• Arkin, P. A., 1988: The global climate for June–August 1987: Mature phase of an ENSO warm episode persists. J. Climate,1, 306–324.

  • Atkinson, R. J., 1984: Automated mesoscale winds determined from satellite imagery. Interim Report on NAS8-34596, General Electric Company, Huntsville, AL, 51 pp. [Available from LMC/GHCC, 977 Explorer Blvd., Huntsville, AL 35806.].

  • Atkinson, R. J., 1987: Automated mesoscale winds determined from satellite imagery. Final Report on NAS8-34596, General Electric Company, Huntsville, AL, 50 pp. [Available from LMC/GHCC, 977 Explorer Blvd., Huntsville, AL 35806.].

  • Bates, J. J., X. Wu, and D. L. Jackson, 1996: Interannual variability of upper-tropospheric water vapor band brightness temperature. J. Climate,9, 427–438.

  • Bell, G. D., and M. S. Halpert, 1998: Climate assessment for 1997. Bull. Amer. Meteor. Soc.,79, S1–S50.

  • Blackwell, K. G., and J. P. McGuirk, 1996: Tropical upper-tropospheric dry regions from TOVS and rawinsondes. J. Appl. Meteor.,35, 464–481.

  • Bony, S., Y. Sud, K. M. Lau, J. Susskind, and S. Saha, 1997: Comparison and satellite assessment of NASA/DAO and NCEP–NCAR reanalyses over tropical ocean: Atmospheric hydrology and radiation. J. Climate,10, 1441–1462.

  • Del Genio, A. D., 1994: Climate implications of the seasonal variation of upper tropospheric water vapor. Geophys. Res. Lett.,21, 2701–2704.

  • Eigenwillig, N., and H. Fischer, 1982: Determination of midtropospheric wind vectors by tracking pure water vapor structures in METEOSAT water vapor image sequences. Bull. Amer. Meteor. Soc.,63, 44–58.

  • Elliot, W. P., and D. J. Gaffen, 1991: On the utility of rawinsonde humidity archived for climate studies. Bull. Amer. Meteor. Soc.,72, 1507–1520.

  • Endlich, R. M., and D. E. Wolf, 1981: Automatic cloud tracking applied to GOES and METEOSAT observations. J. Appl. Meteor.,20, 309–319.

  • Fischer, H., N. Eigenwillig, and H. Muller, 1981: Information content of METEOSAT and Nimbus/THIR water vapor channel data: Altitude association of observed phenomena. J. Appl. Meteor.,20, 1344–1352.

  • Fritz, S., and J. S. Winston, 1962: Synoptic use of radiation measurements from satellite TIROS-II. Mon. Wea. Rev.,90, 1–9.

  • Fuelberg, H. E., and P. J. Meyer, 1984: An analysis of the AVE–SESAME I period using statistical structure and correlation functions. Mon. Wea. Rev.,112, 1562–1576.

  • Gray, W. M., 1984: Atlantic seasonal hurricane frequency. Part I: El Niño and 30 mb quasi-biennial oscillation influences. Mon. Wea. Rev.,112, 1649–1668.

  • Hillger, D. W., and T. H. Vonder Haar, 1988: Estimating noise levels of remotely sensed measurements from satellites using spatial structure analysis. J. Atmos. Oceanic Technol.,5, 206–214.

  • Janowiak, J. E., 1988: The global climate for March–May 1988: The end of the 1986–87 Pacific warm episode and the onset of widespread drought in the United States. J. Climate,1, 1019–1040.

  • Jedlovec, G. J., and R. J. Atkinson, 1996: Quality and control of water vapor winds. Preprints, Eighth Conf. on Satellite Meteorology and Oceanography, Atlanta, GA, Amer. Meteor. Soc., 5–8.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year Reanalysis Project. Bull. Amer. Meteor. Soc.,77, 437–471.

  • Kousky, V. E., and A. Leetmaa, 1989: The 1986–87 Pacific warm episode: Evolution of oceanic and atmospheric anomaly fields. J. Climate,2, 254–267.

  • Laurent, H., 1993: Wind extraction from Meteosat water vapor channel image data. J. Appl. Meteor.,32, 1124–1133.

  • Lindzen, R. S., and A. Y. Hou, 1988: Hadley circulations for zonally averaged heating centered off the equator. J. Atmos. Sci.,45, 2416–2427.

  • Menzel, W. P., and J. F. W. Purdom, 1994: Introducing GOES-I: The first of a new generation of geostationary operational environmental satellites. Bull. Amer. Meteor. Soc.,75, 757–780.

  • Menzel, W. P., W. L. Smith, and T. R. Stewart, 1983: Improved cloud motion wind vector and altitude assignment using VAS. J. Climate Appl. Meteor.,22, 377–384.

  • Merrill, R. T., 1989: Advances in the automated production of wind estimates from geostationary satellite imagery. Preprints, Fourth Conf. on Satellite Meteorology and Oceanography, San Diego, CA, Amer. Meteor. Soc., 246–249.

  • Merrill, R. T., W. P. Menzel, W. Baker, J. Lynch, and E. Legg, 1991: A report on the recent demonstration of NOAA’s upgraded capability to derive cloud motion satellite winds. Bull. Amer. Meteor. Soc.,72, 372–376.

  • Montgomery, H. E., and L. W. Uccellini, 1985: VAS demonstration:(VISSR Atmospheric Sounder) description and final report. NASA Reference Publication 1151, 198 pp. [Available from NASA Center for Aerospace Information, P.O. Box 8757, Baltimore/Washington International Airport, Baltimore, MD 21240.].

  • Nieman, S. J., W. P. Menzel, C. M. Hayden, D. Gray, S. T. Wanzong, C. S. Velden, and J. Daniels, 1997: Fully automated cloud-drift winds in NESDIS operations. Bull. Amer. Meteor. Soc.,78, 1121–1133.

  • Oort, A. H., and E. M. Rasmusson, 1970: The annual variation of the monthly mean meridional circulation. Mon. Wea. Rev.,98, 423–442.

  • Peixoto, J. P., and A. H. Oort, 1996: The climatology of relative humidity in the atmosphere. J. Climate,9, 3443–3463.

  • Picon, L., and M. Desbois, 1990: Relation between METEOSAT water vapor radiance fields and large scale tropical circulation features. J. Climate,3, 865–876.

  • Rind, D., E. W. Chiou, W. Chu, J. Larsen, S. Oltmans, J. Lerner, M. P. McCormick, and L. McMaster, 1991: Positive water vapour feedback in climate models confirmed by satellite data. Nature,349, 500–503.

  • Ropelewski, C. F., 1988: The global climate for June–August 1988:A swing to the positive phase of the Southern Oscillation, drought in the United States, and abundant rain in the monsoon areas. J. Climate,1, 1153–1177.

  • Salathe, E. P., and D. L. Hartmann, 1997: A trajectory analysis of tropical upper-tropospheric moisture and convection. J. Climate,19, 2533–2547.

  • Schmetz, J., and O. M. Turpeinen, 1988: Estimation of the upper tropospheric relative humidity field from METEOSAT water vapor image data. J. Appl. Meteor.,27, 889–899.

  • Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gaertner, A. Koch, and L. van de Berg, 1993: Operational cloud-motion winds from Meteosat infrared images. J. Appl. Meteor.,32, 1206–1225.

  • Schmetz, J., C. Geijo, W. P. Menzel, K. Strabala, L. van de Berg, K. Holmlund, and S. Tjemkes, 1995a: Satellite observations of upper tropospheric relative humidity, clouds and wind field divergence. Contrib. Atmos. Phys.,68, 345–357.

  • Schmetz, J., and Coauthors, 1995b: Monthly mean large-scale analyses of upper-tropospheric humidity and wind field divergence derived from three geostationary satellites. Bull. Amer. Meteor. Soc.,76, 1578–1584.

  • Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc.,74, 2331–2342.

  • Schwartz, B. E., and C. A. Doswell III, 1991: North American rawinsonde observations: Problems, concerns, and a call to action. Bull. Amer. Meteor. Soc.,72, 1885–1896.

  • Soden, B. J., 1998: Tracking upper tropospheric water vapor radiances from satellite. J. Geophys. Res.,103, 17 069–17 081.

  • Soden, B. J., and F. P. Bretherton, 1993: Upper tropospheric relative humidity from the GOES 6.7-μm channel: Method and climatology for July 1987. J. Geophys. Res.,96, 16 669–16 688.

  • Soden, B. J., and F. P. Bretherton, 1994: Evaluation of water vapor distribution in general circulation models using satellite observations. J. Geophys. Res.,99, 1187–1210.

  • Soden, B. J., and F. P. Bretherton, 1996: Interpretation of TOVS water vapor radiances in terms of layer-average relative humidities: Method and climatology for the upper, middle, and lower troposphere. J. Geophys. Res.,101, 9333–9343.

  • Soden, B. J., and J. R. Lanzante, 1996: An assessment of satellite and radiosonde climatologies of upper-tropospheric water vapor. J. Climate,9, 1235–1250.

  • Spencer, R. W., and W. D. Braswell, 1997: How dry is the tropical free troposphere? Implications for global warming theory. Bull. Amer. Meteor. Soc.,78, 1097–1106.

  • Stephens, G. L., D. L. Jackson, and I. Wittmeyer, 1996: Global observations of upper-tropospheric water vapor derived from TOVS radiance data. J. Climate,9, 305–326.

  • Stewart, T. R., C. M. Hayden, and W. L. Smith, 1985: A note on water-vapor wind tracking using VAS data on McIDAS. Bull. Amer. Meteor. Soc.,66, 1111–1115.

  • Sun, D. Z., and R. S. Lindzen, 1993: Distribution of tropical tropospheric water vapor. J. Atmos. Sci.,50, 1643–1660.

  • Trenberth, K. E., 1997: Short-term climate variations: Recent accomplishments and issues for future progress. Bull. Amer. Meteor. Soc.,78, 1081–1096.

  • van de Berg, L., A. Pyomjamsri, and J. Schmetz, 1991: Monthly mean upper tropospheric humidities in cloud-free areas from METEOSAT observations. Int. J. Climatol.,11, 819–826.

  • Velden, C. S., C. M. Hayden, S. J. Nieman, W. P. Menzel, S. Wanzong, and J. S. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc.,78, 173–195.

