• Adler, R. F., , Huffman G. J. , , and Keehn P. R. , 1994: Global tropical rain estimates from microwave-adjusted geosynchronous IR data. Remote Sens. Rev., 11, 125152, doi:10.1080/02757259409532262.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., and et al. , 2003: The Version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 11471167, doi:10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ashcroft, P., , and Wentz F. J. , 2006: AMSR-E/Aqua L2A global swath spatially-resampled brightness temperatures, version 2. NASA National Snow and Ice Data Center DAAC. Subset used: Low resolution swath data, accessed 5 October 2011. [Available online at http://nsidc.org/data/docs/daac/ae_l2a_tbs/v2/ae_l2a_tbs.gd.html.]

  • Cecil, D. J., 2009: Passive microwave brightness temperatures as proxies for hailstorms. J. Appl. Meteor. Climatol., 48, 12811286, doi:10.1175/2009JAMC2125.1.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , Grody N. C. , , and Marks G. F. , 1994a: Effects of surface conditions on rain identification using the SSM/I. Remote Sens. Rev., 11, 195209, doi:10.1080/02757259409532265.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 1994b: Microwave measurements produce global climatic, hydrologic data. Eos, Trans. Amer. Geophys. Union, 75, 337343, doi:10.1029/94EO00988.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , Smith E. A. , , Berg W. , , and Huffman G. J. , 1998: A screening methodology for passive microwave precipitation retrieval algorithms. J. Atmos. Sci., 55, 15831600, doi:10.1175/1520-0469(1998)055<1583:ASMFPM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 2013: An evaluation of microwave land surface emissivities over the continental United States to benefit GPM-era precipitation algorithms. IEEE Trans. Geosci. Remote Sens., 51, 378398, doi:10.1109/TGRS.2012.2199121.

    • Search Google Scholar
    • Export Citation
  • Gopalan, K., , Wang N.-Y. , , Ferraro R. R. , , and Liu C. , 2010: Status of the TRMM 2A12 land precipitation algorithm. J. Atmos. Oceanic Technol., 27, 13431354, doi:10.1175/2010JTECHA1454.1.

    • Search Google Scholar
    • Export Citation
  • Grody, N. C., 1991: Classification of snow cover and precipitation using the Special Sensor Microwave Imager. J. Geophys. Res., 96, 74237435, doi:10.1029/91JD00045.

    • Search Google Scholar
    • Export Citation
  • Grody, N. C., , and Basist A. , 1996: Global identification of snow cover using SSM/I measurements. IEEE Trans. Geosci. Remote Sens., 34, 237249, doi:10.1109/36.481908.

    • Search Google Scholar
    • Export Citation
  • Hong, Y., , Adler R. , , and Huffman G. , 2007: Use of satellite remote sensing data in the mapping of global landslide susceptibility. J. Nat. Hazards, 43, 245256, doi:10.1007/s11069-006-9104-z.

    • Search Google Scholar
    • Export Citation
  • Houghton, J. T., 1979: The future role of observations from meteorological satellites. Quart. J. Roy. Meteor. Soc., 105, 123, doi:10.1002/qj.49710544302.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and et al. , 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 3855, doi:10.1175/JHM560.1.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., , Adler R. F. , , Bolvin D. T. , , and Gu G. , 2009: Improving the global precipitation record: GPCP Version 2.1. Geophys. Res. Lett., 36, L17808, doi:10.1029/2009GL040000.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J., 2007: Validation of rainfall algorithms at the NOAA Climate Prediction Center. Measuring Precipitation from Space: EURAINSAT and the Future, V. Levizzani, P. Bauer, and F. Joseph, Eds., Advances in Global Change Research, Vol. 28, Springer Netherlands, 393–401.

  • Kirschbaum, D. B., , Adler R. , , Hong Y. , , and Lerner-Lam A. L. , 2009: Evaluation of a satellite-based landslide algorithm using global landslide inventories. Nat. Hazards Earth Syst. Sci., 9, 673686, doi:10.5194/nhess-9-673-2009.

    • Search Google Scholar
    • Export Citation
  • Kirstetter, P.-E., and et al. , 2012: Toward a framework for systematic error modeling of spaceborne precipitation radar with NOAA/NSSL ground radar–based National Mosaic QPE. J. Hydrometeor., 13, 12851300, doi:10.1175/JHM-D-11-0139.1.

    • Search Google Scholar
    • Export Citation
  • Kirstetter, P.-E., , Hong Y. , , Gourley J. J. , , Schwaller M. , , Petersen W. , , and Zhang J. , 2013: Comparison of TRMM 2A25 products, version 6 and version 7, with NOAA/NSSL ground radar–based National Mosaic QPE. J. Hydrometeor., 14, 661669, doi:10.1175/JHM-D-12-030.1.

    • Search Google Scholar
    • Export Citation
  • Kucera, P. A., , Ebert E. E. , , Turk F. J. , , Levizzani V. , , Kirschbaum D. , , Tapiador F. J. , , Loew A. , , and Borsche M. , 2013: Precipitation from space: Advancing Earth system science. Bull. Amer. Meteor. Soc., 94, 365375, doi:10.1175/BAMS-D-11-00171.1.

    • Search Google Scholar
    • Export Citation
  • Kummerow, C. D., and et al. , 2001: The evolution of the Goddard Profiling Algorithm (GPROF) for rainfall estimation from passive microwave sensors. J. Appl. Meteor., 40, 18011820, doi:10.1175/1520-0450(2001)040<1801:TEOTGP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kummerow, C. D., , Ringerud S. , , Crook J. , , Randel D. , , and Berg W. , 2011: An observationally generated a priori database for microwave rainfall retrievals. J. Atmos. Oceanic Technol., 28, 113130, doi:10.1175/2010JTECHA1468.1.

    • Search Google Scholar
    • Export Citation
  • Morrissey, M. L., , Maliekal J. A. , , Greene J. S. , , and Wang J. , 1995: The uncertainty of simple spatial averages using rain gauge networks. Water Resour. Res., 31, 20112017, doi:10.1029/95WR01232.

    • Search Google Scholar
    • Export Citation
  • Munchak, S. J., , and Skofronick-Jackson G. , 2013: Evaluation of precipitation detection over various surfaces from passive microwave imagers and sounders. Atmos. Res., 131, 8194, doi:10.1016/j.atmosres.2012.10.011.

    • Search Google Scholar
    • Export Citation
  • National Ice Center, 2008: IMS daily Northern Hemisphere snow and ice analysis at 1 km, 4 km, and 24 km resolutions. National Snow and Ice Data Center. Subset used: 24 km, accessed 27 February 2012, doi:10.7265/N52R3PMC.

  • Petty, G. W., , and Li K. , 2013: Improved passive microwave retrievals of rain rate over land and ocean. Part I: Algorithm description. J. Atmos. Oceanic Technol., 30, 24932508, doi:10.1175/JTECH-D-12-00144.1.

    • Search Google Scholar
    • Export Citation
  • Ramsay, B. H., 1998: The interactive multisensor snow and ice mapping system. Hydrol. Processes, 12, 15371546, doi:10.1002/(SICI)1099-1085(199808/09)12:10/11<1537::AID-HYP679>3.0.CO;2-A.

    • Search Google Scholar
    • Export Citation
  • Sapiano, M., , Janowiak J. E. , , Shi W. , , Higgins R. W. , , and Silva V. B. S. , 2010: Regional evaluation through independent precipitation measurements: USA. Satellite Rainfall Applications for Surface Hydrology, M. Gebremichael and F. Hossain, Eds., Springer Science, 169–191.

