Citation
  • Rapp, A. D., , C. D. Kummerow, , W. Berg, , and B. Griffith, 2005: An evaluation of the proposed mechanism of the adaptive infrared iris hypothesis using TRMM VIRS and PR measurements. J. Climate, 18 , 41854194.

    • Search Google Scholar
    • Export Citation
  • Raymond, D. J., , and D. J. Torres, 1998: Fundamental moist modes of the equatorial troposphere. J. Atmos. Sci., 55 , 17711790.

  • Robinson, W., , C. Kummerow, , and W. S. Olson, 1992: A technique for matching the resolution of microwave measurements from the SSM/I instrument. IEEE Trans. Geosci. Remote Sens., 30 , 419429.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 1976: Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Rev. Geophys. Space Phys., 4 , 609624.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 2000: Inverse Methods For Atmospheric Sounding: Theory and Practice. World Scientific, 238 pp.

  • Rosenfeld, D., , and G. Gutman, 1994: Retrieving microphysical properties near the tops of potential rain clouds by multispectral analysis of AVHRR data. Atmos. Res., 34 , 259283.

    • Search Google Scholar
    • Export Citation
  • Rosenkranz, P. W., 1998: Water vapor microwave continuum absorption: A comparison of measurements and models. Radio Sci., 33 , 919928.

    • Search Google Scholar
    • Export Citation
  • Simpson, J., , C. Kummerow, , W. K. Tao, , and R. F. Adler, 1996: On the Tropical Rainfall Measuring Mission (TRMM). Meteor. Atmos. Phys., 60 , 1936.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2002: The CloudSat mission and the A-Train: A new dimension of space-based observations of clouds and precipitation. Bull. Amer. Meteor. Soc., 83 , 17711790.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., , P. J. Webster, , R. H. Johnson, , R. Engelen, , and T. L’Ecuyer, 2004: Observational evidence for the mutual regulation of the tropical hydrological cycle and tropical sea surface temperature. J. Climate, 17 , 22132224.

    • Search Google Scholar
    • Export Citation
  • Stogryn, A., 1978: Estimates of brightness temperatures from scanning radiometer data. IEEE Trans. Antennas Propagat., AP-26 , 720726.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , and T. Meissner, 2000: AMSR Ocean Algorithm: ATBD version 2. RSS Tech. Proposal 121599A-1, 66 pp. [Available online at http://www.remss.com/papers/amsr/AMSR_Ocean_Algorithm_Version_2.pdf].

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , C. Gentemann, , D. Smith, , and D. Chelton, 2000: Satellite measurements of sea surface temperatures through clouds. Science, 288 , 847850.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. T., 1979a: The effect of wind on the microwave emission from the ocean’s surface at 37 GHz. J. Geophys. Res., 84 , 49214926.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. T., 1979b: A model for the microwave emissivity of the ocean’s surface as a function of wind speed. IEEE Trans. Geosci. Electron., 17 , 244249.

    • Search Google Scholar
    • Export Citation
  • Zuidema, P., , E. R. Westwater, , C. Fairall, , and D. Hazen, 2005: Ship-based liquid water path estimates in marine stratocumulus. J. Geophys. Res., 110 , D20206. doi:10.1029/2005JD005833.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Mean LWP (g m−2) with SST (K) for nonraining warm clouds for (a) MODIS and (b) AMSR-E and for raining warm clouds for (c) MODIS and (d) AMSR-E.

  • View in gallery

    PR attenuation-corrected radar reflectivity (dBZ) vs rainwater content (kg m−3).

  • View in gallery

    Difference in parameters (a) LWP (g m−2), (b) TPW (mm), (c) wind speed (m s−1), and (d) χ2, retrieved with and without cloud fraction, plotted against VIRS cloud fraction.

  • View in gallery

    TMI OE-retrieved parameters (a) LWP (g m−2), (b) TPW (mm), (c) wind speed (m s−1), and (d) χ2, with and without PR rainwater estimates.

  • View in gallery

    Portion of TRMM swath containing warm rain clouds: (a) VIRS infrared TB (K), (b) PR rain rate (mm h−1), (c) optimal estimation LWP (g m−2) without PR rainwater estimation, and (d) optimal estimation LWP (g m−2) using PR rainwater estimate.

  • View in gallery

    Mean TMI OE-retrieved LWP (g m−2) vs SST (K) for (a) nonraining and (b) raining warm clouds.

  • View in gallery

    Ratio of mean TMI OE-retrieved LWP to PR-estimated rainwater vs SST (K).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 21 21 5
PDF Downloads 12 12 4

A Combined Multisensor Optimal Estimation Retrieval Algorithm for Oceanic Warm Rain Clouds

View More View Less
  • 1 Cooperative Institute for Research in Environmental Sciences, University of Colorado, and NOAA Earth System Research Laboratory, Boulder, Colorado
  • | 2 Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado
© Get Permissions
Full access

Abstract

The complicated interactions between cloud processes in the tropical hydrologic cycle and their responses to changes in environmental variables have been the focus of many recent investigations. Most studies that examine the response of the hydrologic cycle to temperature changes focus on deep convection and cirrus production, but recent results suggest that warm rain clouds may be more sensitive to temperature changes. These clouds are prevalent in the tropics and make considerable contributions to the radiation budget and to total tropical rainfall, as well as serving to moisten and precondition the atmosphere for deep convection. A change in the properties of these clouds in climate-change scenarios could have significant implications for the hydrologic cycle. Existing microwave and visible retrievals of warm rain cloud liquid water path (LWP) disagree over the range of sea surface temperatures (SST) observed in the tropical western Pacific Ocean. Although both retrieval methods show similar behavior for nonraining clouds, the two methods show very different warm-rain-cloud LWP responses to SST, both in magnitude and trend. This makes changes to the relationship between precipitation and cloud properties in changing temperature regimes difficult to interpret. A combined optimal estimation retrieval algorithm that takes advantage of the strengths of the different satellite measurements available on the Tropical Rainfall Measuring Mission (TRMM) satellite has been developed. Deconvolved TRMM Microwave Imager brightness temperatures are combined with cloud fraction from the Visible and Infrared Scanner and rainwater estimates from the TRMM precipitation radar to retrieve the cloud LWP in warm rain systems. This algorithm is novel in that it takes into account the water in the rain and estimates the LWP due to only the cloud water in a raining cloud, thus allowing investigation of the effects of precipitation on cloud properties.

Corresponding author address: Anita D. Rapp, CIRES, University of Colorado, and NOAA/ESRL, 325 Broadway, Boulder, CO 80305. Email: anita.d.rapp@noaa.gov

Abstract

The complicated interactions between cloud processes in the tropical hydrologic cycle and their responses to changes in environmental variables have been the focus of many recent investigations. Most studies that examine the response of the hydrologic cycle to temperature changes focus on deep convection and cirrus production, but recent results suggest that warm rain clouds may be more sensitive to temperature changes. These clouds are prevalent in the tropics and make considerable contributions to the radiation budget and to total tropical rainfall, as well as serving to moisten and precondition the atmosphere for deep convection. A change in the properties of these clouds in climate-change scenarios could have significant implications for the hydrologic cycle. Existing microwave and visible retrievals of warm rain cloud liquid water path (LWP) disagree over the range of sea surface temperatures (SST) observed in the tropical western Pacific Ocean. Although both retrieval methods show similar behavior for nonraining clouds, the two methods show very different warm-rain-cloud LWP responses to SST, both in magnitude and trend. This makes changes to the relationship between precipitation and cloud properties in changing temperature regimes difficult to interpret. A combined optimal estimation retrieval algorithm that takes advantage of the strengths of the different satellite measurements available on the Tropical Rainfall Measuring Mission (TRMM) satellite has been developed. Deconvolved TRMM Microwave Imager brightness temperatures are combined with cloud fraction from the Visible and Infrared Scanner and rainwater estimates from the TRMM precipitation radar to retrieve the cloud LWP in warm rain systems. This algorithm is novel in that it takes into account the water in the rain and estimates the LWP due to only the cloud water in a raining cloud, thus allowing investigation of the effects of precipitation on cloud properties.