  • Velden, C. S., T. L. Olander, and S. Wanzong, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part I: Dataset methodology, description, and case analysis. Mon. Wea. Rev.,126, 1202–1218.

  • Weinreb, M. P., H. E. Fleming, L. M. McMillin, and A. C. Nuendorffer, 1981: Transmittances for the TIROS Operational Vertical Sounder. NOAA Tech. Rep. TR NESS 85, NOAA, Washington, DC, 60 pp. [NTIS PB82-155920.].

  • Weldon, R. B., and S. J. Holmes, 1991: Water vapor imagery: Interpretation and applications to weather analysis and forecasting. NOAA Tech. Rep. NESDIS 67, 213 pp. [Available from NOAA/NESDIS, 5200 Auth Rd., Washington, DC 20233.].

  • Young, J. T., J. W. Hagens, and D. M. Wade, 1995: GOES Pathfinder product generation system. Preprints, Ninth Conf. on Applied Climatology, Dallas, TX, Amer. Meteor. Soc., 176–180.

  • View in gallery

    WVTI from (a) NCEP–NCAR Reanalysis at 400 hPa and (b) GOES VAS for 1200 UTC on 14 Jun 1988. The GOES VAS data are shown with streamlines in blue. WVTI units are g kg−1 m s−1.

  • View in gallery

    The MAW algorithm feature-identification and -tracking procedure. A sequence of three images is used to determine image displacements used for estimates of the wind.

  • View in gallery

    An example of products derived from GOES data on 14 Jun 1988: (a) water vapor–based winds (knots), unedited; (b) edited and cloud-filtered winds interpolated to a grid (knots); (c) upper-tropospheric relative humidity (%) derived at wind vector locations and interpolated to a grid; and (d) assigned pressure (hPa) of the winds and humidity. In (a), the red wind flags are those with large accelerations or directional deviations between vector pairs, and blue wind flags correspond to winds in cloudy regions.

  • View in gallery

    Relationship between the logarithm of layer relative humidity [left side of (4)] and the GOES water vapor brightness temperature (simulated from the NCEP–NCAR Reanalysis Jan 1988 monthly mean data). Red dots represent points where satellite zenith angle θ is greater than 75°. Least squares fit line and regression statistics displayed in upper right corner of figure are computed from θ less than 75°.

  • View in gallery

    Scatterplot of relative humidity showing the bias that results from the use of retrieval coefficients not representative of the retrieval period. (a) Jul coefficients applied to data from 5 Jan 1988; (b) Jan coefficients applied to data from 14 Jun 1988. Deviations from the one-to-one fit line represent biases associated with applying the inappropriate monthly coefficients.

  • View in gallery

    Water vapor weighting functions for different moisture profile configurations: (top) AFGL tropical;(middle) AFGL tropical but with relative humidity configured as in Profile A; (bottom) AFGL tropical but with relative humidity configured as in Profile B. Note that for nearly the same absolute moisture amount q the shape and peak of the weighting functions change because of the different vertical distribution of moisture.

  • View in gallery

    Upper-level specific humidity q and wind speed SPD derived from (a), (c) GOES and (b), (d) NCEP–NCAR Reanalysis data for 14 Jun 1988 at 1200 UTC. Model data are at 400 hPa. Units of specific humidity are 10−1 g kg−1 and wind speeds are m s−1.

  • View in gallery

    GOES-derived WVTI for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are g kg−1 m s−1.

  • View in gallery

    GOES-derived mean meridional transport of specific humidity for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are g kg−1 m s−1.

  • View in gallery

    Mean monthly pressure assigned to the layer wind and humidity fields for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are hPa.

  • View in gallery

    Jun 1988 mean vertical velocity estimates from NCEP–NCAR Reanalysis at 400 hPa. Units are 10−2 Pa s−1.

  • View in gallery

    Zonally averaged (30°–120°W) (a), (b) WVTI and (c), (d) meridional transport of specific humidity for Jun, Jul, and Aug 1987 and 1988. Units are g kg−1 m s−1.

  • View in gallery

    Layer estimates of WVTI and derived from the NCEP–NCAR Reanalysis monthly mean data using three different methods: (a), (b) simulated brightness temperature; (c), (d) observed brightness temperature; (e), (f) full weighting function approach. Units are g kg−1 m s−1.

  • View in gallery

    WVTI simulated from the NCEP–NCAR Reanalysis monthly mean temperature and moisture profile for Jun, Jul, and Aug of 1987 and 1988. Units are g kg−1 m s−1.

  • View in gallery

    Meridional transport of moisture simulated from the NCEP–NCAR reanalysis monthly mean temperature and moisture profile for Jun, Jul, and Aug of 1987 and 1988. Units are g kg−1 m s−1.

  • View in gallery

    Fig. A1. Random wind error associated with size of tracking template. The bold lines correspond to unedited wind data and the thin lines correspond to edited winds using a 5 m s−1 acceleration threshold. The medium thickness lines with values read from the right axis show the reduction in good wind vectors because of the editing.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 8 8 0
PDF Downloads 6 6 0

A Satellite-Derived Upper-Tropospheric Water Vapor Transport Index for Climate Studies

View More View Less
  • a NASA/Global Hydrology and Climate Center, Huntsville, Alabama
  • b University of Alabama in Huntsville, Huntsville, Alabama
  • c Lockheed Martin Corporation, Huntsville, Alabama
© Get Permissions
Full access

Abstract

A new approach is presented to quantify upper-level moisture transport from geostationary satellite data. Daily time sequences of Geostationary Operational Environmental Satellite GOES-7 water vapor imagery were used to produce estimates of winds and water vapor mixing ratio in the cloud-free region of the upper troposphere sensed by the 6.7-μm water vapor channel. The winds and mixing ratio values were gridded and then combined to produce a parameter called the water vapor transport index (WVTI), which represents the magnitude of the two-dimensional transport of water vapor in the upper troposphere. Daily grids of WVTI, meridional moisture transport, mixing ratio, pressure, and other associated parameters were averaged to produce monthly fields for June, July, and August (JJA) of 1987 and 1988 over the Americas and surrounding oceanic regions. The WVTI was used to compare upper-tropospheric moisture transport between the summers of 1987 and 1988, contrasting the latter part of the 1986/87 El Niño event and the La Niña period of 1988. A similar product derived from the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) 40-Year Reanalysis Project was used to help to validate the index. Although the goal of this research was to describe the formulation and utility of the WVTI, considerable insight was obtained into the interannual variability of upper-level water vapor transport.

Both datasets showed large upper-level water vapor transport associated with synoptic features over the Americas and with outflow from tropical convective systems. Minimal transport occurred over tropical and subtropical high pressure regions where winds were light. Index values from NCEP–NCAR were 2–3 times larger than that determined from GOES. This difference resulted from large zonal wind differences and an apparent overestimate of upper-tropospheric moisture in the reanalysis model.

A comparison of the satellite-derived monthly values between the summers of 1987 and 1988 provided some insight into the impact of the ENSO event on upper-level moisture and its transport during the period. During July 1987, a large portion of the Tropics in the eastern Pacific Ocean and Caribbean Sea was dominated by strong vapor transport in excess of 4.0 g kg−1 m s−1, with relatively small amounts in the other months. JJA 1988 transport values reached similar magnitude and showed similar patterns for all three months. The meridional transport of upper-level water vapor indicated large poleward transport from the Tropics to the higher latitudes. This transport favored the Southern Hemisphere, with large transport occurring south of the ITCZ, which extended across the eastern Pacific and northern South America. Zonally averaged monthly transport values were shown to provide a simple way to quantify the monthly and interannual changes in water vapor transport. Zonally averaged WVTI values peaked in the Southern Hemisphere subtropics during both austral winters. In the Tropics, a single, more-pronounced peak located over the equator and south latitudes occurred in 1988 as opposed to a dual peak in 1987. The second peak around 20°N latitude is consistent with findings of others in which upper-tropospheric winds were noted to be stronger in this region during warm ENSO events. Zonally averaged meridional transport was southward for all summer months and was stronger in 1988. The asymmetric nature of the zonally averaged meridional transport (more southerly water vapor transport) was enhanced during JJA 1988, thus indicating a stronger upper-level branch of the Hadley circulation during this notably strong La Niña period.

Corresponding author address: Gary J. Jedlovec, Global Hydrology and Climate Center, 977 Explorer Blvd., Huntsville, AL 35806.

gary.jedlovec@msfc.nasa.gov

Abstract

A new approach is presented to quantify upper-level moisture transport from geostationary satellite data. Daily time sequences of Geostationary Operational Environmental Satellite GOES-7 water vapor imagery were used to produce estimates of winds and water vapor mixing ratio in the cloud-free region of the upper troposphere sensed by the 6.7-μm water vapor channel. The winds and mixing ratio values were gridded and then combined to produce a parameter called the water vapor transport index (WVTI), which represents the magnitude of the two-dimensional transport of water vapor in the upper troposphere. Daily grids of WVTI, meridional moisture transport, mixing ratio, pressure, and other associated parameters were averaged to produce monthly fields for June, July, and August (JJA) of 1987 and 1988 over the Americas and surrounding oceanic regions. The WVTI was used to compare upper-tropospheric moisture transport between the summers of 1987 and 1988, contrasting the latter part of the 1986/87 El Niño event and the La Niña period of 1988. A similar product derived from the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) 40-Year Reanalysis Project was used to help to validate the index. Although the goal of this research was to describe the formulation and utility of the WVTI, considerable insight was obtained into the interannual variability of upper-level water vapor transport.

Both datasets showed large upper-level water vapor transport associated with synoptic features over the Americas and with outflow from tropical convective systems. Minimal transport occurred over tropical and subtropical high pressure regions where winds were light. Index values from NCEP–NCAR were 2–3 times larger than that determined from GOES. This difference resulted from large zonal wind differences and an apparent overestimate of upper-tropospheric moisture in the reanalysis model.