  • Schneider, U., , Becker A. , , Finger P. , , Meyer-Christoffer A. , , Rudolf B. , , and Ziese M. , 2011: GPCC full data reanalysis version 6.0 (at 0.5°, 1.0°, 2.5°): Monthly land-surface precipitation from rain-gauges built on GTS-based and historic data. Global Precipitation Climatology Centre. Subset used: 2.5° grid resolution, accessed 12 November 2013, doi:10.5676/DWD_GPCC/FD_M_V6_250.

  • Sudradjat, A., , Wang N.-Y. , , Gopalan K. , , and Ferraro R. R. , 2011: Prototyping a generic, unified land surface classification and screening methodology for GPM-era microwave land precipitation retrieval algorithms. J. Appl. Meteor. Climatol., 50, 12001211, doi:10.1175/2010JAMC2572.1.

    • Search Google Scholar
    • Export Citation
  • Tang, L., , Tian Y. , , and Lin X. , 2014: Validation of precipitation retrievals over land from satellite-based passive microwave sensors. J. Geophys. Res. Atmos., 119, 45464567, doi:10.1002/2013JA018955.

    • Search Google Scholar
    • Export Citation
  • Tedesco, M., , Kelly R. , , Foster J. L. , , and Chang A. T. C. , 2004: AMSR-E/Aqua daily L3 global snow water equivalent EASE-Grids, version 2. NASA National Snow and Ice Data Center DAAC. Subset used Monthly SWE, accessed 23 April 2012. [Available online at http://nsidc.org/data/docs/daac/ae_swe_ease-grids.gd.html.]

  • Wilber, A. C., , Kratz D. P. , , and Gupta S. K. , 1999: Surface emissivity maps for use in satellite retrievals of longwave radiation. NASA Tech. Publ. NASA/TP-1999-209362, 35 pp. [Available online at https://eosweb.larc.nasa.gov/sites/default/files/project/calipso/Wilber.NASATchNote99.pdf.]

  • Wilheit, T. T., , Theon J. S. , , Shenk W. E. , , Allison L. J. , , and Rodgers E. B. , 1976: Meteorological interpretations of the images from the Nimbus 5 electrically scanned microwave radiometer. J. Appl. Meteor., 15, 166172, doi:10.1175/1520-0450(1976)015<0166:MIOTIF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhang, J., and et al. , 2011: National Mosaic and Multi-Sensor QPE (NMQ) system: Description, results, and future plans. Bull. Amer. Meteor. Soc., 92, 13211338, doi:10.1175/2011BAMS-D-11-00047.1.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Screening procedure logic for GPROF2010. Green (red) arrows indicate that the conditions for flagging were met (not met). Blue shading indicates screens designed to identify frozen ground, surface ice, or fallen snow. In GPROF2010, all screening procedures were applied globally, including regions where snow and ice were never observed.

  • View in gallery

    Frequency at which the (left) snow and (right) semiarid land flags were triggered in 2010, corresponding to cases B and G in Fig. 1, respectively.

  • View in gallery

    Frequency that the TB snow flag was triggered in April 2010. Red line indicates the 0% contour of the snow frequency climatology.

  • View in gallery

    (left) NMQ/Q2 rain rates for a line of convective storms from 1920 UTC 30 Apr 2010. (middle) GPROF2010 retrieval without updated screening procedures. Large areas near the core of the convective cells were screened as potential ice surface, flagged, and removed. (right) GPROF2010V2 eliminated screening for surface snow and accurately identifies precipitation in the convective core.

  • View in gallery

    Updated screening procedure for GPROF2010V2. Climatological desert and snow are automatically flagged as unreliable retrievals. All screening procedures shaded red in Fig. 1 are applied globally in GPROF2010V2, while those shaded blue are only used where snow is climatologically possible.

  • View in gallery

    Screening regimes of GPROF2010V2 for (top) January and (bottom) July. Regions where snow was observed in over 75% of days for a particular month were immediately flagged. Snow-likelihood climatology was derived from IMS snow cover data in the Northern Hemisphere and AMSR-E snow cover in the Southern Hemisphere. Heritage screening procedures were used where snow was possible.

  • View in gallery

    (left) Density plot of TMI and AMSR-E collocated observations of 85V and 89V GHz in clear-sky conditions, and (right) 89V GHz after applying correction to AMSR-E brightness temperatures over land.

  • View in gallery

    (top left) GPROF2010 rain-rate retrieval for 15 Apr 2011. Large regions were flagged as potential surface ice contamination and no retrievals were performed. (bottom left) Specific flags triggered during screening routine. Primary flag triggered checks for spatial variability at 89V GHz with an ambiguously cold T24V GHz, indicating a frozen uniform surface is likely (Figs. 1d–f). (top right) Rain-rate retrieval from GPROF2010V2 using the updated screening methodology. (bottom right) Rain accumulation of 1 h measured from NCEP stage IV QPE analysis.

  • View in gallery

    Measure of (top) POD, (middle) FAR, and (bottom) CSI validating GPROF2010V2 against NMQ/Q2 rainfall rates for (left) July 2010. (right) Difference plot shows the relative change of each parameter in comparison to the original GPROF2010 algorithm.

  • View in gallery

    As in Fig. 8, but for January 2010.

  • View in gallery

    Overestimation (red) and underestimation (blue) of monthly accumulations of GPROF2010V2 retrievals relative to GPCC gauge measurements for July 2010.

  • View in gallery

    Daily rain rates averaged over the month of (left) January and (right) July 2010 from (top) GPCP and (middle) GPROF2010V2. (bottom) Changes to daily rain rates between the original version of GPROF2010 and GPROF2010V2 show the spatial changes to the retrieval. Positive (negative) values indicate an increase (decrease) of daily rain rate for GPROF2010V2 relative to GPROF2010. Large decreases of rain totals in winter and at high elevations are related to the elimination of the rain-rate retrieval when snow is expected at the surface.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 28 28 2
PDF Downloads 18 18 2

Updated Screening Procedures for GPROF2010 over Land: Utilization for AMSR-E

View More View Less
  • 1 Cooperative Institute for Climate and Satellites, University of Maryland, College Park, College Park, Maryland
  • | 2 Center for Satellite Applications and Research, NOAA/National Environmental Satellite, Data, and Information Service, College Park, Maryland
  • | 3 I.M. Systems Group, College Park, Maryland
© Get Permissions
Full access

Abstract

The Goddard profiling algorithm 2010 (GPROF2010) was revised for the Advanced Microwave Scanning Radiometer for Earth Observing System (EOS; AMSR-E) instrument. The GPROF2010 land algorithm was developed for the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), which observes slightly different central frequencies than AMSR-E. A linear transfer function was developed to convert AMSR-E brightness temperatures to their corresponding TMI frequency for raining and nonraining instantaneous fields of view (IFOVs) using collocated brightness temperature and TRMM precipitation radar (PR) measurements. Previous versions of the algorithm separated rain from surface ice, snow, and desert using a series of empirical procedures. These occasionally failed to separate raining and nonraining scenes, leading to failed detection and false alarms of rain. The new GPROF2010, version 2 (GPROF2010V2), presented here, prefaced the heritage screening procedures by referencing annual desert and monthly snow climatologies to identify IFOVs where rain retrievals were unreliable. Over a decade of satellite- and ground-based observations from the Interactive Multisensor Snow and Ice Mapping System (IMS) and AMSR-E allowed for the creation of a medium-resolution (0.25° × 0.25°) climatology of monthly snow and ice cover. The scattering signature of rain over ice and snow is not well defined because of complex emissivity signals dependent on snow depth, age, and melting, such that using a static climatology was a more stable approach to defining surface types. GPROF2010V2 was subsequently used for the precipitation environmental data record (EDR) for the AMSR2 sensor aboard the Global Change Observation Mission–Water 1 (GCOM-W1).