Corresponding author address: Anita D. Rapp, CIRES, University of Colorado, and NOAA/ESRL, 325 Broadway, Boulder, CO 80305. Email: anita.d.rapp@noaa.gov

1. Introduction

In an effort to understand better the response of the hydrologic cycle to climate feedbacks, many recent studies have examined the response of tropical cloud processes to changes in SST. The majority of these studies (e.g., Ramanathan and Collins 1991; Lindzen et al. 2001; Hartmann and Larson 2002) focused on deep convective clouds because of their contribution to total tropical rainfall and the large radiative impacts from the detrained anvil cirrus. However, in the midst of the controversy surrounding many of the theories regarding the response of deep convection to climate change, several studies have suggested another cloud type in the tropics that may be sensitive to and important for our understanding of the response of the hydrologic cycle to anthropogenic warming. Petty (1999) found warm-topped clouds to be important to the population of precipitating clouds in the tropics and suggested that the prevailing satellite retrieval methods—namely, infrared and scattering-based passive microwave—might be inadequate to resolve these clouds. At the same time, Johnson et al. (1999) drew attention to the importance of cumulus congestus clouds and suggested a new conceptual model of the tropical hydrologic cycle including this third mode of midlevel clouds, moving away from the more typical model of a bimodal cloud distribution of deep cumulonimbus and trade winds cumulus. In the more recent fourth Intergovernmental Panel on Climate Change assessment, variability in general circulation model (GCM) cloud feedback effects was mostly attributed to differences in the models’ shortwave cloud feedback, which is dominated by the low and midlevel clouds.

In an analysis using Tropical Rainfall Measuring Mission (TRMM; Simpson et al. 1996) data, Lau and Wu (2003) further examined the role of warm precipitating clouds in the tropics. This study exploited the fact that no SST can be retrieved from the TRMM Microwave Imager (TMI) when a grid box is completely filled with rain, as would be the case in deep convection. However, because of the nature of warm rain clouds, they do not typically fill a grid box, and a valid TMI SST can be retrieved. A valid SST in the presence of rain was classified as warm rain, and their results suggest that warm rain clouds are responsible for about 31% of the total rainfall in the tropics. An earlier study by Johnson et al. (1999) indicated that midlevel congestus clouds contribute greater than 25% of the total tropical convective rainfall. In light of the fact that congestus clouds include both warm rain clouds and those that reach the freezing level and glaciate, the Lau and Wu (2003) estimate is slightly higher than these earlier findings. It is also within the range of estimates by Petty (1999), who combined infrared satellite data with surface and ship stations and found that warm rain clouds were associated with 20%–40% of the precipitation reports at these stations. The most recent estimates from CloudSat (Stephens et al. 2002) show that in the western Pacific Ocean low and middle clouds make up about 50% of the population of precipitating clouds and over the entire tropics rain falls as frequently from low clouds as it does from both middle and deep convective clouds combined (Haynes 2008). Each of the previous studies used different definitions of warm rain clouds, which may be responsible for the spread in warm rain estimates. In this study, we define warm rain clouds as clouds that are precipitating and have tops below the freezing level as indicated by infrared brightness temperatures above 273 K.

The Lau and Wu (2003) study also went a step further to investigate the precipitation efficiency of warm rain clouds. Using a climate-model parameterization along with TMI-retrieved cloud liquid water and precipitation, they showed that residence time, defined as the ratio of cloud water to precipitation, decreased with increasing SST in warm rain clouds, especially for low rain rates. Their findings suggest that precipitation efficiency (the inverse of residence time) of warm rain clouds increases by approximately 8% per kelvin degree of increase in SST and that this increase in precipitation efficiency in warm rain may be at the expense of cloud water. In a previous study that examined the influence of SST on deep convective cloud area, Rapp et al. (2005) used the ratio of cloud area, as defined by TRMM Visible and Infrared Scanner (VIRS) infrared brightness temperatures, to rainfall rate from the TRMM precipitation radar (PR) as a proxy for the precipitation efficiency of rain clouds in the tropical western Pacific. The ratio of deep convective cloud area to rainfall was found to be insensitive to SST; however, warm rain clouds showed an approximately 5% decrease in the ratio of cloud area to rainfall per degree rise in SST. This is similar to the findings by Lau and Wu (2003) and provides more observational evidence supporting their suggestion that the increase in precipitation efficiency may be at the expense of cloud water.

Although the findings of these studies are suggestive, each is subject to criticisms. The SST screening method of Lau and Wu (2003) is an ambiguous way of identifying warm rain systems. The study by Rapp et al. (2005) was designed for the identification of deep convection and was not originally intended to examine warm rain systems. The criteria by which the clouds were defined—namely, that the clouds must have a rainfall rate greater than 10 mm h−1—limit the population of warm rain systems observed in that study. With these criticisms in mind, we set out to investigate further the properties of warm rain clouds in the tropics but found that available cloud microphysical property datasets disagreed. Both microwave and optical retrievals are readily available to examine the properties of clouds in the tropics; however, comparison of these products shows discrepancies when the clouds are raining. Moderate Resolution Imaging Spectroradiometer (MODIS) 5-km cloud liquid water path (LWP) retrievals (Platnick et al. 2003; King et al. 1997) from the version-5 Atmosphere Level 2 Joint Product from the Aqua satellite were matched to 0.25° Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) version-5 retrievals from Remote Sensing Systems, Inc., (RSS; Wentz and Meissner 2000) and compared for the tropical western Pacific. Although the 0.25° resolution of the AMSR-E retrieval is coarse, each of the microwave channels used in the retrieval has a different field of view (FOV) and therefore comparing at finer scales would be questionable. Using the MODIS cloud phase and cloud-top temperature along with the AMSR-E rainfall retrieval as a rain mask, we were able to identify warm rain clouds in the tropical western Pacific. We examine this region because it is where warm rain clouds showed the sensitivity to SST in Rapp et al. (2005). Figures 1a and 1b from MODIS and AMSR-E, respectively, show the sensitivity of LWP retrievals to SST for nonraining clouds. Nonraining clouds show similar results for both retrievals, with LWP fairly constant over the observed range of SST. However, when examining the two LWP retrievals for the raining, warm clouds in Figs. 1c and 1d, substantial differences are observed. The MODIS retrieval shows a strong decrease in LWP with increasing SST, whereas the microwave retrieval remains fairly constant. Although not shown here, examination of the MODIS retrievals of effective radius and cloud optical depth also reveals differences between nonraining and raining clouds. In nonraining clouds the effective radius increases from 12 to 18 μm and the optical depth decreases by about 30%, with the resultant LWP remaining nearly constant with SST. For the warm rain clouds, the effective radius has small variations over the range of SST shown in Fig. 1, only varying from about 18 to 21 μm. However, the cloud optical depth decreases by about 60% with SST, which is the reason for the large observed decrease in warm rain cloud LWP in Fig. 1. Results from a modified Nakajima–King retrieval (Nakajima and King 1990; Nakajima and Nakajima 1995) applied to the TRMM VIRS data and RSS TMI retrievals show similar behavior.