A comparison of the satellite-derived monthly values between the summers of 1987 and 1988 provided some insight into the impact of the ENSO event on upper-level moisture and its transport during the period. During July 1987, a large portion of the Tropics in the eastern Pacific Ocean and Caribbean Sea was dominated by strong vapor transport in excess of 4.0 g kg−1 m s−1, with relatively small amounts in the other months. JJA 1988 transport values reached similar magnitude and showed similar patterns for all three months. The meridional transport of upper-level water vapor indicated large poleward transport from the Tropics to the higher latitudes. This transport favored the Southern Hemisphere, with large transport occurring south of the ITCZ, which extended across the eastern Pacific and northern South America. Zonally averaged monthly transport values were shown to provide a simple way to quantify the monthly and interannual changes in water vapor transport. Zonally averaged WVTI values peaked in the Southern Hemisphere subtropics during both austral winters. In the Tropics, a single, more-pronounced peak located over the equator and south latitudes occurred in 1988 as opposed to a dual peak in 1987. The second peak around 20°N latitude is consistent with findings of others in which upper-tropospheric winds were noted to be stronger in this region during warm ENSO events. Zonally averaged meridional transport was southward for all summer months and was stronger in 1988. The asymmetric nature of the zonally averaged meridional transport (more southerly water vapor transport) was enhanced during JJA 1988, thus indicating a stronger upper-level branch of the Hadley circulation during this notably strong La Niña period.

Corresponding author address: Gary J. Jedlovec, Global Hydrology and Climate Center, 977 Explorer Blvd., Huntsville, AL 35806.

gary.jedlovec@msfc.nasa.gov

Introduction

Over the last few years there has been increasing interest in measuring upper-level moisture from satellite observations. This interest originates from the inability to monitor upper-level water vapor accurately with the existing upper-air rawinsonde network and the need to understand better the role of upper-level water vapor in climate processes. Although the global rawinsonde network provides in situ moisture measurements at twice-daily intervals, the coverage is only regional, because the observations for the most part are confined to populated land areas. Upper-level moisture measurements from rawinsondes also are limited in their use for climate studies, because of the difficulty they have in measuring small amounts of moisture at relatively cold temperatures, and because the specific instrumentation used to measure moisture varies (as does its accuracy) throughout the world (Elliot and Gaffen 1991; Schwartz and Doswell 1991). Satellite instruments, equipped with appropriate channels that are sensitive to atmospheric water vapor, have the potential to provide valuable information on the water vapor variability over the globe. The twice-a-day coverage from many polar-orbit satellites gives daily snapshots of the global structure of upper-level water vapor, while geostationary observations capture the small-scale movement of water vapor and cloud features.

Water vapor is the primary greenhouse gas and, along with clouds, exerts a major influence on the energy balance of the earth–atmosphere system. Small changes in upper-level moisture can produce large changes in satellite brightness temperatures or in the outgoing longwave radiation, which may equal or exceed the effects caused by other atmospheric constituents or processes (Blackwell and McGuirk 1996; Spencer and Braswell 1997; Sun and Lindzen 1993). Perturbations in the magnitude and distribution of global upper-level water vapor caused by short-term climate events such as the El Niño–Southern Oscillation (ENSO) or long-term warming from anthropogenic sources must be studied to ascertain their cause and effect on the earth’s hydrologic cycle and corresponding long-term climate response (Trenberth 1997).

Numerous investigators have used satellite data to monitor upper-level water vapor and its role in climate processes (e.g., Rind et al. 1991; Del Genio 1994). Soden and Lanzante (1996) showed that, despite good agreement of regional variations in upper-tropospheric relative humidity, rawinsonde-derived values are more moist than satellite-derived ones and may affect satellite clear-sky water vapor climate descriptions by as much as 10%. Blackwell and McGuirk (1996) suggest that subtropopause moisture can greatly affect water vapor channel radiances and go undetected in rawinsonde or outgoing longwave radiation measurements. Spencer and Braswell (1997) have used satellite data to hypothesize that the free troposphere actually may be drier than earlier indications suggested.

Several studies have shown that High-Resolution Infrared Radiation Sounder 6.7-μm data can be used to infer upper-tropospheric humidity and changes in the Hadley cell circulation associated with ENSO events. Bates et al. (1996) indicate that anomalously high amounts of tropical upper-tropospheric humidity are associated with convection, and lower amounts are associated with convection-free regions during ENSO warm events. These dry regions correspond to the subsiding branches of tropical circulation cells. Stephens et al. (1996) document seasonal changes in upper-level humidity over the Tropics, which may be related to seasonal swings in the Hadley cell circulation. Outside the Tropics, they find that the Southern Hemisphere winter is significantly drier than the Northern Hemisphere winter, which is consistent with the greater extent of subtropical high pressure regions in the Southern Hemisphere. This relationship between upper-level humidity and tropical circulation systems has been confirmed with geostationary satellite data by van de Berg et al. (1991) and with modeling and trajectory studies by Salathe and Hartmann (1997) and Soden (1998).

The use of geostationary water vapor imagery has allowed the determination of both upper-level moisture content and the wind fields that correspond to the water vapor layers, with numerous researchers concluding that upper-level humidity is related intimately to cloud and divergence patterns (Soden and Bretherton 1993; Schmetz et al. 1995a,b), and that these patterns reflect the large-scale circulation (Schmetz et al. 1995a,b; van de Berg et al. 1991; Picon and Desbois 1990). Schmetz et al. (1995a) and Soden (1998) have used water vapor and wind data to investigate the sources and sinks of moisture to particular regions. Most of these studies have used a month or a season of data to show the feasibility of using geostationary satellite data in climate analysis. Despite these efforts, there continues to be a need to develop a long-term climatological description of upper-level moisture and winds that can be used to study quantitatively the upper-level water vapor variability and its transport in association with regional-scale and global climate processes.

The focus of this paper is to describe a technique that can be used to develop a long-term dataset that quantitatively describes upper-level moisture variability and its transport over large hemispheric regions by using geostationary satellite data. The method combines water vapor winds and upper-level humidity derived from Geostationary Operational Environmental Satellite (GOES) Visible Infrared Spin-Scan Radiometer Atmospheric Sounder (VAS) data to contrast water vapor transport between the summers of 1987 and 1988. This period captures a large swing from a warm ENSO event (Arkin 1988) to a notably strong positive phase of the Southern Oscillation (Ropelewski 1988). Schmetz et al. (1995b), Velden et al. (1997), and Soden (1998) have used 6.7-μm satellite images from geostationary satellites to derive upper-level humidity and winds to evaluate climate variability. The research described in this paper builds on these past approaches and develops a water vapor transport index (WVTI), which quantifies the magnitude of upper-level water vapor transport over the hemisphere region. This paper defines the index, describes how it is calculated from geostationary satellite data, presents a preliminary validation of the product, and describes its utility in quantifying upper-level water vapor transport, its variability, and its importance in climate processes. An in-depth analysis of daily, monthly, and seasonal variations in upper-level water vapor transport for the entire GOES Pathfinder period is beyond the scope of this paper.

Water vapor transport

The formulation of an expression for water vapor transport comes from the atmospheric water vapor budget equation for the vertical column of the atmosphere
QtEPD,
where Qt is the time derivative of precipitable water (the vertically integrated specific humidity), or the water vapor tendency; E is a source of atmospheric moisture from evaporation (of moisture from the surface, from precipitation, or from clouds); P is a sink of atmospheric water vapor as a result of condensation and precipitation; and D is the moisture flux divergence (which can be either a source or sink of moisture). In a closed system with no horizontal boundaries, such as the global atmospheric environment, there is no net lateral transport into or out of the system. Therefore the change in precipitable water vapor is the result of water vapor transport across the lower boundary (evaporation of surface water) and water vapor phase changes in the atmosphere (condensation or evaporation). In a limited atmospheric volume (one with lateral and vertical boundaries), however, there is transport of water vapor into and out of the volume, and the flux divergence term is expressed as
D3qV3
where q is the specific humidity, V is the wind vector, and (qV)3 is the three-dimensional moisture flux, or transport of water vapor. If calculated over the entire vertical column, this transport has only a horizontal component. While it is desirable to obtain this information at various levels of the atmosphere as can be done with a numerical model, passive satellite measurements are limited to providing information in (a few) thick layers of the atmosphere as given by the water vapor weighting function. If a layer L of the atmosphere, centered at a given height with a given thickness, and the two-dimensional transport of water vapor in this layer are considered, then (qV)L represents the water vapor transport of the two-dimensional (layer) flux divergence of water vapor,
LqVL
This transport, or two-dimensional flux divergence, occurs exclusively over a layer in the atmosphere. An example of hemispherical water vapor transport (|qV|) from the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) 40-Year Reanalysis Project (Kalnay et al. 1996) is presented in Fig. 1a. The data are valid at 1200 UTC 14 June 1988 at 400 hPa. A similar product derived from GOES data (described below) is shown in Fig. 1b. Both fields are superposed on the corresponding GOES VAS 6.7-μm 1200 UTC water vapor image. These images indicate relatively dry (dark) and moist (relatively bright) regions of the upper troposphere. The contours indicate large transport values associated with tropical convective regions and synoptic weather systems in the midlatitudes of both hemispheres. These are regions where both high humidity and strong winds are present in the reanalysis fields for this time. The utility of the model-diagnosed upper-level water vapor transport is subject to scrutiny since the fields are based on upper-level moisture and winds that may be poorly represented in these models (Soden and Bretherton 1994; Spencer and Braswell 1997). Differences between the model and GOES transport products exist. In this case, the model values are for a single model level (400 hPa), while the GOES data represent a layer of information that varies in its vertical position. Details of these differences will be discussed in a later section.

A goal of this research was to develop a technique to quantify upper-level water vapor transport within a layer in the upper troposphere, (|qV|)L, with satellite data, which does not have the same limitations that model-derived estimates do. To accomplish this goal, estimates of upper-level humidity and winds from GOES VAS are used to define a quantitative measure of transport (the WVTI) in the layer observed by the satellite. The development of the WVTI from GOES data is described in the next section. Although the GOES VAS application is presented, the approach is applicable to other geostationary satellites with similar capabilities to observe upper-level water vapor and winds. The approach capitalizes on the sensitivity of the GOES 6.7-μm water vapor channel to small amounts of water vapor in the upper layers of the troposphere and on the use of this moisture as a passive tracer of the wind. The 6.7-μm channel on the GOES VAS presents an opportunity to quantify the source and sink regions of upper-level moisture and its spatial, seasonal, and interannual variability over a relatively limited period. The GOES VAS Pathfinder dataset (Young et al. 1995) provides an opportunity to quantify hemispheric fluctuations of upper-level water vapor transport on a daily basis during a period of considerable atmospheric moisture change as a result of the 1986/87 El Niño and 1988 La Niña (Kousky and Leetmaa 1989; Janowiak 1988). Application to a longer period of record may be possible with an extension of this work to the imager on the current GOES satellite or to other geostationary satellite datasets. In addition, comparison of the satellite-derived variations with that portrayed by long-term reanalysis project datasets (Kalnay et al. 1996; Schubert et al. 1993) (as in Fig. 1a) could provide insight into model sensitivities and limitations for monitoring upper-level moisture transport and its variability.