Corresponding author address: Patrick Meyers, CICS-MD, University of Maryland, College Park, 5825 University Research Court, Suite 4001, College Park, MD 20740-3823. E-mail: pmeyers@umd.edu

Abstract

The Goddard profiling algorithm 2010 (GPROF2010) was revised for the Advanced Microwave Scanning Radiometer for Earth Observing System (EOS; AMSR-E) instrument. The GPROF2010 land algorithm was developed for the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), which observes slightly different central frequencies than AMSR-E. A linear transfer function was developed to convert AMSR-E brightness temperatures to their corresponding TMI frequency for raining and nonraining instantaneous fields of view (IFOVs) using collocated brightness temperature and TRMM precipitation radar (PR) measurements. Previous versions of the algorithm separated rain from surface ice, snow, and desert using a series of empirical procedures. These occasionally failed to separate raining and nonraining scenes, leading to failed detection and false alarms of rain. The new GPROF2010, version 2 (GPROF2010V2), presented here, prefaced the heritage screening procedures by referencing annual desert and monthly snow climatologies to identify IFOVs where rain retrievals were unreliable. Over a decade of satellite- and ground-based observations from the Interactive Multisensor Snow and Ice Mapping System (IMS) and AMSR-E allowed for the creation of a medium-resolution (0.25° × 0.25°) climatology of monthly snow and ice cover. The scattering signature of rain over ice and snow is not well defined because of complex emissivity signals dependent on snow depth, age, and melting, such that using a static climatology was a more stable approach to defining surface types. GPROF2010V2 was subsequently used for the precipitation environmental data record (EDR) for the AMSR2 sensor aboard the Global Change Observation Mission–Water 1 (GCOM-W1).

Corresponding author address: Patrick Meyers, CICS-MD, University of Maryland, College Park, 5825 University Research Court, Suite 4001, College Park, MD 20740-3823. E-mail: pmeyers@umd.edu

1. Introduction

An accurate record of the global hydrological cycle is critical to identify regions of climatic events, such as droughts and sustained inundation, and short-term hazards, including landslides, flash floods, and tropical cyclones (Hong et al. 2007; Kirschbaum et al. 2009; Kucera et al. 2013). Ground-based gauges measure rain accumulation at a single point to represent quantitative precipitation estimation (QPE) on a much larger spatial scale, which reduces the reliability of rain measurements in locations lacking a dense rain network (Morrissey et al. 1995). Ground-based radars produce a volumetric estimate of raindrop size and distribution to determine accurate near-surface rain rates; however, it is not a source of worldwide rainfall measurements. Radar coverage is sparse on a global scale and has problems with ground clutter, beam blockage, and beam spreading. Low-Earth-orbiting meteorological satellites fill in the observational gaps by providing near-global rain-rate estimates every 3–4 h.

The microwave spectrum is particularly sensitive to water in all states, allowing for retrievals of water vapor, liquid precipitation, and surface snow cover. Rainfall interacts with the microwave emissions from the earth’s surface, such that convective regions can be identified by the scattering of surface emissions by suspended snow, ice, and water (Ferraro et al. 1998).

Spaceborne passive microwave (PMW) imagers have been used to monitor global precipitation since the late 1970s on research missions (Wilheit et al. 1976; Houghton 1979), and operational satellites starting with the Special Sensor Microwave Imager (SSM/I) in 1987, and now several other missions, including the recent launch of the Global Precipitation Measurement (GPM) Microwave Imager (GMI) in 2014.

Historically, the classification of rain/no rain was performed via a scattering index calculation based on observed brightness temperatures (TB) (Grody 1991; Ferraro et al. 1998; Kummerow et al. 2001). Ice particles preferentially scatter microwave emissions at higher frequencies, such that TB depression at T89V relative to lower frequencies indicated precipitation was possibly present. The naming convention for specific TB channels in this manuscript is: [T] [central frequency rounded to the nearest integer] [vertical/horizontal polarization direction]. For example, the TB at 89 GHz with vertical polarization is referred to as T89V. When the scattering index [T24V–T89V] > 8 K (Kummerow et al. 2001) and T89H < 270 K (adapted from Adler et al. 2003), rain was determined to be possible. Unfortunately, this simple relationship was complicated by surface emissions with a similar TB signature in rain-free conditions. Subsequent screening was necessary to separate raining instantaneous fields of view (IFOVs) from scenes contaminated by desert sand, semiarid land, snow, and ice.

Surface screening procedures predominantly rely upon heritage algorithms (Grody 1991; Adler et al. 1994; Ferraro et al. 1994b; Grody and Basist 1996; Ferraro et al. 1998) developed during the SSM/I era of passive microwave remote sensing, based predominantly on TB relationships in the form of discriminant functions. Rain-rate algorithms, including Goddard profiling algorithm 2010 (GPROF2010), combined the strengths of these individual algorithms to create a comprehensive procedure to separate a precipitation signal from background contamination (Kummerow et al. 2001; Gopalan et al. 2010). However, those procedures have become scientifically outdated and can be partially eliminated through the use of ancillary information, which are easily obtainable and suitable for near-real-time use. Additionally, the enhanced spatial resolution of sensors developed after the SSM/I—for example, TMI and AMSR-E—can yield some severe deficiencies when adapting these functions to other sensors.

GPROF2010 was the production rainfall algorithm for TMI and AMSR-E and is the primary rainfall-rate algorithm used in this study. The algorithm was developed for TRMM and therefore is dependent on TBs observed by TMI at a frequency of 10–85 GHz (Table 1). GPROF2010 can be applied to other passive microwave imagers that observe a similar spectrum, including AMSR-E, SSM/I, and SSMIS. The differences between the background characteristics of land and ocean necessitate separate theoretical basis for geophysical retrievals over each surface.

Table 1.

Corresponding frequencies of TMI and AMSR-E. Coefficients for translation from AMSR-E to TMI, with the form TB(TMI) = m × TB(AMSR-E) + b.

Table 1.

The ocean is radiometrically cold because of the low emissivity of water surfaces, so precipitation is identified as a warming signal due to the emission by rain droplets. The ocean segment of GPROF2010 is a Bayesian retrieval, which matches AMSR-E brightness temperatures to an a priori database populated with TRMM precipitation radar (PR) profiles (Kummerow et al. 2011). Each profile has an estimate of surface rain rate with corresponding TMI brightness temperatures. In the retrieval, observed AMSR-E TBs are compared to a priori profiles with similar sea surface temperatures and total precipitable water environments. A priori profiles with the smallest differences from the observed TBs are most heavily weighted when determining the retrieved rain rate. The stable and well-characterized emissivity of the ocean lends itself to such a Bayesian retrieval approach.

The GPROF2010 retrieval algorithm for rain rates over land is most recently described in Gopalan et al. (2010). The algorithm was developed with collocated TMI and PR observations from TRMM. The 85-/89-GHz channels, subsequently referred to as the high-frequency channels, are most sensitive to rain over land. In convective systems, scattering by ice particles aloft of land surface creates a depression in the high-frequency channels relative to the 24-GHz channel, which is less sensitive to ice in the atmosphere and represents an approximate measure of the background land surface. Convective and stratiform rain rates are determined by a third- and first-order polynomial, respectively, dependent on the high-frequency channels. The proportion of convective to stratiform rain is function of T10V, T37V, and the high-frequency channels, as well as the spatial variability and the polarization difference, TV − TH, of the high-frequency channels. Equations for the GPROF2010 land algorithm are explicitly stated in Eqs. (2)–(6) in Gopalan et al. (2010). Currently, a Bayesian retrieval over land is not operationally realistic due to a lack of understanding of the temporal and spatial variability of emissivity (Ferraro et al. 2013; Petty and Li 2013); however, the next generation of GPROF, GPROF2014, developed for GPM, aims to resolve this problem through the use of surface emissivity information.