It is not surprising given the limitations of each of these methods that there are differences, but it illustrates the potential issues with using these retrievals for some climate studies. Zuidema et al. (2005) estimate up to 15%–20% differences in retrieved LWP depending on the gaseous absorption and liquid dielectric models used in microwave retrievals. The microwave retrieval also suffers from coarse resolution and may be viewing either clear sky or multiple cloud types within a single footprint containing a warm rain cloud. It is also possible that the coarse resolution affects warm rain cloud identification since these clouds may not fill the entire footprint and the signal may be too weak to retrieve rainfall. This could be why the MODIS cloud optical depth slightly decreases with SST for nonraining clouds, because clouds may be misidentified as nonraining. Another issue with the microwave retrieval is that the microwave retrieval is sensitive to both cloud drops and raindrops, and therefore most retrieval algorithms use an LWP threshold to delineate clouds and rainfall, which can bias the results as shown by Berg et al. (2006) and O’Dell et al. (2008). In the microwave LWP retrieval shown in Fig. 1, the LWP in raining clouds has a minimum threshold of 180 g m−2 and is parameterized as a function of the rain rate and an assumed cloud height based on SST. Because of the retrieved LWP dependence on the rain rate, the shape of this curve is very similar to the shape of the mean rain-rate curve for warm rain clouds. O’Dell et al. (2008) estimate a combined systematic error of 15%–30% in satellite microwave retrievals that is due to a combination of beam-filling effects, cloud temperature and height assumptions, sensitivities of LWP to retrieved total precipitable water (TPW) and wind speed, and cloud/rain partitioning.

The visible/near-infrared retrieval calculates LWP as the residual of the retrieval of optical depth from the nonabsorbing visible wavelengths and the retrieval of effective radius from the absorbing near-infrared wavelengths. Systematic errors in visible/near-infrared retrievals typically stem from uncertainties in the measurements and assumptions regarding the uniform profile of cloud particle size with altitude. A number of studies (e.g., Han et al. 1994; Nakajima and Nakajima 1995; Platnick and Valero 1995) have shown satellite retrievals of effective radius and LWP to agree with in situ measurements to within 10% and about 20%, respectively. Another issue that may be affecting Fig. 1 is that the retrieval of effective radius is only sensitive to the cloud top, and therefore the calculation of LWP could be biased, especially in a raining cloud in which the larger drops are more concentrated near the cloud base. It is also possible that in a thick, precipitating cloud, much of the visible radiation may be scattered by the cloud drops before it reaches the rain in the lower portion of the cloud.

Because we are most interested in how the rainfall is affecting the properties of the cloud, rather than the total column properties, we have developed a combined optimal estimation microwave retrieval algorithm that takes advantage of the strengths of these different datasets as well as uses precipitation information that is available on the TRMM satellite to retrieve a microwave cloud LWP in a warm rain cloud. We stress the word cloud, because this retrieval estimates the LWP that is due to only the cloud water in a raining cloud. Note that, in reality, rainwater and cloud water are not necessarily physically distinct quantities. The microwave radiometer is sensitive to emission from the total integrated water content of the atmosphere and only has a minor dependence on the drop size distribution (DSD) because the absorption efficiency in the Mie regime is slightly higher than in the Rayleigh regime. Radar reflectivity is proportional to drop sizes to the sixth power, which means that the larger, precipitation-sized droplets dominate the signal and the signal from cloud-sized particles is nearly negligible. In this study, the distinction of cloud and rainwater is based solely on the sensitivity of the PR. Using deconvolved TMI brightness temperatures along with cloud fraction information from VIRS and rainwater estimates from the PR, this retrieval improves on some of the resolution and sensitivity problems of microwave-only or visible/near-infrared retrievals and also takes into account the emission and scattering from the rainfall to allow an estimate of LWP associated with the cloud water in a raining cloud.

2. Data

The retrieval algorithm developed in this study combines data from microwave, visible/infrared, and precipitation radar sensors in an optimal estimation (OE) retrieval framework. We test the algorithm with data from the TRMM satellite observed from December 2005 to February 2006 at 30°S–30°N, 130°E–170°W. However, this algorithm could also be applied by combining a visible/infrared sensor with the upcoming Global Precipitation Measurement mission sensors. And though more difficult because the CloudSat radar does not scan the entire AMSR-E field of view, it may be possible to apply this algorithm to a combination of AMSR-E, MODIS, and CloudSat.

a. Microwave sensor

The brightness temperatures used in this retrieval are from the TMI (Kummerow et al. 1998). The TMI is a conically scanning passive microwave radiometer that has eight channels that measure both vertical (V) and horizontal (H) polarizations at frequencies of 10.7, 19.4, 37.0, and 85.5 GHz, and a ninth channel at 21.3 GHz that measures only the vertical polarization. The resolution of the measurements varies with the frequency and ranges from 7 km × 5 km at 85.5 GHz up to 63 km × 37 km at 10.7 GHz. The resolution difference can be a problem because of the inhomogeneity of the scenes viewed by the different channels. A deconvolution algorithm was applied to resample all of the TMI channels to a common resolution of the 19.4-GHz channel at 30 km × 18 km. For the algorithm developed in this study a uniform FOV is a necessity because we are using the VIRS data to calculate cloud fraction within the microwave footprint. This requires that the microwave data be at a common resolution. The deconvolution algorithm follows the method of Backus and Gilbert (1970) that has been shown to be successful at resolution modification of remotely sensed microwave data in numerous studies (e.g., Stogryn 1978; Robinson et al. 1992; Farrar and Smith 1992; Long and Daum 1998). This method uses a weighted sum of the observed brightness temperatures to construct a set of effective brightness temperatures at a single resolution. The Backus–Gilbert method calculates the weighting coefficients by choosing a factor that minimizes the error in the fit of the solution and minimizes the associated noise amplification. Calculating these weighting coefficients is time consuming, but, because the TMI antenna patterns and scan geometry are known, the coefficients only need to be calculated once and can then simply be applied to each orbit. Although it is possible to resample the data to the resolution of the highest-frequency channels, the associated noise becomes very large. Making the enhancement of the 10-GHz channel to the 19.4-GHz FOV increased the noise level from 0.54 to 1.5 K. Resampling to a higher resolution would result in an even larger increase in noise, above which we felt the accuracy of the retrievals would suffer. As a result, the 19.4-GHz-channel resolution was chosen because it was the best combination of enhanced resolution of the low-frequency channels with an acceptable level of noise.

b. Visible/infrared sensor

The retrieval we have developed utilizes data from the VIRS on TRMM to characterize the cloud field within a TMI footprint. The VIRS (Kummerow et al. 1998) is a five-channel (0.63, 1.6, 3.7, 10.8, and 12.0 μm) scanning radiometer on the TRMM satellite with a 2.11-km FOV at nadir and a swath width of 720 km. For this study, we use the visible and infrared channels as a cloud mask to determine the cloud fraction of a given TMI footprint. The VIRS pixels are collocated and matched to the TMI footprints, and the cloud fraction is calculated as the ratio of matching VIRS pixels identified as cloud to the total matching VIRS pixels. A VIRS pixel is considered to be cloudy if the visible reflectance at 0.63 μm is greater than the characteristic clear-sky reflectance or if the 10.8-μm infrared brightness temperature is lower than a threshold determined by the underlying SST. The pixel is further tested for ice contamination by checking that the infrared brightness temperature is not below 270 K and by checking the difference between the infrared channels to identify thin cirrus. The retrieval algorithm developed in this study is only designed for microwave footprints that contain water clouds, and therefore any TMI footprint that contains matched VIRS pixels identified as ice clouds is flagged and no retrieval is performed.