Retrieval methodology

To derive the WVTI from GOES data, satellite measurements of upper-level humidity and winds are required. These parameters need to be consistent in space and time and therefore must come from the same sequence of images. The individual parameters are combined to produce a daily estimate of the upper-tropospheric water vapor transport as indicated by the WVTI. A discussion of these procedures is presented below.

Winds

The standard approach taken in generating wind fields from geostationary satellite data uses a sequence of two or more images to track identifiable image features (determine image displacements). In this paper, the winds are derived from GOES water vapor imagery with the Marshall Automated Wind (MAW) algorithm (Atkinson 1984, 1987) with modifications described below. The algorithm uses a minimum-difference template-matching scheme for feature identification and tracking. In the tracking procedure, the first of a pair of images is divided into image subscenes called templates, while the second image contains subscenes called search areas (Fig. 2). The template (T1) is an array of picture elements, and the spatial location of a template is designated as the template’s center picture element location in the image (i.e., i, j). To determine feature displacement or motion (wind), each template T1i,j in image 1 is translated to all possible positions S2i,j within a corresponding search area (having a specified radius) in the second image to look for the best match. The best match is simply the position of the template T2i+x,j+y in image 2, which, when compared with the template position in image 1, gives the smallest mean difference value. Once the best match is found, its position within the search area (i + x, j + y), along with its initial position in image 1, determines the template displacement between the two images in satellite coordinates. Accurate navigation and registration of the two images allows for the determination of displacement or velocity vectors relative to the earth (u and υ components of the wind). The MAW approach as applied here actually uses a sequence of three images from which two vectors are produced (displacement of feature in image 1 to image 2, and displacement of feature in image 2 to image 3) for the movement of each feature (Fig. 2). The vector pairs V1 and V2 are used in quality checks, as discussed below, and are averaged for mean wind field determination.

When using the MAW tracking scheme, there are several decisions to be made that affect the quality of the resulting motion vectors (winds): radiometric accuracy, spatial and temporal resolution of the images, template size, and search constraints. Experience indicates that the highest-quality winds come from the appropriate match of spatial and temporal resolution. Use of high–temporal resolution data with coarse spatial resolution produces poor winds because the pixel displacement of the feature is small, in which case navigation and registration uncertainties heavily influence the results. On the other hand, poor temporal sampling introduces errors in tracking because of the evolution and subsequent decorrelation of features over time. Merrill (1989) and Schmetz et al. (1993) also discussed the effect of image resolution on the ability to track image features. The GOES Pathfinder dataset used in this study provides hourly full-disk imagery of much of the Western Hemisphere in one visible and three infrared channels over a period from May 1987 to November 1988. The dataset contains the 6.7-μm imagery, which has a nadir resolution of 16 km (nominal) in each direction, based on the use of the VAS large detectors. Each VAS pixel oversampled the scene in the east–west direction by a large amount. The nominal 8 km × 8 km pathfinder dataset was produced by repeating every line in the north–south direction and by using every other element of the oversampled data in the east–west direction. In this way, each pixel represents an approximately 8 km × 7 km footprint (nominal) at satellite nadir (see Montgomery and Uccellini 1985). A 49 × 49 pixel (8 km × 7 km) template (about a 392 km × 350 km area at nadir) was used to track and to match features in the water vapor imagery. This large size was selected based on the relatively coarse image resolution available with the Pathfinder dataset, on the relatively poor radiometric quality of the radiances as compared with more recent instrumentation (Menzel and Purdom 1994), and on the lack of small-scale structure in the cloud-free water vapor data. Jedlovec and Atkinson (1996) have shown that increasing the template size reduces the random noise in the water vapor winds by locating and tracking large-scale features (as discussed in the appendix). Thus, large templates reduce the need for editing of the final wind dataset but produce winds that are representative of only the large-scale environment. In this study, the extent of the search (distance away from the initial template position in image 1 that is searched in image 2 to find a match) was selected to be 31 pixels (about 250 km at nadir) from the original template location in image 1. This conservative approach allows for winds in excess of 70 m s−1 (at satellite subpoint) without any preconceived directionality. These ground sizes (GOES nadir footprint, template size, and search radius) increase with increasing off-nadir viewing angle and can be determined for any location, based on GOES navigation parameters. To a first approximation, this increase varies as the sec2θ (e.g., at 35°N, 115°W, the GOES nadir 16 km × 16 km footprint increases in size to about 28 km × 28 km), where θ is the satellite viewing zenith angle.

An example of the resultant wind product derived with the MAW algorithm in the above way and used in this research is shown in Fig. 3a. A sequence of three hourly images of GOES VAS data (middle time is 1200 UTC) on 14 June 1988 that cover much of the Western Hemisphere was used for this example. Three images allow for the calculation of two vectors corresponding to each feature. The average of the two vectors for each location is plotted in the figure. The initial distribution of wind vectors is uniform because of the simple indexing scheme used to process the image data. The winds show good spatial consistency in many areas. The calculation of two wind vectors per feature is common in both research and operational processing of satellite data (Nieman et al. 1997; Velden et al. 1997; Laurent 1993; Schmetz et al. 1993) and allows for continuity or symmetry checks between the vector pairs. Differences between the magnitude of the winds of vector 1 and vector 2 for a given location that are greater than 15 m s−1 or direction differences greater than 30° are flagged as “bad.” These threshold values were determined based on a comprehensive error analysis, as detailed in the appendix. This filtering approach is different from the operational scheme of the National Environmental Satellite, Data, and Information Service (Nieman et al. 1997; Velden et al. 1998), in which a 5 m s−1 u and υ vector pair threshold and spatial comparisons (i.e., “buddy checks”) are used. The approach here is less restrictive in that it allows large deviations but flags (as bad) vector pairs that show totally different flow characteristics (large direction differences between vector pairs). It does not employ a spatial check. Vector pairs that show considerable agreement (pass the acceleration filters) are averaged together to form a single vector, valid at the middle image time and assigned a location on the earth that is based on the average displacement of the two vectors. In Fig. 3a, both the “good” winds and the bad winds (those that failed the quality control procedures) are shown. It is apparent that the bad winds (red wind flags) are inconsistent spatially with many of their neighbors, which provides confidence in the filtering procedure. These bad vectors occur throughout the image and result from 1) lack of trackable image structure, 2) changes in image structure over the 2-h tracking sequence, or 3) multiple solutions in the matching approach. Further information on approaches to minimize bad vectors is presented in the appendix. There are many good wind vectors (yellow and blue) throughout the image, which show consistent flow patterns associated with outflow from tropical convective systems, strong winds from synoptic systems in both hemispheres, and large-scale winds away from convective regions.

GOES imagery often shows gradients from both upper-lower water vapor and high clouds. Pure water vapor motions can have distinctively different flow characteristics than the cloud features do, and, as discussed below, the upper-tropospheric humidity cannot be determined adequately in the presence of clouds. Accordingly, only winds in cloud-free regions of the GOES imagery were used in this study. Winds in the presence of clouds were identified with a relative humidity threshold of 99% (described in section 3b) and are shown as blue vectors in Fig. 3a for 14 June 1988.

To provide a uniform field of winds representative of the flow over this region for quantitative calculations and subsequent analysis, the cloud-free good winds were interpolated to a 1° × 1° grid with a Barnes objective analysis scheme. The tuning parameters in the scheme were adjusted to preserve as much spatial structure as possible at the smallest resolvable scale (800–1200 km) without generating noisy fields or fictitious signatures. An example of these gridded winds is presented in Fig. 3b. It is readily apparent that the gridded winds provide a uniform distribution of the flow features over the region that is representative of the large-scale characteristics of the wind field. In data-sparse regions, the Barnes scheme uses trends in the surrounding data to fill the voids. Care was taken to ensure that reasonable values were provided in these regions. When the original sequence of images used in the tracking procedure is animated beneath the gridded wind field, it also is apparent that the wind field portrays the major motions observed in the satellite imagery. Gridded wind fields were produced for every day of June, July, and August of 1987 and 1988 with the above methodology and subsequently were used in the quantitative calculations discussed below.

Humidity

The WVTI requires the calculation of specific humidity to represent quantitatively the moisture variations in the upper layers of the atmosphere. Several methods have been developed for retrieving upper-tropospheric humidity (UTH) from water vapor channel satellite measurements. Schmetz and Turpeinen (1988) describe the operational algorithm in use by the European Space Operations Centre for deriving UTH from Meteosat 6.3-μm water vapor channel data. Their physical approach employs radiative transfer calculations with European Centre for Medium-Range Weather Forecasts temperature profile data to generate lookup tables to convert the water vapor radiances to UTH. A modification to the approach (Schmetz et al. 1995a) has been used for synoptic analysis of Meteosat moisture data. Soden and Bretherton [1993; 1996 (referred to subsequently as SB96)] also developed a technique for converting water vapor brightness temperatures to UTH. Simplified radiative transfer theory is used to arrive at their relationship:
i1520-0450-39-1-15-e4
In (4), r is the layer-averaged relative humidity, Po is the normalized pressure [i.e., Po = P(T = 240 K)/300 hPa], A and B are the regression coefficients, and TB is the satellite-measured brightness temperature. This approach was developed originally for GOES-7 VAS water vapor measurements and since has been applied to Television Infrared Observational Satellite Operational Vertical Sounder (e.g., Soden and Bretherton 1996; Stephens et al. 1996) and Special Sensor Microwave humidity sounder (SSM/T2) (Spencer and Braswell 1997) water vapor data. The approach used here is fundamentally the same as that of SB96 but is applied to GOES data with regression coefficients that are derived on a monthly basis to account for temperature and humidity variations not properly represented in a single set of regression coefficients. In addition, the satellite zenith angle is included explicitly in (4) to derive the coefficients.