Despite extensive efforts to create piecewise global algorithms that separate precipitation from surface emission, there are regions where a unified screening algorithm does not work. After decades of remote sensing, a sufficient record exists to create accurate maps of climatologically expected surface conditions. Additionally, improved computer resources and connectivity across various data centers make it feasible to utilize these datasets as opposed to two decades ago, when a self-contained system was required to support operational satellite products. A previous study (Sudradjat et al. 2011) suggested using dynamic predictors for surface-type classification, which works best in a research environment where ancillary surface data can be processed and compiled well after the time of a satellite overpass. GPROF2010 was intended for operational rain-rate calculation, requiring all ancillary data to be available in near–real time. The updates to GPROF2010 presented here did not attempt to improve active detection of snow and ice surfaces, but rather acknowledged the inaccuracies in separating precipitation from frozen surfaces by applying a climatological snow screen. Likewise, retrievals over desert regions were also treated as ambiguous.

2. Methodology/data

a. Collocation

Portability of the GPROF2010 between observing platforms is a strength of the algorithm. Rather than recalculating the rain-rate function for individual satellite platforms, it was advantageous to translate observed TBs, such that they were equivalent to TMI. To correct AMSR-E to TMI, a simple regression with collocated TBs was performed. It should be noted that this is best suited for sensors with similar spatial resolutions, which is the case between TMI and AMSR-E.

The low-inclination orbit of TRMM created daily intersections between TMI and AMSR-E data. A matchup dataset of 3.7 million observations was built from coincident measurements within 30 min and 1 km from the year of 2010. From these data, a simple linear correction function was applied to translate AMSR-E TBs to corresponding TMI frequencies. Subsampling monthly data did not show substantial seasonal trends in the correction function. Additionally, the nature of the conical scanning with near-constant Earth incidence angles limited the necessity for additional view angle corrections between AMSR-E and TMI.

If rain was determined to be likely after completing the screening procedures, a different set of coefficients was used for adjustment to TMI. Presence of rain for the collocation dataset was determined using the TRMM 2A25 PR, version 6 (V6), dataset. Rain detection differences between 2A25 V6 and version 7 (V7) were negligible, such that V6 was suitable for this study (Kirstetter et al. 2013). This correction was only applied to TBs over land because the database for the Bayesian retrieval over the ocean was already adjusted for AMSR-E frequencies. Although simplistic, the collocation and regression process captured the primary differences between the sensors and avoids pitfalls of highly sensitive land surface emissivity modeling.

b. Screening for GPROF2010

Prior to calculating a rain rate, each pixel passed a series of screening tests to determine if the IFOV contained precipitation. Precipitation was radiometically characterized at higher frequencies by the scattering of radiation by ice crystals. Lower frequencies were less sensitive to precipitation, leading to a depression of T89V relative to T24V. The presence of rain was identified when T89H was less than 270 K and the scattering index [T24V – T89H] was greater than 8 K. Subsequent calculations were necessary to determine if the IFOV was contaminated by desert, snow, or ice surfaces. The screening procedures for GPROF2010 were built off an assortment of heritage techniques designed to separate raining and nonraining scenes. An overview of the original screening protocol for GPROF2010 over land is provided in Fig. 1.

Fig. 1.
Fig. 1.

Screening procedure logic for GPROF2010. Green (red) arrows indicate that the conditions for flagging were met (not met). Blue shading indicates screens designed to identify frozen ground, surface ice, or fallen snow. In GPROF2010, all screening procedures were applied globally, including regions where snow and ice were never observed.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

IFOVs were classified as “land” when >95% of the viewable area was land using an internal 1/16° land/sea mask. The land screening routine first determined if surface water including rivers and small lakes contaminated the TBs. Liquid water surfaces were radiometrically cold relative to land, such that water over land was identified by low T24V when a depression at T89H was not present (Fig. 1a; Adler et al. 1994). This flag was typically triggered near wide river basins and swamplands. When T24V < 269.1 K, the Grody (1991) criteria were used to screen for cold surfaces, such as ice and fallen, melted, or refrozen snow. Heavily precipitating convective regions can contain large ice particles, causing a decrease in T24V below the threshold value. A secondary condition needed to be met to separate snow from heavy precipitation based on a linear relationship between T24V and T89H (Fig. 1b). Cold desert surfaces were identified when the polarization difference at 19 GHz, T19V – T19H, was greater than 20 K. Furthermore, semiarid land was identified globally as a function of polarization difference at T19 in relation to T89H (Fig. 1c). GPROF2010 also calculated the local variance of T89V in a 5 pixel × 5 pixel box surrounding the central IFOV. Precipitation had high spatial variability over a small area, whereas cold surfaces are more uniform. If the standard deviation did not exceed 30 K, then the IFOV was flagged as an ambiguous cold surface. After completing the screening procedures, a uniformity check eliminated isolated IFOVs completely surrounded a dissimilar flag.

The screening procedures over land in GPROF2010 were applied globally regardless of typical surface conditions. The percentage of observations where particular snow (Fig. 1, case B) and semiarid land flags (Fig. 1, case G) were triggered throughout 2010 showed that the screening procedures typically produce realistic surface conditions (Fig. 2). The snow screening captured the widespread snow cover in the Northern Hemisphere winter, the snowmelt in the spring into summer, and fresh snow accumulation in late fall. The snow screening also identified persistent snow at high latitudes and high elevations. While the seasonality and long-term features were captured on monthly time scales, some IFOVs at the swath level were identified as snow-covered ground in areas where snow is nearly never observed (Fig. 3). Although misclassifications were infrequent, the incorrectly flagged IFOVs can result in missed heavy rain events. Semiarid and desert surfaces were persistently observed throughout the year (Fig. 2). The TB screening procedures were not able to identify all instances of surface contamination in transition zones between semiarid and thicker vegetation.

Fig. 2.
Fig. 2.

Frequency at which the (left) snow and (right) semiarid land flags were triggered in 2010, corresponding to cases B and G in Fig. 1, respectively.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

Fig. 3.
Fig. 3.

Frequency that the TB snow flag was triggered in April 2010. Red line indicates the 0% contour of the snow frequency climatology.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

c. New screening logic

All of the screening algorithms were applied across all seasons and surface regimes, such that observed convective cores would often be flagged as possible surface snow or ice in warmer months (Fig. 4). A simple change in logic addressed this problem (Fig. 5). In GPROF2010, version 2 (V2), all climatological snow and desert surfaces were immediately flagged, and ice screening procedures were only executed where the presence of snow or ice was observed 5% of the time based on climatology. This eliminated false flagging of precipitation as ice or snow, particularly in convective summer months and in regions where snow rarely occurred.

Fig. 4.
Fig. 4.

(left) NMQ/Q2 rain rates for a line of convective storms from 1920 UTC 30 Apr 2010. (middle) GPROF2010 retrieval without updated screening procedures. Large areas near the core of the convective cells were screened as potential ice surface, flagged, and removed. (right) GPROF2010V2 eliminated screening for surface snow and accurately identifies precipitation in the convective core.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

Fig. 5.
Fig. 5.