c. Precipitation radar

One of the most important additions to this algorithm is the information provided by the TRMM PR. The PR is a 13.8-GHz cross-track scanning phased-array radar with a 4.3-km horizontal resolution at nadir and 215-km FOV. We are using the 2A25 dataset (Iguchi et al. 2000) that contains PR estimates of total precipitation water content, which is the vertically integrated precipitation water content calculated from the attenuation-corrected radar reflectivity at each range bin from the surface to the rain top. This allows us to calculate the emission and scattering of the rainwater. In this way, we can separately calculate the contribution of the rainwater to the forward modeled brightness temperatures and retrieve only the LWP associated with the cloud water. As previously mentioned, almost all other microwave retrieval algorithms typically use an LWP threshold to distinguish rain, but this method has been shown to result in discrepancies. Using rainwater information from the PR removes the need for using a threshold and allows the more realistic treatment of cloud and rainwater that is necessary for this retrieval.

d. Ancillary data

In this study, we specify SST, temperature lapse rate, and water-vapor scale height for calculating the upwelling radiances in the forward model. The SST data used in this study are retrieved daily from the TMI (Wentz and Meissner 2000; Wentz et al. 2000) on a 0.25° × 0.25° grid. To describe the atmospheric temperature lapse rate and water-vapor scale heights, daily values are computed from National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) data on a 2.5° × 2.5° grid. The temperature lapse rate is computed as the average lapse rate from the surface to 250 hPa. To calculate the daily water-vapor scale height, reanalysis specific humidity profiles from the surface to 300 hPa are fit with
i1558-8432-48-11-2242-e1
where SH is the specific humidity at height Z, SH0 is the specific humidity at the surface, and H is the water-vapor scale height. Uncertainties due to these parameters must be specified in the forward model. For each 2.5° × 2.5° grid box, the standard deviation in lapse rate and scale height is computed for the 3-month period of study. For the period from January to February 2005, the region of study in the western Pacific has an average lapse rate of 6.4 K km−1, with an average standard deviation of 0.5 K km−1. The average water-vapor scale height for this time period and region is 2.6 km, with an average standard deviation of 0.6 km. The uncertainties in the forward-model brightness temperatures that are due to the NCEP–NCAR reanalysis lapse rates and water-vapor scale heights are computed from these average standard deviations.

3. Retrieval algorithm

Retrieval of atmospheric properties from microwave measurements is dependent on a number of factors—the forward model, assumptions about the model atmosphere, the uncertainties of the forward model itself, the assumptions in the forward model, and uncertainties in the measurements. In this study we employ the optimal estimation retrieval technique (Rodgers 1976, 2000; Marks and Rodgers 1993) for the inversion. An earlier version of our algorithm for nonprecipitating clouds is thoroughly described in Elsaesser and Kummerow (2008); however, we have made several modifications to the forward model that allow us to account for partially cloud filled TMI footprints, as well as emission and scattering from rainwater.

a. Forward model

Following Elsaesser and Kummerow (2008), the surface reflection and emission are calculated using the Deblonde and English (2001) model, which takes into account nonspecular reflection to improve surface emissivity calculations at large viewing angles, and the Kohn (1995) model, which improved on the Wilheit (1979a,b) model with better treatment of multiple reflections, sea surface roughness parameters, and sea foam. A modified version of the Rosenkranz (1998) model is used to compute gaseous absorption by oxygen, nitrogen, and water vapor. In this study, we are retrieving cloud systems that may only partially fill the microwave pixels and in which scattering may be present. Because the microwave pixels may contain both clear and cloudy areas, the calculation of brightness temperatures TB in the forward model is formulated as
i1558-8432-48-11-2242-e2
where α is the cloud fraction as defined by VIRS, TBCLR is the modeled clear-sky brightness temperature, and TBCLD is the modeled brightness temperature of the cloudy area. The use of cloud fraction is of great practical benefit to the retrieval; however, it is prescribed rather than retrieved because the TMI and PR measurements contain little information about the cloud fraction. However, the VIRS visible and infrared radiances provide an excellent high-resolution cloud mask. It is likely that there are slant-path effects that influence our calculation of cloud fraction; however, this is limited somewhat by the fact that we are only using data within the 215-km swath width of the PR. The possible errors associated with the cloud-fraction calculation are taken into account within the OE framework and are discussed in the following section. In the absence of rain the Rayleigh approximation is assumed and cloud liquid water absorption calculations are based on the Liebe et al. (1991, 1993) model. When rain is present and the Rayleigh assumption is no longer valid, the calculation of TBCLD includes Mie-scattering effects. Using the rainwater estimate and top of the rain column from the PR and assuming a Marshall–Palmer DSD (Marshall and Palmer 1948), we calculate the contribution of the rainwater to the upwelling brightness temperature according to Lorenz–Mie theory. These calculations include many assumptions regarding the DSD and the accuracy of the PR rainwater estimates, which must be accounted for within the retrieval. The effects of these assumptions are investigated in more detail in the following section.

b. Retrieval approach

Following the work of Rodgers (1976), we employ the optimal estimation approach to the retrieval as in Elsaesser and Kummerow (2008). With the forward model described previously being denoted as F, we can express the TMI satellite measurements y as
i1558-8432-48-11-2242-e3
where x is the retrieved atmospheric state [TPW, surface wind speed (“WIND”), and LWP], b represents other unretrieved surface and atmospheric parameters in the forward model, and ε is an error term containing the uncertainties in the measurements, forward model, and the forward-model assumptions. The problem is then to invert this equation by estimating the atmospheric state x that most likely produced the TMI measurements y. In this case, we are interested in retrieving the amount of cloud water regardless of whether the scene is precipitating or not; therefore our state vector is the same for all scenes and rainwater is prescribed for precipitating scenes from the PR estimates. Using the Bayes theorem, the probably of x being the true retrieved state given a set of TMI measurements is proportional to the product of the probability of observing the TMI measurements y given a simulated state x and the a priori probability that is the atmospheric state,
i1558-8432-48-11-2242-e4
The solution of retrieved state x occurs when a cost function Φ is minimized. The cost function is given by
i1558-8432-48-11-2242-e5
where 𝗦y represents the uncertainties associated with the measurements and the forward model, 𝗦a represents the a priori variability of the retrieved state x, and xa is the a priori guess at the atmospheric state x. Using Newtonian iteration, the value for x that minimizes the cost function can be found with
i1558-8432-48-11-2242-e6
where
i1558-8432-48-11-2242-e7
is the error covariance matrix of the retrieved parameters and 𝗞 is the kernel matrix expressing the sensitivity of the forward model to a perturbation in the retrieved parameters. The solution is found by iterating until the difference in retrieved states between successive iterations is less than the number of independent retrieved parameters. One of the assumptions of the optimal estimation retrieval is that the state vector parameters follow a Gaussian distribution with random error variances. The validity of the retrieved solution is dependent on how well the parameters are represented by a Gaussian distribution. Elsaesser and Kummerow (2008) found that the assumption of a lognormal probability distribution function (PDF) instead of a Gaussian PDF yielded up to a 10% change in the retrieved LWP but less than a 1% difference in retrieved wind speed and TPW. Therefore, in this retrieval, a lognormal transformation of LWP is performed as in Elsaesser and Kummerow (2008). This is done implicitly, with final results showing LWPs transformed out of lognormal space.

c. Retrieval error diagnostics

One of the benefits of the optimal estimation approach to inversion is that it provides several diagnostics that indicate the quality of the retrieval. In Eq. (7), the error covariance of the retrieved parameters provides an estimate of the uncertainty in the retrieved state due to uncertainties in the measurements, in the forward model, and in the a priori parameters. Diagonal elements of Eq. (7) represent the errors associated with each retrieved variable, and the off-diagonal elements represent the correlations between errors in the retrieved variables. From Eq. (7), it is obvious that the error of the retrieved parameters is dependent on both the uncertainties in the a priori parameters and the uncertainties in the measurements and the forward model. TMI measurement uncertainties are described in Kummerow et al. (1998), and a thorough description of the a priori and forward-model parameter uncertainties for nonraining scenes is given in Elsaesser and Kummerow (2008). Table 1 gives the uncertainties from Elsaesser and Kummerow (2008) for the SST and cloud height, and the values for the uncertainty in lapse rate and scale height have been recomputed from the daily NCEP–NCAR reanalysis values described previously in section 2. However, the microwave TB deconvolution, the addition of fractional cloudiness from the VIRS, and the PR rain characteristics introduce other error sources that must be accounted for.