For the current study, the NCEP–NCAR monthly mean reanalysis (Kalnay et al. 1996) was used to obtain Po and q, and saturation vapor pressure over ice was used to get relative humidity. These Po values are different from SB96, in which climatological estimates were used. Since the reanalysis model pressure-level moisture data extend only up to 300 hPa, relative humidity was allowed to decay exponentially above this level to satisfy the “weighting” criteria in the SB96 technique. The model thermodynamic data serve as input for the forward radiative transfer calculations [transmittances calculated as in Weinreb et al. (1981) with the GOES-7 6.7-μm channel spectral response function values] to generate simulated channel brightness temperatures at each model grid point. In the radiative transfer calculations, the satellite zenith angle (based on a 0°, 75°W satellite subpoint) for each reanalysis model grid point (within the GOES-7 view area of 60°N–60°S, 0°–150°W) was used to calculate a simulated VAS TB under nonnadir conditions. Simulated brightness temperatures and the expression on the left side of (4) were used to investigate the relationship between the natural logarithm of layer relative humidity (RH) and the water vapor brightness temperature developed by SB96. The results of this analysis are presented in Fig. 4. For relatively warm brightness temperatures (TB > 235 K) the relationship is highly linear with minimal scatter. Significant departure from a linear fit is observed for zenith angles greater than 75° (red points). This result seems to be consistent with the other studies (Soden and Bretherton 1993, 1996; Schmetz et al. 1995a; Spencer and Braswell 1997), although specifics are lacking, and consistent with Soden (1998).

The RH slope and intercept coefficients derived from the monthly mean NCEP–NCAR Reanalysis data (2.5° × 2.5°) for 1988 with satellite zenith angles less than 75° using (4) are presented in Table 1. The linear correlation coefficients between TB and the left side of (4) are greater than about 0.95 for all months. Although it may appear that an increase in the intercept coefficient might be compensated for by a decrease in the slope coefficient (as in the boreal summer months), the retrieved humidity is not the same in each case. To illustrate this point, two sets of regression coefficients (January and July 1988) were applied to a single day in both June (14 June 1988) and January (5 January 1988) to demonstrate the effect of variation in coefficients on the humidity retrievals. The results are shown in Fig. 5. A reversal in the bias, associated with using the wrong monthly coefficient, is evident for humidity below about 60%. For example, when using July coefficients on January data (Fig. 5a) in dry regions (<60% RH), the retrieved humidity is less moist (up to 10% absolute difference). The opposite holds true when January coefficients are applied to summer data (Fig. 5b). For relatively moist conditions (>60% RH), there is less bias but more scatter. These figures therefore illustrate that summer coefficients applied to winter data will enhance moisture gradients, while winter coefficients applied to summer data will tend to weaken slightly the humidity gradients. This variation is undesirable in this study, where spatial variations in moisture over the hemispheric region and monitoring of month-to-month changes in regionally averaged RH values are the focus of the research. Therefore, in this study, the set of monthly regression coefficients in Table 1 for June, July, and August was used to retrieve RH from GOES data for the summers of 1987 and 1988. This approach will reduce retrieved relative humidity bias from the use of a single set of coefficients. To limit the regression relationship to input data with satellite zenith angles of less than 75° also is consistent with the wind retrieval procedures.

For the retrieval of average RH (over the layer represented by the water vapor channel weighting function) with GOES VAS data, special data processing methods were used to ensure consistency with the water vapor winds. Since the winds were generated from tracking features in a 49 × 49 pixel template, an average brightness temperature was obtained over this template. This temperature was used to retrieve a layer-averaged RH value using the coefficients in Table 1. If this humidity value exceeded 99%, the template region was assumed to contain clouds, in which case the wind vector was flagged as cloudy and the humidity value was not used. In this way, the two datasets (humidity and wind fields) represent the same scale of features in the cloud-free regions of the imagery. This is unlike Soden (1998) and Soden and Bretherton (1993), who applied the retrieval coefficients to an average of a few cloud-free pixels. The RH values were interpolated to a uniform 1° × 1° grid identical to that of the wind field. An example of the resultant RH product is displayed in Fig. 3c. The derived humidity field shows excellent agreement with the cloud-free regions of the GOES 6.7-μm image and does not capture the detailed structure in the cloudy regions (as expected). The coarse-scale RH field does show gradients in the water vapor field around cloudy regions, especially around the tropical convective regions and midlatitude weather systems. Subsidence zones in the subtropics exhibit very low RH values (below 20%), even in the large-scale analysis. It is believed that this approach accurately quantifies the large-scale patterns and gradients observed over the cloud-free areas of the analysis region.

For quantifying vapor transport and its variability, specific humidity is required. The template-averaged satellite brightness temperature is a good estimate for the mean temperature of the water vapor layer and is used in Teten’s formula to convert relative humidity to absolute or specific humidity, that is, qes(UTH) (Peixoto and Oort 1996). Since most of the water vapor band brightness temperatures are below −10°C, the saturation vapor pressure over ice (es) is used in the conversion.

Layer height assignment

The moisture and wind fields derived above represent the spatially averaged humidity and wind in the layer of the atmosphere sensed by the GOES water vapor channel. The height and depth of this layer change with the temperature and, more important, with the vertical structure of water vapor in the atmosphere (Fischer et al. 1981). The energy measured by the satellite in this channel is emitted from a layer of the atmosphere that contains a constant optical depth (water vapor burden times the water vapor absorption coefficient integrated from the top of the atmosphere to some level below) that is equivalent to about 1–2 mm of integrated vapor content. For regions with a dry upper troposphere, this nearly constant water vapor burden or optical depth is not encountered until somewhere in the midtroposphere. For a moist upper troposphere, this burden is encountered at lower pressures. Satellite viewing geometry also is a factor. When the satellite views an off-nadir point, the slant-path water vapor burden is detected. Thus, for large viewing angles, the water vapor burden is reached at lower pressures than for near-nadir points. Since the WVTI is a layer quantity, the actual height of the humidity and wind does not enter directly into the calculation. However, the height is important for interpreting the change in vertical position of the water vapor transport. In all cases, the water vapor channel weighting function determines the layer of the atmosphere sensed by the satellite. Since the satellite brightness temperature measures the integrated temperature of the emission layer, matching this temperature to an appropriate thermodynamic profile can be used to estimate the height of the humidity and wind layer (assumed to be the position of the peak of the weighting function). This approach is similar to the traditional one suggested by Fritz and Winston (1962), and applied by others to opaque cloud regions (e.g., Schmetz et al. 1993). In this research, the water vapor layer is opaque to upwelling radiation from the surface, and this traditional infrared height assignment method is applied. To determine this height, the NCEP–NCAR Reanalysis data for 1200 UTC with a 5° × 5° spacing provided an estimate of the temperature profile at the wind and humidity location. This temperature profile was used to obtain a pressure that corresponds to the template-averaged brightness temperature at each wind and humidity location. Individual pressure values were interpolated to a uniform grid that is consistent with the wind and humidity fields. An example of the pressure assigned to the wind and humidity on 14 June 1988 at 1200 UTC is presented in Fig. 3d. Note the variability of pressure around cloudy regions and the more gradual changes in the heights in subsidence zones. In the latter areas, mean pressures can reach in excess of 400 hPa, indicating that the water vapor signature occurs more in the mid- rather than upper troposphere.

This height assignment method is not error-free. A basic assumption in this approach is that the pressure corresponds to the peak of the channel weighting function and that the shape of the weighting function is roughly the same from region to region. The latter will be the case only when the vertical structure of moisture is the same. To understand this point better and to assess the range of possible height assignment errors with this approach, numerous extreme profiles of moisture were examined. Figure 6 presents GOES water vapor weighting functions that correspond to several of these profiles. The top diagram shows the Air Force Geophysics Laboratory (AFGL) tropical profile, while the second and third diagrams result from changes made to the vertical structure of moisture in the mid- and upper troposphere. The three thermodynamic profiles yield nearly identical brightness temperatures and a corresponding pressure of 380 hPa with this height assignment technique. Evaluation of the location of the weighting function peak in each diagram will produce three different answers that vary by nearly 100 hPa. This type of variability is not uncommon, as noted by Weldon and Holmes (1991). Therefore, in this study the pressure assigned to each wind and humidity value should be interpreted as only an estimate of the layer height of the feature observed in the GOES water vapor channel imagery.

Results

In this section, results from applying the WVTI retrieval technique to GOES data for June, July, and August of 1987 and 1988 are presented and discussed. Before presentation of some of the results, some of the specific procedures used for the current application are summarized.

Water vapor winds were derived from a sequence of three hourly images in the 6.7-μm water vapor channel of the GOES-7 VAS instrument using data at nominal times of 1100, 1200, and 1300 UTC each day. Between 1 June and 19 July 1987, data at 1400, 1500, and 1600 UTC were used because of changes in the data collection schedule for GOES-7. The MAW algorithm was applied to this data with a 49 × 49 pixel (8 km × 7 km) square template using a 31-pixel search distance. Initial quality control procedures rejected vectors with template minima on the edge of the search area. Final editing included the elimination of vector pair discrepancies of 15 m s−1 in speed and 30° in direction. This automated editing typically reduced the number of wind vectors by about 40%. Water vapor wind vector pairs that passed these criteria were averaged to get a mean wind valid at the middle image time (1200 UTC) and were assigned spatially to the mean location of each individual vector (of the vector pair). Only radiance data with satellite zenith angles of less than 70° were used, to avoid uncertainties with the extremely large satellite fields of view near the edge of the full disk of the earth. Relative humidity was derived from the template-averaged brightness temperature at each good wind location using a modification of the SB96 scheme. If the RH derived from this radiance was in excess of 99%, the scene was deemed to be cloudy and its wind and humidity values were rejected for use in subsequent analyses. Layer RH was converted to specific humidity using the template-averaged brightness temperature as the mean layer temperature in the conversion process. Pressure was assigned to the water vapor winds and specific humidity of the layer using the template-averaged brightness temperature and the NCEP–NCAR Reanalysis dataset at a reduced 5° resolution at 1200 UTC. A Barnes objective analysis scheme was applied to the winds, humidity, and pressure retrievals to produce the 1° × 1° analysis for every day during the summer months of 1987 and 1988. The analysis region extends from 30°S to 45°N latitude and 30° to 120°W longitude to account for frequent missing data south of 30°S and to exclude retrievals for large satellite zenith angles. Grid values of winds and specific humidity then were used to calculate the WVTI (i.e., |qV|).