Updated screening procedure for GPROF2010V2. Climatological desert and snow are automatically flagged as unreliable retrievals. All screening procedures shaded red in Fig. 1 are applied globally in GPROF2010V2, while those shaded blue are only used where snow is climatologically possible.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

In winter, surface snow and ice presented challenges in identifying precipitation. Warm-season convection produced a well-defined scattering signature compared to surface radiance; however, rain in the cold season had a similar signature to surface snow and ice. Based on current screening methodologies, it was not possible to entirely separate rain and cold surfaces globally exclusively using TB relationships. Because of a lack of confidence of rain-rate retrievals over cold surfaces, a monthly snow and ice climatology was used to flag IFOVs where the likelihood of surface snow is greater than 75%. The rain-rate retrieval was not performed for these IFOVs because of a lack of confidence separating rain from snow and ice. Likewise, SWE retrieval algorithms struggled in separating precipitation from snow (Grody 1991; Ferraro et al. 1994a), often employing a snow climatology to ensure the observed TB signal was likely snow and not precipitation scattering.

A similar use of climatological surface conditions was used over the desert. Historically, screening algorithms relied solely on TB relationships to identify arid land (Ferraro et al. 1994b) measured by the polarization difference at 19 GHz. Barren surfaces were successfully screened in most cases, but the screening protocol was supplemented by flagging all IFOVs over desert regions. Even if precipitation was identified by GPROF2010 or TRMM PR, typical dry desert conditions cause falling rain to evaporate, such that surface rain rates cannot be accurately quantified. Figure 6 summarizes screening regimes for January and July.

Fig. 6.
Fig. 6.

Screening regimes of GPROF2010V2 for (top) January and (bottom) July. Regions where snow was observed in over 75% of days for a particular month were immediately flagged. Snow-likelihood climatology was derived from IMS snow cover data in the Northern Hemisphere and AMSR-E snow cover in the Southern Hemisphere. Heritage screening procedures were used where snow was possible.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

d. Data

1) Surface-type climatologies

The climatology used to determine the expected surface snow conditions was built from the Interactive Multisensor Snow and Ice Mapping System (IMS) snow cover analysis in the Northern Hemisphere (Ramsay 1998; National Ice Center 2008). IMS manually merges snow cover estimates from a wide variety of observing platforms, including visible imagery from geostationary and polar-orbiting satellites, PMW SWE retrievals, and ground observations. Daily polar stereographic maps determining snow or no snow at ~25-km resolution were accumulated from January 1999 to April 2012. A monthly climatology of the likelihood of snow, defined as a percentage of daily IMS analyses with snow present, was produced. In this context, a climatology value of 75% indicated that snow existed for 75% of days in the IMS data record.

The IMS product is not produced for the Southern Hemisphere, so monthly AMSR-E level 3 SWE retrievals from 2002 to 2011 were used as an alternative (Tedesco et al. 2004). In the Southern Hemisphere, snow was typically restricted to the Andes mountain range and parts of New Zealand. Snow was assumed always present in Antarctica. The lack of large-scale shifts in snow cover permitted the use of a single snow product in the Southern Hemisphere. The Northern and Southern Hemisphere products were merged to create a 0.25° Mercator projection of monthly snow likelihood.

The desert mask was extracted from the Moderate Resolution Imaging Spectroradiometer (MODIS) surface-type classification annual climatology from NASA Langley Research Center’s Clouds and the Earth’s Radiant Energy System Surface and Atmosphere Radiation Budget (CERES SARB) following the International Geosphere–Biosphere Programme surface identification standards (Wilber et al. 1999). IFOVs were flagged on a 0.25° grid if a majority of IFOVs within a 0.5° radius were classified as barren or desert.

2) Rainfall

GPROF2010 was run internally from AMSR-E level 2A, version 2.10, brightness temperatures (Ashcroft and Wentz 2006). The modular GPROF system allowed for interchangeable routines to implement updated code without tampering with the overall code structure. Swath rain-rate data from TMI was extracted from NASA’s 2A12 hydrometeor profile product and the 2A25 precipitation radar rainfall rate, version 7. Validation over the continental United States (CONUS) was performed with the National Mosaic and Multi-Sensor QPE (NMQ) next-generation (Q2) analysis (Zhang et al. 2011; Kirstetter et al. 2012). The NMQ/Q2 system assimilates radar reflectivities to produce precipitation rates at 1 km every 5 min that are bias corrected with hourly automated rain gauge measurements, creating high-resolution rain-rate maps that can be directly compared to AMSR-E retrievals. The postprocessed Q2 dataset was provided by the National Severe Storms Laboratory with additional quality controls for satellite comparisons.

3. Results

PMW rain rates are used for applications on hourly to yearly time scales. This requires accurate estimates of instantaneous rain rates at the swath level, as well as realistic near-global seasonal. Validation of GPROF2010V2 was performed across several time scales to justify implementing a TMI algorithm for AMSR-E retrievals and using climatological desert and snow information in the process of determining the presence of rain.

a. Brightness temperature correction

To use TMI TB–rain-rate relationships for AMSR-E, linear regressions were performed. Table 1 summarizes the regression parameters for AMSR-E frequencies adjustments to TMI. The simple correction had the form TB′ = m × TBAMSR-E + b, where TB′ is the AMSR-E brightness temperature adjusted to the corresponding TMI frequency. Differences between collocated TMI and corrected AMSR-E data were independent of brightness temperature and centered around a mean bias near 0 (Fig. 7). At the time of the study, GPM’s intersatellite-calibrated level 1C (L1C) brightness temperatures were still evolving. Residual biases between corrected AMSR-E and TMI TBs persisted when using preliminary L1C corrections. Future updates to GPROF2010 can be developed from the L1C TB data.

Fig. 7.
Fig. 7.

(left) Density plot of TMI and AMSR-E collocated observations of 85V and 89V GHz in clear-sky conditions, and (right) 89V GHz after applying correction to AMSR-E brightness temperatures over land.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

b. Surface flagging improvement: Case study

To better understand further the impact of each individual screen, the situation of improper screening of surface snow cover was examined. Strong convective cores were frequently flagged as possible snow cover using the original screening procedures. It was speculated that the scattering was observed by the AMSR-E 24-GHz channel because of its enhanced spatial resolution as compared to the SSM/I 22-GHz channel, where the scattering effects were much less pronounced or not observed at all. Figure 8 highlights a case in the northern Midwest from 15 April 2011. An afternoon overpass identified several cells of heavy precipitation moving north-northeast along a cold front. This system produced some quarter-sized hail near the time of observation and spawned a tornado several hours later.

Fig. 8.
Fig. 8.

(top left) GPROF2010 rain-rate retrieval for 15 Apr 2011. Large regions were flagged as potential surface ice contamination and no retrievals were performed. (bottom left) Specific flags triggered during screening routine. Primary flag triggered checks for spatial variability at 89V GHz with an ambiguously cold T24V GHz, indicating a frozen uniform surface is likely (Figs. 1d–f). (top right) Rain-rate retrieval from GPROF2010V2 using the updated screening methodology. (bottom right) Rain accumulation of 1 h measured from NCEP stage IV QPE analysis.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

To further decompose the impact of the screening on the overall rain-rate retrieval, IFOVs were mapped and color-coded based on each discriminant function shown in Fig. 1. Blue regions in Fig. 8 were flagged as an unambiguous cold surface because T24V was less than a threshold value for cold scenes (269.1 K) and the standard deviation of T89V in a 5 × 5 grid around the center IFOV was less than 20 K (corresponding to the flag in Fig. 1e). Small spatial variability of T89V typically indicates a homogenous ice surface, instead of the typically irregular signal produced by precipitation. Areas in light blue identify where the Grody (1991) screening criteria determined snow was likely contaminating the retrieval based on the relationship of T24V and T89V. Hail on the ground could generate an accurately designated screening flag (Cecil 2009), but the hail swath would be much narrower than the flagged region of this case study.