The contribution to the forward model and measurement error covariance matrix 𝗦y due to the resampling of the microwave TB to a common resolution is computed at each channel and given as σBG in Table 2. Note that the value given in Table 2 for σBG is only for the pixel in the middle of the scan, since it varies with scan position.

Utilizing cloud fraction from the VIRS also introduces uncertainty in the forward-model calculations. Errors in the cloud mask due to thresholding techniques and slant-path effects influence the weighting of the computed clear and cloudy TB. Because developing a better cloud mask is not the goal of this study and there is little information on the uncertainty in the cloud mask, here we use a simple cloud-masking technique and assume up to a 30% error in our estimates of cloud fraction. Perturbing cloud fraction by 30%, we simulate the associated TB and calculate the difference from the unperturbed cloud-fraction TB to estimate the contribution of errors in assumed cloud fraction σFCLD that are shown with the other sources in Table 2.

In retrieving rainy-scene parameters, the largest source of error comes from uncertainties in the PR-estimated rainwater used as input in the forward model. To calculate the uncertainty in the rainwater estimates, we have used the PR attenuation-corrected reflectivity along with the PR-derived rainwater for warm rain clouds, shown in Fig. 2. This figure shows that, at any given reflectivity bin, there is a wide range of rainwater values depending on whether a convective or stratiform reflectivity–rainwater relationship is used in the 2A25 PR rainfall algorithm. The two different classifications are evident in Fig. 2 by the split between clusters of points at higher reflectivities; however, over 90% of the warm rain profiles are below about 30 dBZ where the convective and stratiform curves are close to one another. The lack of certainty in storm classification and the scatter within the classifications suggest a mean uncertainty in PR-estimated rainwater content of about 50%. This uncertainty in rainwater stems from assumptions made in the PR algorithm regarding the DSD of the rainwater. In our forward model, rainwater absorption and scattering are calculated as functions of temperature, DSD, and PR rainwater content. In these calculations we are assuming an exponential DSD:
i1558-8432-48-11-2242-e8
with a distribution intercept value N0 of 8 × 106 m−4 as given by the Marshall–Palmer distribution and prescribing the rainwater content and temperature. To maintain consistency between N0 and the PR rainwater content, we solve for the appropriate slope of the distribution λ. Because N0 is prescribed and may not represent the actual distribution or that assumed by the PR, errors in this assumed value translate into errors in the calculation of absorption and scattering coefficients and ultimately into errors in the simulated TB.

To test the forward-model sensitivity to errors in the rainwater and DSD, the PR rainwater and N0 values are perturbed by 50% and microwave TBs are simulated. The difference in brightness temperature is calculated from TB using the original PR estimate and from TB modeled with the Marshall–Palmer distribution N0 value. The total error due to PR rainwater and DSD assumptions is given as the sum of the square of the errors. This gives a reasonable estimate of the sensitivity of the forward-model TB computation to PR-rainwater uncertainty and DSD assumptions and is shown as σRW in Table 2.

One useful diagnostic that results from the optimal estimation technique used in this study to evaluate the quality of the retrieval is the χ2 test, in which
i1558-8432-48-11-2242-e9
In general terms, χ2 indicates how well the forward model TBs fit the observations. This should approximately follow a χ2 distribution with the number of degrees of freedom equal to the number of dimensions of the observations y, if the forward-modeled TBs agree with the observations within the error range. If χ2 is too small, then measurement errors may have been overestimated or the a priori is too loosely constrained. If χ2 is very large, then either the forward model inadequately represents the physics or the assumed errors and Gaussian error distribution do not correctly describe the uncertainties in the measurements and the forward model. In this study, χ2 values range from near 0 to values exceeding 100. For a problem with eight degrees of freedom (nine microwave frequencies), the theoretical χ2 critical value is approximately 20. A retrieved solution with a χ2 value below this threshold can be considered to be a good fit to the measurements at the 99% confidence level.

4. Results

To understand better the results from this retrieval algorithm, it is important to recognize the sensitivity of the retrieval to the addition of the parameters described in this paper. Results after the addition of the VIRS cloud fraction and PR rainwater are compared with the optimal estimation retrieval without this additional information to test the sensitivities. Elsaesser and Kummerow (2008) validated the nonraining retrieval against both optical and other microwave retrievals, but it is nearly impossible to validate LWP retrievals in the presence of rain, especially because we are only retrieving the cloud LWP and not the total. However, we will compare the results of our warm rain retrieval algorithm with the microwave and optical retrievals from Fig. 1.

a. Sensitivity to cloud fraction

We test the retrieval’s sensitivity to cloud fraction by examining the differences in the OE retrieval with and without the VIRS cloud-fraction information for the same three months of data in the tropical western Pacific. Figures 3a–d show the differences between the retrieved cloud properties (retrieval with cloud fraction minus retrieval without cloud fraction) along with the standard deviation for a given prescribed cloud fraction. Not surprising is that utilizing the VIRS cloud fraction tends to increase the retrieved cloud LWP especially at very low cloud fractions; however, there is little sensitivity in the TPW and wind speed except at cloud fractions below 20%. This lack of sensitivity to cloud fraction is most likely due to the more uniform nature of the TPW and wind fields in compared with the more variable LWP field. From middle to high cloud fractions, the use of the VIRS cloud-fraction information improves the fit of the retrieved properties, as shown by slightly lower χ2 values in Fig. 3d, but at low-to-midrange cloud fractions, for which even the TPW and wind field differences and standard deviations increase because of inclusion of cloud fraction, the resulting retrieved solution tends to be a poorer fit than without cloud fraction. Such low cloud fractions typically do not occur for TMI pixels containing rain, and therefore it will not affect the warm rain cloud retrievals in which we are most interested.

b. Sensitivity to rainwater

One of the most important additions to this OE retrieval is the inclusion of the PR rainwater estimates so that the cloud LWP can be retrieved in warm rain scenes. By modeling the contribution of the rainwater on the microwave TB, we retrieve the LWP associated with the cloud and examine the role that precipitation plays in the properties of clouds. Figures 4a–d show the retrieved properties and their standard deviations for warm raining scenes for an OE retrieval with and without the PR rainwater estimates. The inclusion of the rainwater inherently lowers the cloud LWP but only affects the TPW at high amounts of column water. At low wind speeds, the inclusion of rainwater increases the retrieved wind speed, whereas the opposite occurs for the higher wind speeds. Although not shown, the majority of the large differences in TPW and wind speeds occurred at the highest rain rates, with the LWP differences being fairly constant for high rain rates. The lower χ2 values indicate that the retrieved solution is a substantially better fit with the inclusion of the PR rainwater estimates. However, not reflected in these figures are pixels for which the retrieval cannot converge to a solution with the inclusion of the PR rainwater. Although not frequent, it typically occurs at the heaviest rain rates. Figures 5a and 5b depict a portion of a TRMM swath that contains many warm rain clouds, as shown by the cloud-top temperatures in the VIRS IR TB and the PR rain-rate map. Figures 5c and 5d show the resultant OE LWP retrieval without and with the PR rainwater information, respectively. In Fig. 5c, the rainfall is evident in the high retrieved LWP, which corresponds to areas of rain in Fig. 5b. By taking into account the rainwater, the retrieved LWP in Fig. 5d represents the water in the cloud and not the total water in the column and is much more uniform, as is expected in a cloud LWP field. The black areas of the retrieval within the clouds in Fig. 5d are not pixels with zero LWP but rather are those in which the retrieval could not converge to a solution. As previously mentioned, they correspond to higher-rain-rate PR pixels, a fact that most likely indicates that the PR rainwater estimate is too high or that our assumptions regarding the rain DSD are inappropriate for this situation.