Application of WVTI to daily data from 14 June 1988

The above procedures have been applied to daily GOES VAS Pathfinder data for 6 months during June–August of 1987 and 1988. For discussion and initial evaluation purposes, the data from a single day are examined. An example of the WVTI for 14 June 1988 is presented in Fig. 1b. The bold red lines correspond to the WVTI. Values range from near 0 where the winds are light to 8–10 g kg−1 m s−1 around midlatitude weather features. A streamline analysis of the water vapor winds is shown in blue. Much water vapor transport is associated with and occurs around synoptic features in both hemispheres. The patterns generally are consistent with that from the NCEP–NCAR Reanalysis 400-hPa data for this day and time (Fig. 1a). Both datasets identify the major transport centers in Canada, the western Atlantic Ocean, and the elongated region straddling northern Argentina and Chile. Upper-level transport around regions of tropical convection to regions of subtropical high pressure also is apparent in both datasets. The datasets depict minimal transport over the continental United States and the southeast Pacific Ocean subtropical high pressure region. The height of the transport layer in the tropical Pacific region (Fig. 3d) indicates that the WVTI represents transport at low pressures near the tropical convection. Away from areas of tropical convection, the transport occurs at higher pressures and corresponds to subsidence regions in both the Tropics and subtropics.

It is instructive to examine these two datasets more closely on this day to understand better the reason for their differences. The reanalysis dataset shows transport that is about 2–3 times stronger than is shown by the satellite data in most regions. This difference occurs for several reasons. First, water vapor transport from the NCEP–NCAR model in Fig. 1a is valid at a single pressure level (400 hPa), while the satellite-derived water vapor transport occurs over a thick layer of the atmosphere that varies in height, as is apparent from the pressure map in Fig. 3d. Second, the transport itself depends on both winds and specific humidity in each dataset. The representation of these two parameters in the satellite and model datasets may be very different (Bony et al. 1997; Spencer and Braswell 1997; Soden and Bretherton 1994). Indeed, the individual fields of humidity and winds from GOES and the NCEP–NCAR model presented in Fig. 7 for 14 June 1988 are very different. Specific humidity q from the satellite-derived product generally is smaller in all regions, ranging from 0.1 g kg−1 in the cold dry upper troposphere of the Southern Hemisphere midlatitudes to greater than 0.4 g kg−1 in the Tropics. Since the NCEP–NCAR data presented are valid at 400 hPa, which typically is lower in the troposphere than the mean GOES wind height, greater values of q can be anticipated. This expectation is confirmed in Fig. 7b, in which q values range from 0.4–2.0 g kg−1. In regions where pressures coincide to within 50 hPa, such as in the subtropical dry zones, the agreement is better, but the NCEP–NCAR Reanalysis still provides somewhat larger values. Note that the moist bias in the reanalysis data might be attributed partially to the fact that the model calculations of specific humidity are done using saturated vapor pressure with respect to water, not ice. The greatest differences between the model- and satellite-derived values are observed in the Tropics, where the difference in height assigned to the humidity values can be as large as 100 hPa and the resultant specific humidity difference is in excess of 1.0 g kg−1. The magnitude of the 400-hPa wind tends to be in better agreement with GOES. Spatial contour maps (Figs. 7c,d) agree well in the Southern Hemisphere and in the Northern Hemisphere extratropics. The NCEP–NCAR winds do not achieve the same magnitude in the Southern Hemisphere extratropics as do the satellite-derived data, which may be more capable of depicting synoptic flow features such as jet streaks and other divergent circulations in data-sparse regions. Use of 300-hPa reanalysis data improves the moisture agreement in this region, but the overall agreement in the fields is reduced as a result of the differences in the winds (not shown).

Monthly transport for summer 1987 and 1988

It is apparent from the daily analysis of the WVTI that the GOES VAS data are very useful for delineating regions of upper-tropospheric water vapor transport and its spatial variation. It is appropriate to examine its ability to characterize the spatial and temporal variability of water vapor transport in terms of monthly averages. Figure 8 shows the monthly mean WVTI for June, July, and August (JJA) of 1987 and 1988 over most of the Western Hemisphere as observed by the GOES-7 satellite. These monthly fields represent an average of daily grids. The WVTI shows weak horizontal transport associated with subtropical high pressure zones and strong transport that corresponds to synoptic disturbances at higher latitudes and around tropical convection. The magnitude of this variation can be very large, with values ranging from less than 1.0 g kg−1 m s−1 over the western United States and Bermuda high pressure region to 4.0–6.0 g kg−1 m s−1 in the Southern Hemisphere midlatitudes and around active areas of tropical convection. While many of the large-scale transport patterns show continuity from month to month (such as the maximum located east of the northeastern United States and a minimum in the Bermuda high pressure region), a detailed comparison of the index values indicates large changes in the magnitude of the water vapor transport during the individual summer months and between 1987 and 1988.

In July 1987, large values (>4.0 g kg−1 m s−1) are observed across the western Atlantic, Caribbean Sea, and tropical eastern Pacific. In 1988, large values are present in these regions during all three months. During some of the months there exists a preferred region for strong tropical transport just south of the ITCZ in the eastern tropical Pacific (0°–10°N, 75°–110°W). This preference is a direct result of the superposition of relatively large values of moisture in the upper troposphere and, more important, persistently high wind speeds. Transient vapor transport regions are identifiable in maps of WVTI that may not be discernible as easily in other fields such as the meridional transport of specific humidity (). For example, July 1988 is characterized by relatively weak transport over most of the study domain with relatively flat gradients in vapor transport. As will be discussed shortly, inspection of just the monthly mean meridional transport fields indicates meridional fluxes of the same magnitude in July and August 1988 (Fig. 9) when, in fact, more upper-tropospheric water vapor transport (WVTI in Fig. 8) occurs in August 1988. This discrepancy is due to the substantial zonal transport of moisture during this period. Another transient feature is the presence of a region of relatively strong transport that extends into the Gulf of Mexico during June 1988 (Fig. 8b). This feature is not as prominent in the corresponding map of (Fig. 9b), partly because the monthly mean flow in this region is more zonal (u wind component not shown).

The WVTI also is useful in noting the intraseasonal changes in flow patterns as is the case of the maturing austral winter in the Southern Hemisphere extratropics. Amid the large number of contrasts between the summers of 1987 and 1988, the WVTI over the South Pacific during June and August 1988 is notably weaker than in July 1988. Another interesting feature that occurs in both 1987 and 1988 is the transport maximum during June observed over southern Brazil and northern Argentina. The interannual variations between 1987 and 1988 in other regions (e.g., the Tropics), though, are numerous. The local minimum of transport over the Andes Mountains in Chile is caused in part by the topography of this physical barrier to the flow. In addition, the Andes present a difficulty in retrieving wind vectors for all satellite-derived wind algorithms because the mountains protrude into the middle troposphere. During dry upper-tropospheric conditions (as is often the case over the Andes Mountains region), the amount of water vapor above the mountains is small; thus the water vapor channel actually can detect the mountain surface. Good wind vectors in this region usually are obtained only when water vapor structure or high clouds occur above 300 hPa. Manual editing of the winds was carried out for the daily data such that the daily gridpoint values are not influenced by the mountains.

Although the WVTI indicates the magnitude of transport, it does not provide directional information. Decomposing the satellite-derived water vapor winds into u and υ components allows for the determination of zonal and meridional transport of upper-level moisture (Fig. 9). Consistent with the findings of Lindzen and Hou (1988) and Oort and Rasmusson (1970), the upper branch of the Hadley circulation favors cross-equatorial transport such that the outflow patterns from the regions of strongest diabatic heating associated with the ITCZ near 10°–15°N are directed southward. More of the poleward-transported upper-level tropical moisture expelled from deep convection along the ITCZ is directed toward the Southern Hemisphere, as is evident over northern South America and, especially, in the eastern tropical Pacific in the mean monthly maps of (Fig. 9). In fact, based on just the analysis of the meridional flux of vapor fields (i.e., no vertical velocity, precipitation, or lower-level wind field information), only a single Hadley circulation cell is evident in the viewing domain. This asymmetrical flux of moisture continues for JJA of both 1987 and 1988. The strongest southward transport generally is observed in the region just to the south of the most active convection (as determined by the lowest mean height assignment). June 1988 indicates an exceptionally strong moisture flux both in the Tropics and midlatitudes (Figs. 8b and 9b). Despite the relatively weaker Northern Hemisphere extratropical circulation features that normally occur during the boreal summer, long-wave ridge–trough patterns are discernible over North America and the North Atlantic (especially during June 1988). The smaller amplitude waves in the Southern Hemisphere extratropics can be viewed in terms of the weak (<1.0 g kg−1 m s−1) despite the exceptionally strong WVTI (>4.0 g kg−1 m s−1) caused by the strong zonal wind component during August 1987 and August 1988. With the exception of the southward maximum of in the tropical eastern Pacific, the interannual and intraseasonal patterns in meridional transport vary considerably, such that no discernible dominant feature persists from month to month or year to year.

Figure 10 presents the mean monthly pressure associated with the wind and humidity fields based on the height assignment method described in section 3c. These contour lines approximate the mean height of the layer that contributes to the water vapor channel measurements over a monthly period. Its variation in time and space is the result of changes in mid- to upper-tropospheric humidity and temperature. These spatial variations are consistent with other meteorological fields (e.g., circulation patterns and corresponding brightness temperature, not shown). Large subtropical high pressure regions, such as the Bermuda and southeastern Pacific high pressure regions, are represented by weaker gradients in pressure height. These areas tend to coincide with regions of low but not necessarily with low WVTI, as is apparent in Fig. 8c, over the western Atlantic. A semipermanent horizontal orientation of isobars between 10°S and 10°N located west of 75°W marks the strongest gradients in moisture observed between the ITCZ and the strong southeastern Pacific subsidence zone. Imagery verifies that this region is characterized by gradually increasing brightness temperatures south of the pressure height minimum (∼10°N) adjacent to the most intense convective regions. The NCEP–NCAR Reanalysis–estimated vertical velocities in Fig. 11 for June 1988 at 400 hPa corroborate the information from the pressure height and meridional transport fields discussed above: negative vertical velocities coincide with the outflow from the tropical convection as it descends southward toward the tropical eastern Pacific subsidence zone. The interannual variation in pressure height assignment sheds some interesting information on the “dryness” of the upper troposphere. During JJA 1988, the assigned pressures are greater on average than in 1987. This result implies that less water vapor is present at upper levels and outgoing longwave emissions are increased. This situation particularly was the case in July 1988 when pressure heights greater than 320 hPa dominated the Bermuda high pressure region of the Atlantic. More upper-tropospheric moisture distinguishable by the lower pressures in July 1987 is related to the strong water vapor transport (Fig. 8c) observed in this region. That is, the magnitude of water vapor transport seems to be influenced more by UTH in the Tropics than by the winds in the extratropics.