The implementation of the new screening logic eliminated flagging for cold surfaces because snow was unlikely, so unflagged rain rates were retrieved over the domain. It was clear that the regions that were previously flagged produced realistic rain rates with the new screening regime. Characteristics of the GPROF retrieval match well with the stage IV hourly rainfall estimates. Local precipitation maxima compared well; however, the spatial extent of the precipitation was overestimated by GPROF. Excessive screening caused the mean rainfall rate to be underestimated by 25%. In the original case, the average rain rate over the mapped domain was 1.34 mm h−1. The updated algorithm produced 1.85 mm h−1, close to the stage IV–derived rate of 1.77 mm h−1. This incorrect flagging was not an isolated incident and occurred in many convective cells within synoptic systems. Missing heavy precipitation events not only causes errors in near-real-time applications but also causes an underestimate of precipitation on monthly time scales.

c. Validation over CONUS: Q2 analysis

Q2 gauge-adjusted radar-derived rain-rate estimates provided independent measurements with high spatial (1 km) and temporal (5 min) resolution. Radar and satellite retrieval algorithms infer surface rain rates from the radiometric signature of a volume of hydrometeors. Different scan volumes and observation fundamentals created differences in observed rain rates; however, the Q2 measurements provided a stable ground truth for satellite-derived values. The most recent AMSR-E crossing of 40°N defined the time of the Q2 data. Q2 rain rates were averaged from a 5 km × 5 km grid around the center of AMSR-E FOV at 89 GHz, weighted by the distance to the center of the FOV. The 5 km × 5 km averaging corresponded to the approximate footprint size of the 89-GHz channel, which is most sensitive to precipitation. The high spatial and temporal resolution of the Q2 data allowed nearly instantaneous comparisons of absolute rain rates as well as skill parameters, including probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) (Table 2). These skill parameters were calculated from the contingency table where rain was considered present when the rain rate exceeded 1 mm h−1. Comparisons were not made in light precipitation because satellite detection of rain rates below 1 mm h−1 is unreliable for sensors similar to TMI (Munchak and Skofronick-Jackson 2013).

Table 2.

Summary of GPROFV2 contingency table used to calculate skill parameters. Values in the table represent the total count of observations satisfying the rain-rate conditions, where POD = H/(H + M); FAR = F/(H + F); CSI = H/(H + M + F).

Table 2.

In summer, eliminating the snow screening procedures led to increased rainfall detection over much of CONUS. POD was typically near 80%, with an 8% improvement using GPROF2010V2 (Fig. 9). Radar-based rain rates westward of the Rocky Mountains have been shown to be less reliable (Janowiak 2007; Sapiano et al. 2010; Tang et al. 2014) due to beam blockage and rough terrain. Additionally, the scattering index (T24V – T89V) of the uncorrected AMSR-E brightness temperatures was typically 3 K lower than the scattering index from TMI, leading to a higher frequency of rain and false alarms for GPROF2010V2, particularly over semiarid surfaces. Outside of these regions, CSI showed an overall 5% improvement in determining presence of rainfall, which supports the universal use of the GPROF2010V2 screening procedures.

Fig. 9.
Fig. 9.

Measure of (top) POD, (middle) FAR, and (bottom) CSI validating GPROF2010V2 against NMQ/Q2 rainfall rates for (left) July 2010. (right) Difference plot shows the relative change of each parameter in comparison to the original GPROF2010 algorithm.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

Persistent snow cover throughout the Northeast, northern Midwest, and Rocky Mountain regions inhibited rain-rate retrievals over much of CONUS in January 2010 (Fig. 10). GPROF2010V2 performed well in detecting rainfall over much of the Southwest and mid-Atlantic. The updated algorithm improves detection by about 10% throughout these regions and the Southwest. These improvements were mainly attributed to eliminating the snow screening procedures in areas where snow was climatologically unlikely. The gains in POD were coupled with increased FAR; however, CSI indicated an overall improvement to rainfall detection for January 2010.

Fig. 10.
Fig. 10.

As in Fig. 8, but for January 2010.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

d. Flagging: Monthly time scales

Although the screening procedures correctly characterized the expected monthly distribution, the algorithm sometimes produced a snow or semiarid flag in unlikely regions. Inspection of Fig. 2 revealed the snow screen was occasionally triggered in central Africa and over Paraguay and southern Brazil throughout the year. This misclassification reduced rainfall totals and created noncontinuous precipitation fields, as highlighted in the previous section. Figure 3 shows where the snow screen was activated over the United States. IFOVs throughout the Southeast were flagged due to expected snow and ice contamination. The simple correction to this problem was to only execute the snow screening procedures where snow climatologically occurred.

e. Monthly mean comparisons

1) GPCC gauges

The Global Precipitation Climatology Centre (GPCC) global network of rain gauges is considered one of the best globally maintained datasets and provided a ground-based system to validate GPROF retrievals (Schneider et al. 2011). Gauge estimates were compared to AMSR-E monthly rain retrievals from GPROF2004 and GPROF2010V2. GPROF2004 was the primary algorithm for AMSR-E rain-rate retrievals, and this was used as the benchmark for this comparison. Monthly estimates of rainfall from gauges were accumulated on a 2.5° × 2.5° grid for January, April, and July 2010 (Fig. 11). The monthly AMSR-E rainfall were computed by simply aggregating all of the instantaneous rain rates over the month within the grid box, computing the mean value and multiplying by the number of hours in the month. Grid boxes with fewer than five gauges were removed from the sample. Observations were highly concentrated over the United States, western Europe, East Asia, and coastal Australia, with sparse observations through much of South America, Africa, and central Asia.

Fig. 11.
Fig. 11.

Overestimation (red) and underestimation (blue) of monthly accumulations of GPROF2010V2 retrievals relative to GPCC gauge measurements for July 2010.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

GPROF2004 underestimated rain rates relative to the GPCC gauges with an overall negative bias of 1.1 mm day−1. RMSE was 2.3 mm day−1 and the slope of the regression line was 0.41. The GPROF2010V2 updated algorithm dramatically improves the rain-rate estimates in comparison to the GPCC gauges. The bias and RMSE were reduced to 0.4 and 2.0 mm day−1, respectively. Significant variability between gauges and satellite retrievals remained; however, GPROF2010 improved correlations by 12%.

Subsampling the data to monthly values did not reveal significant seasonal trends in overall accuracy of GPROF retrievals. Comparing the geographical distribution of errors of July 2010 showed the reduction of the low bias occurred predominately in coastal regions and higher latitudes. The overall increased rain estimates of GPROF2010V2 led to overestimation of monthly accumulation over the midlatitudes in North America and eastern Europe.

2) GPCP monthly means

The Global Precipitation Climatology Project (GPCP), version 2.2, combined precipitation set merges monthly rain estimates from the GPCC gauge network and microwave and infrared satellites (Adler et al. 2003; Huffman et al. 2009). GPCP rain-rate measurements reduce regime-based systematic errors in satellite retrievals, such as in snowy regions. The GPCP analysis highlighted the erroneous rainfall measurements year-round in the Himalayas and in the winter over the Rocky Mountains using the original GPROF2010 algorithm. In these regions the screening procedures routinely failed to identify snow contamination, producing unrealistic rain rates. Utilizing the climatological snow screen in GPROF2010V2, the areas with snow contamination were effectively removed. Rather than producing unreliable rain rates, GPROF2010V2 flagged all areas where snow could cause ambiguous retrievals. Elsewhere in the winter hemisphere, GPROF continued to struggle to detect light rain because of the lack of ice scattering signal.