c. Results for warm rain clouds

The results in this section represent a total of over 300 000 TMI pixels containing warm rain clouds with no ice contamination from December 2005 to February 2006 at 30°S–30°N, 130°W–170°E. Of the pixels with no ice contamination, precipitating clouds represent 10% of the total population of warm clouds. These results also show that 20% of the total rainfall is due to warm rain clouds. Both of these estimates are probably biased slightly negatively because of our strict definition of warm rain clouds, as well as the fact that we are excluding TMI pixels that may contain ice. Over the range of SSTs observed in this region (∼290–304 K), nearly 75% of the warm rain clouds occurred in a 4-K range of SST from 298 to 302 K. The results from this retrieval are very different for warm rain clouds when compared with the AMSR-E microwave retrieval in Fig. 1. This is expected since the AMSR-E retrieval is more indicative of the total LWP and includes the contribution from the rainwater. Figures 6a and 6b show the nonraining and warm rain cloud LWP retrievals, respectively, plotted against the underlying SST. Similar to Fig. 1, the nonraining cloud LWP in Fig. 6a shows very little sensitivity to SST, except for a small spike around 293 K. The nonraining OE retrieval does show a slightly lower mean LWP, around 75 g m−2, with SST than either the RSS AMSR-E (∼100 g m−2) or the MODIS retrieval (∼90 g m−2). These differences in the mean LWP with SST are most likely due to the different rain masks used to detect precipitation. Because we are using the higher-resolution PR data as a rain mask, our retrieval has a higher sensitivity to rainfall and results in a lower mean nonraining LWP than the retrievals using AMSR-E as a rain mask. Without additional information, it is not possible to assess which retrieval might be more accurate. We can nonetheless conclude that, irrespective of the method, the trends show that mean LWP of nonraining clouds is fairly insensitive to SST.

Of interest is that the warm rain OE retrieval in Fig. 6b shows a very strong decrease in cloud LWP with increasing SST. This is in good agreement with the MODIS warm rain LWP results in Fig. 1. In a thick cloud that contains rain, it is very likely that the majority of the visible radiation has been scattered by the time it reaches the top of the rain column, and therefore the MODIS retrieval is mostly sensitive to the water in the cloud and not in the rain. This may be why the two retrievals operating on entirely different principles—the visible/infrared, which inadvertently does not see the rain, and the microwave OE retrieval, which directly accounts for the rain—yield comparable results. Although these results are suggestive and indicate that cloud water may be more efficiently converted to rainfall at higher SST, we can examine the relationship between the rainfall and the cloud properties from our retrievals and determine how their interaction may be affected by surface temperature. Because the warm rain cloud LWP is decreasing with SST in Fig. 6b, which is counterintuitive, it suggests that in an increased SST environment the convection may become more vigorous and the conversion of the cloud water to precipitation may be enhanced, resulting in the scavenging of cloud water for the production of more rainfall. To test this, the ratio of the cloud LWP to PR rainwater is compared in Fig. 7 to gain a better understanding of how rainfall may be affecting the cloud with changing SST.

The ratio of cloud LWP to PR rainwater in Fig. 7 also shows a strong decrease with SST. At the lowest SSTs, the ratio of LWP to rainwater is almost 2:1 but drops to almost 0.5:1 at the higher SSTs. There also appears to be a shift in the slope of the decrease in this ratio near 297 K, with the ratio decreasing faster at lower SSTs than at higher SSTs. Because we are examining a large region that includes both deep tropics as well as subtropical latitudes, this shift may indicate a change in rain regimes that is not apparent when just examining the relationship of LWP with SST. In the deep tropics, there are very few SSTs that fall below 297 K and examining the locations of the clouds shows that over 99% of the warm rain clouds with SSTs below 297 K occur between 20°–30°S and 20°–30°N. Above 297 K, warm rain clouds occur both in the subtropical and tropical regions of our domain. Though not shown here, examining the LWP and rainwater ratio for warm rain clouds only in the subtropical area of the domain shows that the strong decrease of about 10% per degree of SST in Fig. 7 for clouds with SSTs below 297 K actually extends throughout the full SST range. In the tropical area of our domain, between 20°S and 20°N, where SSTs are above 297 K, the decrease is only about 6% per degree of SST. Because clouds in the tropical region of the domain are more numerous, they dominate the trend seen in Fig. 7 above 297 K. The dynamics influencing the clouds in the subtropical and tropical regions of our domain are very different, but over the range of SSTs observed in the two different regions both show that the rainwater is increasing at the expense of water in the cloud.

One issue that has not been addressed is the use of PR rain estimates as the rain/no rain threshold. The PR is only sensitive to rain rates that are greater than approximately 0.5 mm h−1, and therefore there is light rainfall from warm clouds that is not being detected. To be sure that this threshold is not biasing the nonraining cloud results, we use a technique suggested by Rosenfeld and Gutman (1994) to identify clouds that may be precipitating below the sensitivity threshold of the TRMM PR. Using the VIRS effective radius retrieved from a modified Nakajima and King scheme, we employ an effective radius threshold of 14 μm to identify possible raining clouds that may have been included in the previous results and recalculate the mean LWP with SST for the nonraining clouds. Though not shown here, the overall results are not affected by the exclusion of possible precipitating clouds, with both the mean and trends remaining virtually unchanged. Although inclusion of these clouds with the warm rain cloud results might be a better representation for all raining clouds, it is impossible with the available data to estimate the amount of rainwater in these clouds, which could be anywhere from just above zero up to just below the sensitivity threshold of the PR. However, if the results in Fig. 1 of Lau and Wu (2003) are valid, suggesting that the ratio of cloud water to rain rate decreases with SST much more rapidly for low rain rates, then this would only serve to enhance the decreasing trends observed in Figs. 6 and 7.

5. Discussion

To understand better the interaction between cloud properties and precipitation, an optimal estimation retrieval algorithm was developed for oceanic warm rain clouds that combines information from multiple sensors. By utilizing cloud fraction from the VIRS and rainwater estimates from the PR, we not only have eliminated some of the issues with microwave cloud property retrievals but also have enabled the retrieval of the water associated with the cloud instead of the total LWP. An examination of the difference in retrieved parameters due to the addition of VIRS cloud fraction showed very little difference in TPW and wind speed for all but the lowest cloud fractions, with increased LWP for all cloud fractions. A comparison of the OE retrieval with and without the PR rainwater estimates showed that the cloud LWP was always reduced in the presence of rain but the wind speed could be affected in either direction and that the TPW retrieval was lower for higher column water amounts, with both of these differences occurring for the highest rain rates. Evaluation of the results also shows that the added information generally lowered the χ2 diagnostic, suggesting that the extra information allows for a solution that better fits the measurements.