Zonally averaged quantities

Zonally averaged quantities of vapor transport provide a simple way to view and to quantify both annual and interannual variability in the water vapor fluxes and to mark the seasonal cycle and magnitude of tropical convection and extratropical circulation. The WVTI and meridional transport of specific humidity averaged over 30°–120°W at every 1° latitude for JJA 1987 and 1988 are shown in Fig. 12. The zonally averaged WVTI [panels (a) and (b)] represents the latitudinal distribution of the total water vapor transport in a layer of the upper troposphere represented by the GOES water vapor weighting function. There exist three distinct peaks: one near the equator and two at the study domain boundaries (i.e., near 45°N and 30°S). During the austral winter, the dominant peak for both years is found in the Southern Hemisphere subtropics. The Northern Hemisphere peak during the summer months is attributed to weak and moderate upper-level disturbances that track mostly across the eastern half of the United States and the western North Atlantic Ocean.

There is considerable interannual variation in the tropical zonally averaged peak of the WVTI. In 1988, a single, more pronounced peak centered near the equator is indicative of the strong vapor transport just south of the ITCZ. In 1987, the tropical transport maximum is broader, with dual maximum in transport at 5° and 15°N. Indeed, analysis of the spatial maps of the individual months in Fig. 8 reveals strong zonal moisture transport in the western Atlantic and Caribbean Sea (as previously noted) that was not present during JJA 1988. This result is consistent with recent observations by Bell and Halpert (1998) for which upper-tropospheric winds tend to be stronger in the Tropics and subtropics during warm ENSO events [such as was observed during 1987 (Janowiak 1988)]. It is believed that these anomalous winds tend to suppress tropical storm activity in the Atlantic and Caribbean since the wind shear and convergence patterns are less favorable for tropical storm development (Gray 1984).

Since upper-level humidity at high latitudes is smaller than in the Tropics, the midlatitude peaks tend to reflect the relatively stronger winds associated with extratropical synoptic disturbances. The WVTI is useful to distinguish subtropical jet features that may not be apparent in the meridional flux calculations. For example, an anomalous zonally averaged WVTI peak near 25°N during June of 1988 is caused by the strong zonal flow in the Gulf of Mexico [which, of course, is not discernible in the zonally averaged meridional fluxes (Fig. 9d)].

The zonally averaged meridional fluxes (Figs. 12c,d) illustrate the asymmetry about the equator. A sharp gradient is observed from 15° to 5°N for both summers, which is indicative of the relatively strong southward vapor transport south of the ITCZ. The inflection point in the sharp transition from 5° to 15°N corresponds to the ITCZ region in which maximum outflow from tropical convection may be observed. Although monthly averages vary from year to year, the latitudinal variations in seasonal averages are remarkably similar in shape, with slightly greater southward transport in 1988. The poleward transport of moisture is most evident in the winter hemisphere, which is consistent with previous investigations of the Hadley cell circulation (Oort and Rasmusson 1970; Lindzen and Hou 1988; van de Berg et al. 1991).

Despite the interannual and month-to-month variability in the zonally averaged WVTI, the mean transport for each entire summer is close between the two years, as shown in Table 2. The mean meridional transport of moisture was southward for all summer months but clearly was stronger during 1988. The asymmetrical nature of the meridional moisture fluxes is enhanced during La Niña episodes such as was the case during JJA 1988. The similarity in the specific humidity for all summer months reflects the negligible month-to-month and interannual differences in the total mean brightness temperatures, therefore suggesting the differences in to be caused largely by the interannual variability of the meridional wind component.

Intercomparison of GOES and NCEP–NCAR monthly transport

It is informative to compare the monthly WVTI with a similar product derived from a numerical model such as the NCEP–NCAR Reanalysis to assess the amount of agreement and to reconcile differences in light of the limitations of the satellite data and the constraints of the model reanalysis fields. Although models have limitations on their representation of upper-level moisture, they do incorporate available rawinsonde data assimilated within constraints of a large-scale dynamical model. The satellite-derived WVTI represents the quasi-horizontal transport of moisture in a relatively thick layer of the mid- and upper troposphere. The vertical displacement of this layer varies spatially, as was shown in Fig. 3d, as a result of changes in the vertical temperature and moisture of the atmosphere (Fig. 6). Numerical model diagnostic fields provide thermodynamic parameters that are valid at particular model levels instead of in layers. Therefore, a direct comparison of model parameters with the WVTI (as introduced in Fig. 1) is not very complete. However, using the vertical profile information from the model and assuming the vertical position and thickness of the transport layer are best represented by the water vapor channel weighting function, an estimate of water vapor transport in the layer can be synthesized from the NCEP–NCAR data. In this way, the water vapor transport from NCEP–NCAR data can be compared with that from the satellite for similar layers. To achieve this comparison, the GOES VAS TB was simulated using the model data and an appropriate radiative transfer code. A representative pressure was obtained using a technique similar to that applied to the observed GOES data but using the simulated brightness temperature (i.e., the corresponding height of the brightness temperature in the model temperature profile is assumed to be the peak of the weighting function). The specific humidity and the u and υ wind components were determined by using the values that correspond to the wind height in the reanalysis profile data (log-linearly interpolating, when necessary, between model pressure levels). The result of this approach for June 1998 in producing fields of WVTI and from the NCEP–NCAR data is presented in Figs. 13a and 13b. A second approach to produce a WVTI and from the NCEP–NCAR model is to use the observed brightness temperatures in place of the simulated ones derived from the model gridpoint values. The results of this approach are shown in Figs. 13c and 13d. Differences between these two sets of figures result from the difference between the simulated and observed GOES water vapor brightness temperatures. A comparison of the figures shows only small discrepancies in terms of absolute magnitude of the transport parameters. A third method to synthesize the WVTI from the model data involves the use of the complete weighting function information corresponding to each model thermodynamic profile. The weighting function is used to weight the basic model parameters (q, u, and υ) over the representative layer of the atmosphere, rather than using values at a single pressure level that varies spatially. The result of this third approach is presented in Figs. 13e and 13f. Upon examination of these figures, one finds only minor differences in the three approaches, and all depict the same transport features. The third method produces larger-scale spatial patterns that at times have larger magnitudes. This outcome is the result of the weighting of information from different pressure levels above and below the ones used in 13a–d (not shown). The interpretation of water vapor transport from any of the three approaches would be nearly identical. For simplicity in the following analysis, the first approach (use of simulated brightness temperatures) has been used to produce NCEP–NCAR estimates of WVTI and for comparison with the GOES products.

The model-derived WVTI and values for JJA 1987 and 1988 using the above approach are presented in Figs. 14 and 15, respectively. These figures indicate the large-scale transport of water vapor in the upper troposphere, as represented in the NCEP–NCAR Reanalysis data. The monthly mean fields of WVTI from the model data clearly show strong vapor transport in the northern and southern midlatitude regions as well as strong transport over northern South America and the tropical eastern Pacific. The tropical maximum strengthens over the course of both summers, although transport is considerably weaker during JJA 1987. The Southern Hemisphere transport patterns associated with synoptic weather systems strengthen throughout the period, while, in the Northern Hemisphere, a persistent transport maximum is present off the northeastern United States coast during both summers.

The model WVTI can be compared with the satellite-derived values of Fig. 8. The differences between the NCEP–NCAR and GOES estimates of WVTI are considerable. It readily is apparent that the NCEP–NCAR method portrays stronger gradients and a larger magnitude range in the transport index fields throughout the analysis region. Despite these differences, both datasets agree on the general positions of maximum and minimum transport centers. For example, in June of 1987 the tropical eastern Pacific displayed weak moisture transport that is captured by both the model and GOES estimates. Similarities appear in the depiction of weak transport in the subsidence zones by both datasets. The month-to-month and interannual changes in the transport centers are consistent between the two datasets. The monthly averages of the daily grids presented in Table 2 generally show that the satellite and model mean values of WVTI are very close.

The model-derived meridional transport presented in Fig. 15 can be compared with the GOES estimates of in Fig. 9. In a mean sense, satellite estimates of for JJA of both summers clearly indicate stronger southward transport. GOES estimates of in the Tropics depict stronger southward transport south of the preferred position of the ITCZ (near 10°N). We feel this difference most likely is the result of poorly diagnosed meridional transport by the model-estimated wind field, since this region is notorious for a lack of the conventional upper-air measurements that normally influence the reanalysis fields. The model, on the other hand, portrays stronger meridional transport in the extratropics. Both datasets do a good job of diagnosing more prominent circulation features such as the anticyclonic flow in the central and western Atlantic Ocean during June 1988 (Figs. 15b and 9b).

Summary and conclusions

This paper presented a new approach to quantify upper-level moisture transport from geostationary satellite data. Daily time sequences of GOES-7 water vapor imagery were used to produce gridded estimates of winds and specific humidity in the upper troposphere over cloud-free regions viewed by the satellite. The winds and specific humidity values were combined to produce the water vapor transport index, which represents the magnitude of the two-dimensional transport of water vapor in the layer represented by the GOES water vapor channel vapor weighting function. Daily WVTI values and associated parameters (u and υ wind components, specific humidity, zonal and meridional transport, and the mean pressure corresponding to these parameters) were averaged to produce monthly fields for June, July, and August 1987 and 1988 over the Americas and surrounding oceanic regions.

The winds were produced using the Marshall Automated Wind algorithm, a feature-identification and-tracking algorithm that uses a sequence of three GOES water vapor images separated by 60 min to produce a pair of wind vectors quasi-uniformly spaced over the analysis region. Vector-pair consistency checks based on wind speed and direction differences between the temporal vector pairs provided the main quality control for the winds. Specific humidity values were obtained at wind locations from the observed brightness temperature, and an estimate of the relative humidity in the layer was derived from a modified version of the Soden and Bretherton (1996) technique. A monthly varying set of humidity regression coefficients was calculated from the NCEP–NCAR Reanalysis thermodynamic data and used in the humidity retrieval technique.