GPROF2010V2 performed better in convective regimes, including over South America and Africa. It accurately identified the spatial extent of precipitation and locations of local maxima and minima. A persistent problem with microwave retrievals is overestimation of rainfall over Africa and underestimation over South America (Huffman et al. 2007). Screening changes in GPROF2010V2 enhanced summer rainfall totals in South America, bringing measurements closer to the GPCP analysis (Fig. 12). A similar increase to rainfall occurred over India during the summer monsoon season.

Fig. 12.
Fig. 12.

Daily rain rates averaged over the month of (left) January and (right) July 2010 from (top) GPCP and (middle) GPROF2010V2. (bottom) Changes to daily rain rates between the original version of GPROF2010 and GPROF2010V2 show the spatial changes to the retrieval. Positive (negative) values indicate an increase (decrease) of daily rain rate for GPROF2010V2 relative to GPROF2010. Large decreases of rain totals in winter and at high elevations are related to the elimination of the rain-rate retrieval when snow is expected at the surface.

Citation: Journal of Atmospheric and Oceanic Technology 32, 5; 10.1175/JTECH-D-14-00149.1

4. Discussion and summary

Through the use of primarily ancillary-derived climatologies of snow cover and arid land regions, GPROF2010V2 was shown to improve precipitation retrievals for AMSR-E. Results showed that utilizing the monthly climatological snow and a fixed desert data reduces the frequency of incorrectly identifying surface emission signals as rain. Further improvements could be obtained by utilizing even more enhanced climatologies, especially with the arid land surfaces, or even fairly current estimates of those surfaces from near-real-time analyses, such as those used in the assimilation of NWP models. Furthermore, the heritage screening procedures usually correctly identify surface snow and ice, but when these checks failed, the algorithm produced implausible convective rain retrievals in cold weather regimes. While a subjective observer can identify these false rain rates, applications that rely on satellite retrievals would be negatively impacted. GPROF2010V2 excludes possibly inaccurate retrievals rather than producing low-confidence retrievals. Conversely, in warmer seasons, the snow cover climatology eliminated the overproduction of IFOVs flagged due to ice contamination. Some convective storms produced strong scattering signals that triggered screens intended to identify cold surfaces. By only executing snow screening protocols when surface snow was climatologically possible, more continuous and realistic rain fields were produced.

Results from GPROF2010V2 provide a baseline for the next generation of GPROF’s land algorithm being developed for GPM. GPROF2014 attempts to classify land emissivity and perform a Bayesian inversion retrieval similar to GPROF2010’s ocean rain-rate algorithm. Regime-specific rain-rate algorithms are expected to replace universal algorithms as understanding and measurement of emissivity and hydrometeor microphysics improves. A persistent artifact in microwave retrievals is an overestimation of rain rates in Africa and an underestimation over South America. The empirical nature of GPROF2010V2 did not explicitly account for the varying hydrometeor microphysics between these regions. Reduction of incorrect flagging of heavy precipitation increased daily average rain rates over the Amazon rain forest more than in Africa.

Retrievals over coastal regimes continue to be a problem. A modified version of the screening procedures over land is applied to coastal IFOVs, although a detailed methodology must be developed to account for the orientation of the IFOV relative to the coastline and the radiometric contribution from land and ocean. Identification of contamination by surface ice and snow will continue to be a problem for microwave precipitation retrievals. Future PMW radiometers that measure higher frequencies beyond 89 GHz, such as GPM, will be better equipped to identify snow and ice contamination.

PMW rain-rate retrievals are still an imperfect science. The use of climatological ancillary information bypasses the inherent uncertainties associated with detecting rain over snow and desert. A benefit of the design of the algorithm is that it can be easily ported to other sensors, including TMI, SSM/I, and SSMIS. Additionally, GPROF2010V2 was the day 1 algorithm for AMSR2 aboard the Global Change Observation Mission–Water 1 (GCOM-W1) satellite in conjunction with NOAA’s Joint Polar Satellite System.

Acknowledgments

The authors thank Drs. Chris Kummerow, Dave Randel, and Wesley Berg at Colorado State University for their help implementing the GPROF2010 algorithm. We thank Dr. Pierre Kirstetter from the University of Oklahoma and the National Severe Storms Laboratory for the NMQ/Q2 data collocated with AMSR-E overpasses. GPCC and GPCP precipitation data were provided by NOAA/OAR/ESRL/PSD, Boulder, Colorado. The authors thank the anonymous reviewers, whose insights improved the quality of this manuscript.

REFERENCES

  • Adler, R. F., , Huffman G. J. , , and Keehn P. R. , 1994: Global tropical rain estimates from microwave-adjusted geosynchronous IR data. Remote Sens. Rev., 11, 125152, doi:10.1080/02757259409532262.

    • Search Google Scholar
    • Export Citation
  • Adler, R. F., and et al. , 2003: The Version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 11471167, doi:10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ashcroft, P., , and Wentz F. J. , 2006: AMSR-E/Aqua L2A global swath spatially-resampled brightness temperatures, version 2. NASA National Snow and Ice Data Center DAAC. Subset used: Low resolution swath data, accessed 5 October 2011. [Available online at http://nsidc.org/data/docs/daac/ae_l2a_tbs/v2/ae_l2a_tbs.gd.html.]

  • Cecil, D. J., 2009: Passive microwave brightness temperatures as proxies for hailstorms. J. Appl. Meteor. Climatol., 48, 12811286, doi:10.1175/2009JAMC2125.1.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , Grody N. C. , , and Marks G. F. , 1994a: Effects of surface conditions on rain identification using the SSM/I. Remote Sens. Rev., 11, 195209, doi:10.1080/02757259409532265.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 1994b: Microwave measurements produce global climatic, hydrologic data. Eos, Trans. Amer. Geophys. Union, 75, 337343, doi:10.1029/94EO00988.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , Smith E. A. , , Berg W. , , and Huffman G. J. , 1998: A screening methodology for passive microwave precipitation retrieval algorithms. J. Atmos. Sci., 55, 15831600, doi:10.1175/1520-0469(1998)055<1583:ASMFPM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 2013: An evaluation of microwave land surface emissivities over the continental United States to benefit GPM-era precipitation algorithms. IEEE Trans. Geosci. Remote Sens., 51, 378398, doi:10.1109/TGRS.2012.2199121.

    • Search Google Scholar
    • Export Citation
  • Gopalan, K., , Wang N.-Y. , , Ferraro R. R. , , and Liu C. , 2010: Status of the TRMM 2A12 land precipitation algorithm. J. Atmos. Oceanic Technol., 27, 13431354, doi:10.1175/2010JTECHA1454.1.

    • Search Google Scholar
    • Export Citation
  • Grody, N. C., 1991: Classification of snow cover and precipitation using the Special Sensor Microwave Imager. J. Geophys. Res., 96, 74237435, doi:10.1029/91JD00045.

    • Search Google Scholar
    • Export Citation
  • Grody, N. C., , and Basist A. , 1996: Global identification of snow cover using SSM/I measurements. IEEE Trans. Geosci. Remote Sens., 34, 237249, doi:10.1109/36.481908.