Examination of the warm rain cloud results shows that the LWP in the cloud, when the rainwater contribution is taken into account, decreases with underlying SST. Both the magnitude and trend in LWP agree well with the MODIS LWP retrieval in the presence of rain but not the AMSR-E retrieval, since it is sensitive to the total liquid water in the cloud and has the previously mentioned rain/LWP thresholding problem. The ratio of cloud LWP to PR rainwater also decreases with SST. This result is telling because it shows that the increase in rainwater at the expense of the water in the cloud is enhanced as the surface temperature increases, supporting our conjecture and previous findings. These results suggest that in an increased surface temperature scenario the rate at which cloud is converted to precipitation will increase, leaving less cloud water to moisten the lower and middle troposphere. Because cloud-resolving models typically use an autoconversion threshold that is a function of cloud water content and cloud drop number concentration (e.g., Khairoutdinov and Kogan 2000), for a given number concentration the availability of more water at higher SSTs will lead to higher cloud water contents and the rate of autoconversion should be increased. Whether climate models can reproduce this relationship is questionable. In a follow-up to their 2003 study, Lau et al. (2005) examined the sensitivity of a GCM to microphysical processes that describe the conversion of cloud water to precipitation. They found that increasing the autoconversion rate did indeed lead to more rainfall but less cloud produced by the model. However, this study manually prescribed increases in the autoconversion rates and did not have an interactive autoconversion rate. Del Genio et al. (2005) discussed the deficiencies in many GCM schemes to represent accurately the cloud and precipitation processes and showed that an interactive cumulus scheme can produce results similar to this study for clouds below the freezing level.

The implications of this increase in precipitation at the expense of cloud water have further impacts than just changes to the cloud and radiative properties. In a modeling study, Raymond and Torres (1998) showed that shallow and midlevel convective precipitation efficiency controlled the low- and midlevel moistening and found that the moistening provided by these clouds is necessary to precondition the environment for deep convection. Johnson et al. (2001) supported this assertion with observations from the Tropical Ocean and Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) that evaporation of cumulus congestus convection in the lower and middle troposphere preconditions the environment for deep convection. In a study of the mutual regulation of the tropical hydrologic cycle and sea surface temperature, Stephens et al. (2004) also present the humidistat feedback with evidence from TOGA COARE and TRMM suggesting a “destabilization phase” wherein shallow convection increases and moistens the lower troposphere to condition the atmosphere for deep convection. This moistening of the lower troposphere has also been linked to the time scales for deep convective outbreaks by Bladé and Hartmann (1993), Hu and Randall (1994), and Kemball-Cooke and Weare (2001) and more recently in the previously mentioned Lau et al. (2005) modeling study. The increase in the rate of conversion of cloud water to precipitation as shown by the results in this study leaves less water available to moisten the atmosphere, which many studies suggest may have implications for the onset of deep convection.

Acknowledgments

This research was supported by NASA Earth System Science Fellowship NNG05GP53H and NASA Grant NNX07AD75G.

REFERENCES

  • Backus, G., , and F. Gilbert, 1970: Uniqueness in the inversion of inaccurate gross earth data. Philos. Trans. Roy. Soc. London, A266 , 123192.

    • Search Google Scholar
    • Export Citation
  • Berg, W., , T. L’Ecuyer, , and C. Kummerow, 2006: Rainfall climate regimes: The relationship of regional TRMM rainfall biases to the environment. J. Appl. Meteor. Climatol., 45 , 434454.

    • Search Google Scholar
    • Export Citation
  • Bladé, I., , and D. L. Hartmann, 1993: Tropical intraseasonal oscillations in a simple nonlinear model. J. Atmos. Sci., 50 , 29222939.

  • Deblonde, G., , and S. J. English, 2001: Evaluation of the FASTEM-2 fast microwave ocean surface emissivity model. Tech. Proc. 11th Int. TOVS Study Conf., Budapest, Hungary, 67–78.

    • Search Google Scholar
    • Export Citation
  • Del Genio, A. D., , W. Kovari, , M. S. Yao, , and J. Jonas, 2005: Cumulus microphysics and climate sensitivity. J. Climate, 18 , 23762387.

  • Elsaesser, G. S., , and C. D. Kummerow, 2008: Toward a fully parametric retrieval of the nonraining parameters over the global oceans. J. Appl. Meteor. Climatol., 47 , 15991618.

    • Search Google Scholar
    • Export Citation
  • Farrar, M. R., , and E. A. Smith, 1992: Spatial-resolution enhancement of terrestrial features using deconvolved SSM/I microwave brightness temperatures. IEEE Trans. Geosci. Remote Sens., 30 , 349355.

    • Search Google Scholar
    • Export Citation
  • Han, Q., , W. B. Rossow, , and A. A. Lacis, 1994: Near-global survey of effective droplet radii in liquid water clouds using ISCCP data. J. Climate, 7 , 465497.

    • Search Google Scholar
    • Export Citation
  • Hartmann, D. L., , and K. Larson, 2002: An important constraint on tropical cloud–climate feedback. Geophys. Res. Lett., 29 , 1951. doi:10.1029/2002GL015835.

    • Search Google Scholar
    • Export Citation
  • Haynes, J., 2008: The near-global distribution of light rain from CloudSat. Ph.D. dissertation, Dept. of Atmospheric Science, Colorado State University, 229 pp.

  • Hu, Q., , and D. A. Randall, 1994: Low-frequency oscillations in radiative–convective systems. J. Atmos. Sci., 51 , 10891099.

  • Iguchi, T., , T. Kozu, , R. Meneghini, , J. Awaka, , and K. Okamoto, 2000: Rain-profiling algorithm for the TRMM precipitation radar. J. Appl. Meteor., 39 , 20382052.

    • Search Google Scholar
    • Export Citation
  • Johnson, R. H., , T. M. Rickenbach, , S. A. Rutledge, , P. E. Ciesielski, , and W. H. Schubert, 1999: Trimodal characteristics of tropical convection. J. Climate, 12 , 23972418.

    • Search Google Scholar
    • Export Citation
  • Johnson, R. H., , P. E. Ciesielski, , and J. A. Cotturone, 2001: Multiscale variability of the atmospheric mixed layer over the western Pacific warm pool. J. Atmos. Sci., 58 , 27292750.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kemball-Cooke, S. R., , and B. C. Weare, 2001: The onset of convection in the Madden–Julian oscillation. J. Climate, 14 , 780793.

  • Khairoutdinov, M., , and Y. Kogan, 2000: A new cloud physics parameterization in a large-eddy simulation model of marine stratocumulus. Mon. Wea. Rev., 128 , 229243.

    • Search Google Scholar
    • Export Citation
  • King, M. D., , S-C. Tsay, , S. E. Platnick, , M. Wang, , and K-N. Liou, 1997: Cloud retrieval algorithms for MODIS: Optical thickness, effective particle radius, and thermodynamic phase. MODIS Algorithm Theoretical Basis Doc. ATBD-MOD-05, MOD06-Cloud product, 83 pp.

    • Search Google Scholar
    • Export Citation
  • Kohn, D. J., 1995: Refinement of a semi-empirical model for the microwave emissivity of the sea surface as a function of wind speed. M.S. thesis, Dept. of Meteorology, Texas A&M University, 44 pp.

  • Kummerow, C., , W. Barnes, , T. Kozu, , J. Shiue, , and J. Simpson, 1998: The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Oceanic Technol., 15 , 808816.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and H. T. Wu, 2003: Warm rain processes over tropical oceans and climate implications. Geophys. Res. Lett., 30 , 2290. doi:10.1029/2003GL018567.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , H. T. Wu, , Y. C. Sud, , and G. K. Walker, 2005: Effects of cloud microphysics on tropical atmospheric hydrologic processes and intraseasonal variability. J. Climate, 18 , 47314751.

    • Search Google Scholar
    • Export Citation
  • Liebe, H. J., , G. A. Hufford, , and T. Manabe, 1991: A model for the complex permittivity of water at frequencies below 1 THz. Int. J. Infrared Millimeter Waves, 12 , 659675.

    • Search Google Scholar
    • Export Citation
  • Liebe, H. J., , G. A. Hufford, , and M. G. Cotton, 1993: Propagation modeling of moist air and suspended water particles at frequencies below 1000 GHz. Proc. Conf. on Atmospheric Propagation Effects through Natural and Man-Made Obscurants for Visible to MM-Wave Radiation, AGARD-CP-542, Neuilly sur Seine, France, Advisory Group for Aerospace Research and Development, 3-1–3-10.

    • Search Google Scholar
    • Export Citation
  • Lindzen, R. S., , M-D. Chou, , and A. Hou, 2001: Does the earth have an adaptive infrared iris? Bull. Amer. Meteor. Soc., 82 , 417432.

  • Long, D. G., , and D. L. Daum, 1998: Spatial resolution enhancement of SSM/I data. IEEE Trans. Geosci. Remote Sens., 36 , 407417.

  • Marks, C. J., , and C. D. Rodgers, 1993: Determination of characteristic features of cloud liquid water from satellite microwave measurements. J. Geophys. Res., 98 , 50695092.

    • Search Google Scholar
    • Export Citation
  • Marshall, J. S., , and W. M. Palmer, 1948: The distribution of raindrops with size. J. Meteor., 5 , 165166.

  • Nakajima, T., , and M. D. King, 1990: Determinations of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I: Theory. J. Atmos. Sci., 47 , 18781893.

    • Search Google Scholar
    • Export Citation
  • Nakajima, T. Y., , and T. Nakajima, 1995: Wide-area determination of cloud microphysical properties from NOAA AVHRR measurements for FIRE and ASTEX regions. J. Atmos. Sci., 52 , 40434059.

    • Search Google Scholar
    • Export Citation
  • O’Dell, C. W., , F. J. Wentz, , and R. Bennartz, 2008: Cloud liquid water path from satellite-based passive microwave observations: A new climatology over the global oceans. J. Climate, 21 , 17211739.

    • Search Google Scholar
    • Export Citation
  • Petty, G. W., 1999: Prevalence of precipitation from warm-topped clouds over eastern Asia and the western Pacific. J. Climate, 12 , 220229.

    • Search Google Scholar
    • Export Citation
  • Platnick, S., , and F. P. Valero, 1995: A validation of a satellite cloud retrieval during ASTEX. J. Atmos. Sci., 52 , 29853001.

  • Platnick, S., , M. D. King, , S. A. Ackerman, , W. P. Menzel, , B. A. Baum, , J. C. Riédi, , and R. A. Frey, 2003: The MODIS cloud products: Algorithms and examples from Terra. IEEE Trans. Geosci. Remote Sens., 41 , 459473.

    • Search Google Scholar
    • Export Citation
  • Ramanathan, V., , and W. Collins, 1991: Thermodynamic regulation of ocean warming by cirrus clouds deduced from observations of the 1987 El Niño. Nature, 351 , 2732.

    • Search Google Scholar
    • Export Citation
  • Rapp, A. D., , C. D. Kummerow, , W. Berg, , and B. Griffith, 2005: An evaluation of the proposed mechanism of the adaptive infrared iris hypothesis using TRMM VIRS and PR measurements. J. Climate, 18 , 41854194.

    • Search Google Scholar
    • Export Citation
  • Raymond, D. J., , and D. J. Torres, 1998: Fundamental moist modes of the equatorial troposphere. J. Atmos. Sci., 55 , 17711790.

  • Robinson, W., , C. Kummerow, , and W. S. Olson, 1992: A technique for matching the resolution of microwave measurements from the SSM/I instrument. IEEE Trans. Geosci. Remote Sens., 30 , 419429.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 1976: Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Rev. Geophys. Space Phys., 4 , 609624.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 2000: Inverse Methods For Atmospheric Sounding: Theory and Practice. World Scientific, 238 pp.

  • Rosenfeld, D., , and G. Gutman, 1994: Retrieving microphysical properties near the tops of potential rain clouds by multispectral analysis of AVHRR data. Atmos. Res., 34 , 259283.

    • Search Google Scholar
    • Export Citation
  • Rosenkranz, P. W., 1998: Water vapor microwave continuum absorption: A comparison of measurements and models. Radio Sci., 33 , 919928.

    • Search Google Scholar
    • Export Citation
  • Simpson, J., , C. Kummerow, , W. K. Tao, , and R. F. Adler, 1996: On the Tropical Rainfall Measuring Mission (TRMM). Meteor. Atmos. Phys., 60 , 1936.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2002: The CloudSat mission and the A-Train: A new dimension of space-based observations of clouds and precipitation. Bull. Amer. Meteor. Soc., 83 , 17711790.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., , P. J. Webster, , R. H. Johnson, , R. Engelen, , and T. L’Ecuyer, 2004: Observational evidence for the mutual regulation of the tropical hydrological cycle and tropical sea surface temperature. J. Climate, 17 , 22132224.

    • Search Google Scholar
    • Export Citation
  • Stogryn, A., 1978: Estimates of brightness temperatures from scanning radiometer data. IEEE Trans. Antennas Propagat., AP-26 , 720726.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , and T. Meissner, 2000: AMSR Ocean Algorithm: ATBD version 2. RSS Tech. Proposal 121599A-1, 66 pp. [Available online at http://www.remss.com/papers/amsr/AMSR_Ocean_Algorithm_Version_2.pdf].

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , C. Gentemann, , D. Smith, , and D. Chelton, 2000: Satellite measurements of sea surface temperatures through clouds. Science, 288 , 847850.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. T., 1979a: The effect of wind on the microwave emission from the ocean’s surface at 37 GHz. J. Geophys. Res., 84 , 49214926.

    • Search Google Scholar
    • Export Citation
  • Wilheit, T. T., 1979b: A model for the microwave emissivity of the ocean’s surface as a function of wind speed. IEEE Trans. Geosci. Electron., 17 , 244249.

    • Search Google Scholar
    • Export Citation
  • Zuidema, P., , E. R. Westwater, , C. Fairall, , and D. Hazen, 2005: Ship-based liquid water path estimates in marine stratocumulus. J. Geophys. Res., 110 , D20206. doi:10.1029/2005JD005833.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Mean LWP (g m−2) with SST (K) for nonraining warm clouds for (a) MODIS and (b) AMSR-E and for raining warm clouds for (c) MODIS and (d) AMSR-E.

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 2.
Fig. 2.

PR attenuation-corrected radar reflectivity (dBZ) vs rainwater content (kg m−3).

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 3.
Fig. 3.

Difference in parameters (a) LWP (g m−2), (b) TPW (mm), (c) wind speed (m s−1), and (d) χ2, retrieved with and without cloud fraction, plotted against VIRS cloud fraction.

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 4.
Fig. 4.

TMI OE-retrieved parameters (a) LWP (g m−2), (b) TPW (mm), (c) wind speed (m s−1), and (d) χ2, with and without PR rainwater estimates.

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 5.
Fig. 5.

Portion of TRMM swath containing warm rain clouds: (a) VIRS infrared TB (K), (b) PR rain rate (mm h−1), (c) optimal estimation LWP (g m−2) without PR rainwater estimation, and (d) optimal estimation LWP (g m−2) using PR rainwater estimate.

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 6.
Fig. 6.

Mean TMI OE-retrieved LWP (g m−2) vs SST (K) for (a) nonraining and (b) raining warm clouds.

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Fig. 7.
Fig. 7.

Ratio of mean TMI OE-retrieved LWP to PR-estimated rainwater vs SST (K).

Citation: Journal of Applied Meteorology and Climatology 48, 11; 10.1175/2009JAMC2156.1

Table 1.

Forward-model error sources (K) for each TMI channel resulting from assumptions in SST (σSST), water-vapor scale height (σSCLHT), temperature lapse rate (σLR), and cloud height (σCLDHT).

Table 1.
Table 2.

Forward-model error sources (K) for each TMI channel resulting from brightness temperature deconvolution (σBG), assumed cloud fraction errors (σFCLD), and errors in PR rainwater estimates and assumed DSD (σRW).

Table 2.
Save