    • Search Google Scholar
    • Export Citation
  • Hong, Y., , Adler R. , , and Huffman G. , 2007: Use of satellite remote sensing data in the mapping of global landslide susceptibility. J. Nat. Hazards, 43, 245256, doi:10.1007/s11069-006-9104-z.

    • Search Google Scholar
    • Export Citation
  • Houghton, J. T., 1979: The future role of observations from meteorological satellites. Quart. J. Roy. Meteor. Soc., 105, 123, doi:10.1002/qj.49710544302.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and et al. , 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 3855, doi:10.1175/JHM560.1.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., , Adler R. F. , , Bolvin D. T. , , and Gu G. , 2009: Improving the global precipitation record: GPCP Version 2.1. Geophys. Res. Lett., 36, L17808, doi:10.1029/2009GL040000.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J., 2007: Validation of rainfall algorithms at the NOAA Climate Prediction Center. Measuring Precipitation from Space: EURAINSAT and the Future, V. Levizzani, P. Bauer, and F. Joseph, Eds., Advances in Global Change Research, Vol. 28, Springer Netherlands, 393–401.

  • Kirschbaum, D. B., , Adler R. , , Hong Y. , , and Lerner-Lam A. L. , 2009: Evaluation of a satellite-based landslide algorithm using global landslide inventories. Nat. Hazards Earth Syst. Sci., 9, 673686, doi:10.5194/nhess-9-673-2009.

    • Search Google Scholar
    • Export Citation
  • Kirstetter, P.-E., and et al. , 2012: Toward a framework for systematic error modeling of spaceborne precipitation radar with NOAA/NSSL ground radar–based National Mosaic QPE. J. Hydrometeor., 13, 12851300, doi:10.1175/JHM-D-11-0139.1.

    • Search Google Scholar
    • Export Citation
  • Kirstetter, P.-E., , Hong Y. , , Gourley J. J. , , Schwaller M. , , Petersen W. , , and Zhang J. , 2013: Comparison of TRMM 2A25 products, version 6 and version 7, with NOAA/NSSL ground radar–based National Mosaic QPE. J. Hydrometeor., 14, 661669, doi:10.1175/JHM-D-12-030.1.

    • Search Google Scholar
    • Export Citation
  • Kucera, P. A., , Ebert E. E. , , Turk F. J. , , Levizzani V. , , Kirschbaum D. , , Tapiador F. J. , , Loew A. , , and Borsche M. , 2013: Precipitation from space: Advancing Earth system science. Bull. Amer. Meteor. Soc., 94, 365375, doi:10.1175/BAMS-D-11-00171.1.

    • Search Google Scholar
    • Export Citation
  • Kummerow, C. D., and et al. , 2001: The evolution of the Goddard Profiling Algorithm (GPROF) for rainfall estimation from passive microwave sensors. J. Appl. Meteor., 40, 18011820, doi:10.1175/1520-0450(2001)040<1801:TEOTGP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kummerow, C. D., , Ringerud S. , , Crook J. , , Randel D. , , and Berg W. , 2011: An observationally generated a priori database for microwave rainfall retrievals. J. Atmos. Oceanic Technol., 28, 113130, doi:10.1175/2010JTECHA1468.1.

    • Search Google Scholar
    • Export Citation
  • Morrissey, M. L., , Maliekal J. A. , , Greene J. S. , , and Wang J. , 1995: The uncertainty of simple spatial averages using rain gauge networks. Water Resour. Res., 31, 20112017, doi:10.1029/95WR01232.

    • Search Google Scholar
    • Export Citation
  • Munchak, S. J., , and Skofronick-Jackson G. , 2013: Evaluation of precipitation detection over various surfaces from passive microwave imagers and sounders. Atmos. Res., 131, 8194, doi:10.1016/j.atmosres.2012.10.011.

    • Search Google Scholar
    • Export Citation
  • National Ice Center, 2008: IMS daily Northern Hemisphere snow and ice analysis at 1 km, 4 km, and 24 km resolutions. National Snow and Ice Data Center. Subset used: 24 km, accessed 27 February 2012, doi:10.7265/N52R3PMC.

  • Petty, G. W., , and Li K. , 2013: Improved passive microwave retrievals of rain rate over land and ocean. Part I: Algorithm description. J. Atmos. Oceanic Technol., 30, 24932508, doi:10.1175/JTECH-D-12-00144.1.

    • Search Google Scholar
    • Export Citation
  • Ramsay, B. H., 1998: The interactive multisensor snow and ice mapping system. Hydrol. Processes, 12, 15371546, doi:10.1002/(SICI)1099-1085(199808/09)12:10/11<1537::AID-HYP679>3.0.CO;2-A.

    • Search Google Scholar
    • Export Citation
  • Sapiano, M., , Janowiak J. E. , , Shi W. , , Higgins R. W. , , and Silva V. B. S. , 2010: Regional evaluation through independent precipitation measurements: USA. Satellite Rainfall Applications for Surface Hydrology, M. Gebremichael and F. Hossain, Eds., Springer Science, 169–191.

  • Schneider, U., , Becker A. , , Finger P. , , Meyer-Christoffer A. , , Rudolf B. , , and Ziese M. , 2011: GPCC full data reanalysis version 6.0 (at 0.5°, 1.0°, 2.5°): Monthly land-surface precipitation from rain-gauges built on GTS-based and historic data. Global Precipitation Climatology Centre. Subset used: 2.5° grid resolution, accessed 12 November 2013, doi:10.5676/DWD_GPCC/FD_M_V6_250.

  • Sudradjat, A., , Wang N.-Y. , , Gopalan K. , , and Ferraro R. R. , 2011: Prototyping a generic, unified land surface classification and screening methodology for GPM-era microwave land precipitation retrieval algorithms. J. Appl. Meteor. Climatol., 50, 12001211, doi:10.1175/2010JAMC2572.1.

    • Search Google Scholar
    • Export Citation
  • Tang, L., , Tian Y. , , and Lin X. , 2014: Validation of precipitation retrievals over land from satellite-based passive microwave sensors. J. Geophys. Res. Atmos., 119, 45464567, doi:10.1002/2013JA018955.

    • Search Google Scholar
    • Export Citation
  • Tedesco, M., , Kelly R. , , Foster J. L. , , and Chang A. T. C. , 2004: AMSR-E/Aqua daily L3 global snow water equivalent EASE-Grids, version 2. NASA National Snow and Ice Data Center DAAC. Subset used Monthly SWE, accessed 23 April 2012. [Available online at http://nsidc.org/data/docs/daac/ae_swe_ease-grids.gd.html.]

  • Wilber, A. C., , Kratz D. P. , , and Gupta S. K. , 1999: Surface emissivity maps for use in satellite retrievals of longwave radiation. NASA Tech. Publ. NASA/TP-1999-209362, 35 pp. [Available online at https://eosweb.larc.nasa.gov/sites/default/files/project/calipso/Wilber.NASATchNote99.pdf.]

  • Wilheit, T. T., , Theon J. S. , , Shenk W. E. , , Allison L. J. , , and Rodgers E. B. , 1976: Meteorological interpretations of the images from the Nimbus 5 electrically scanned microwave radiometer. J. Appl. Meteor., 15, 166172, doi:10.1175/1520-0450(1976)015<0166:MIOTIF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhang, J., and et al. , 2011: National Mosaic and Multi-Sensor QPE (NMQ) system: Description, results, and future plans. Bull. Amer. Meteor. Soc., 92, 13211338, doi:10.1175/2011BAMS-D-11-00047.1.

    • Search Google Scholar
    • Export Citation
Save