The WVTI compared favorably with a similar product derived from NCEP–NCAR Reanalysis–fixed pressure level data for the individual days examined. Both datasets showed strong upper-level water vapor transport associated with synoptic features in both hemispheres and with outflow from tropical convective systems, while minimal transport occurred over tropical and subtropical high pressure regions. However, index values from the NCEP–NCAR data were 2–3 times larger than that determined from the satellite data and displayed stronger spatial gradients. The WVTI represents the mean transport of water vapor over a thick layer of the upper troposphere whose vertical position and thickness are determined by the water vapor channel weighting function. The vertical position of the weighting function changes from tropical regions to higher latitudes such that a dry upper troposphere represents transport of a layer centered around 350–400 hPa. As moisture in the upper troposphere increases, the transport is determined to be at lower pressures (occasionally less than 300 hPa). Thus, differences between the satellite-derived WVTI and that from NCEP–NCAR data at a fixed model pressure level are expected. However, it does not explain fully a moist bias in reanalysis data, even when various model levels are considered. Based on previous literature citations and examination of the individual wind and humidity fields, it is believed that differences between the two datasets likely are caused by the manner in which the NCEP–NCAR Reanalysis determines the wind and humidity characteristics of the upper troposphere. Another possible explanation for the observed differences is the limitation of the satellite data in determining winds in regions where the satellite imagery shows minimal spatial structure. Because these differences come from systematic biases in the NCEP–NCAR or from sporadic occurrences in the satellite dataset, they do not preclude the use of the daily data to produce monthly averaged values for seasonal and interannual variability studies.

An analysis of the monthly mean gridded fields allowed for an investigation of interannual variability of upper-tropospheric water vapor transport depicted by the satellite data. Significant month-to-month variation was observed in the WVTI, which had spatial and temporal continuity, providing additional credibility to the derived index values. In the midlatitudes, large WVTI values were associated with transient synoptic features, while transport in the Tropics was more varied. In the Tropics, large transport values were observed south of the ITCZ on a frequent basis, although this feature also showed monthly variations. As was expected, considerable variability was observed in weak transport regions associated with the large-scale subsidence zones. This variability was related to the position and intensity of the surrounding convective activity and to the impact of transient weather systems on the monthly averaged data. The comparison of the summer monthly values between 1987 and 1988 provided some insight into the impact of the ENSO event during this period. Both years showed similar features in the midlatitudes, with large transport in the Southern (winter) Hemisphere, which seemed to intensify during each of the 3-month periods. In the Tropics, the eastern Pacific Ocean and the Caribbean Sea were dominated by large index values (>4.0 g kg−1 m s−1) during July 1987, with weaker transport during June and August. In contrast, July 1988 saw relatively weaker tropical water vapor transport, while June and August were notably stronger.

While the WVTI provides a quantitative measure of upper-level water vapor transport, it does not provide directional information. The meridional transport of upper-level water vapor was determined by combining the υ component of the wind with the specific humidity over the layer sensed by the satellite sensor. Monthly averaged values of meridional transport revealed a more pronounced midlatitude ridge–trough pattern in the Northern Hemisphere than in the Southern Hemisphere. Large poleward transport of upper-level moisture from the Tropics to the higher latitudes dominates the monthly fields. Meridional transport from the tropical convection favors the Southern (winter) Hemisphere, with a strong southward flux of upper-level moisture south of the ITCZ that suggests a single Hadley cell circulation pattern. The meridional transport fields provided even better month-to-month continuity than the WVTI did, with less interannual variability, particularly in the Tropics, which indicated that the differences resulted from the large variability of the zonal wind (transport). An analysis of the pressure fields associated with the upper-level water vapor transport showed that transport occurred at low pressures around convective regions and at much higher pressures near tropical and subtropical subsidence zones. A sharp gradient in pressure was present on the south side of the ITCZ region, where large meridional transport occurred. This gradient may imply that outflow from the tropical convection descended as it moved southward toward the eastern Pacific subsidence zone. The NCEP–NCAR vertical motion fields corroborated this vertical circulation during the period.

Monthly averaged WVTI and meridional transport fields from the NCEP–NCAR Reanalysis data were compared with the satellite values. It was shown that transport values from the NCEP–NCAR data interpolated to specific pressure levels compared favorably with values derived from a layer-weighting scheme (based on the water vapor channel weighting function). In comparison with the GOES-derived values, the monthly WVTI fields from the reanalysis data showed the same general patterns as the satellite data but had somewhat larger values, with much stronger spatial gradients. Agreement varied from month to month for both years. The model-derived fields of WVTI during July 1988 placed strong zonal transport of humidity over the Tropics, whereas the satellite data depicted transport of about 4.0–6.0 g kg−1 m s−1 less. Interestingly, the opposite trend was observed in the fields of meridional transport for July 1988. The reanalysis data indicated weaker meridional transport in the tropical regions where large southward (poleward) transport associated with the ITCZ was observed in the satellite data. This difference leads us to conclude that in the Tropics the satellite-derived meridional wind component is stronger than that of the model, while the model zonal wind speeds exceed those of the satellite-derived values.

Zonally averaged monthly values were shown to provide a simple way to quantify the monthly and interannual changes in water vapor transport. Zonally averaged WVTI values peaked in the Southern Hemisphere subtropics during all months for each year. In the Tropics, a single, more-pronounced peak located over the equator and south latitudes occurred in 1988, as opposed to a dual peak in 1987. The second peak around 20°N latitude is consistent with findings of Bell and Halpert (1998) in which upper-tropospheric winds were noted to be stronger in this region during warm ENSO events. Negative values of zonally averaged meridional transport (southward transport) were observed for all summer months, with greater intensity during JJA 1988. The asymmetric nature of the zonally averaged meridional transport (more southerly water vapor transport) was enhanced during the La Niña episode of 1988.

The findings of this study are very encouraging for the continued use of the WVTI to quantify seasonal and interannual changes in upper-level water vapor transport. While the reasonableness of the results has been demonstrated, future applications will benefit from the use of data from additional satellites and from employing enhanced processing methodologies. This demonstration included the use of 6 months of the GOES VAS Pathfinder dataset; the calculation of these transport parameters for all months from the Pathfinder period (May 1987–November 1988) would provide additional continuity to monitor monthly transport changes as the tropical forcing makes the transition from the warm event of 1987 to the cold one of 1988. Application of this methodology to the exceptionally strong 1997/98 ENSO event would provide an important contrast with the 1987/88 period. The 1997/98 analysis would benefit from the use of data from the next generation of GOES satellites (beyond the VAS series), which have demonstrated superior spatial resolution and radiometric accuracy. The use of these new data would improve feature identification and tracking in the difficult subtropical subsidence zones.

Acknowledgments

The authors acknowledge the helpful comments by and interactions with the three reviewers of this manuscript. We also thank Dr. Pete Robertson and Mr. Ron Suggs for their inspiration and useful comments on this research. The analysis of daily GOES data for this period was possible only as a result of the availability of the GOES Pathfinder dataset, produced by the University of Wisconsin with funding from NASA. Continued efforts to make large amounts of operational geostationary satellite data available for research will be of great benefit to the climate community. Portions of this research were funded by NASA’s Office of Earth Sciences and NOAA’s GOES I/M Product Assurance Project. The faithful support of this research through these programs by Drs. Ramesh Kakar, James Dodge, and James Purdom is sincerely appreciated.

REFERENCES

  • Arkin, P. A., 1988: The global climate for June–August 1987: Mature phase of an ENSO warm episode persists. J. Climate,1, 306–324.

  • Atkinson, R. J., 1984: Automated mesoscale winds determined from satellite imagery. Interim Report on NAS8-34596, General Electric Company, Huntsville, AL, 51 pp. [Available from LMC/GHCC, 977 Explorer Blvd., Huntsville, AL 35806.].

  • Atkinson, R. J., 1987: Automated mesoscale winds determined from satellite imagery. Final Report on NAS8-34596, General Electric Company, Huntsville, AL, 50 pp. [Available from LMC/GHCC, 977 Explorer Blvd., Huntsville, AL 35806.].

  • Bates, J. J., X. Wu, and D. L. Jackson, 1996: Interannual variability of upper-tropospheric water vapor band brightness temperature. J. Climate,9, 427–438.

  • Bell, G. D., and M. S. Halpert, 1998: Climate assessment for 1997. Bull. Amer. Meteor. Soc.,79, S1–S50.

  • Blackwell, K. G., and J. P. McGuirk, 1996: Tropical upper-tropospheric dry regions from TOVS and rawinsondes. J. Appl. Meteor.,35, 464–481.

  • Bony, S., Y. Sud, K. M. Lau, J. Susskind, and S. Saha, 1997: Comparison and satellite assessment of NASA/DAO and NCEP–NCAR reanalyses over tropical ocean: Atmospheric hydrology and radiation. J. Climate,10, 1441–1462.

  • Del Genio, A. D., 1994: Climate implications of the seasonal variation of upper tropospheric water vapor. Geophys. Res. Lett.,21, 2701–2704.

  • Eigenwillig, N., and H. Fischer, 1982: Determination of midtropospheric wind vectors by tracking pure water vapor structures in METEOSAT water vapor image sequences. Bull. Amer. Meteor. Soc.,63, 44–58.

  • Elliot, W. P., and D. J. Gaffen, 1991: On the utility of rawinsonde humidity archived for climate studies. Bull. Amer. Meteor. Soc.,72, 1507–1520.

  • Endlich, R. M., and D. E. Wolf, 1981: Automatic cloud tracking applied to GOES and METEOSAT observations. J. Appl. Meteor.,20, 309–319.

  • Fischer, H., N. Eigenwillig, and H. Muller, 1981: Information content of METEOSAT and Nimbus/THIR water vapor channel data: Altitude association of observed phenomena. J. Appl. Meteor.,20, 1344–1352.

  • Fritz, S., and J. S. Winston, 1962: Synoptic use of radiation measurements from satellite TIROS-II. Mon. Wea. Rev.,90, 1–9.

  • Fuelberg, H. E., and P. J. Meyer, 1984: An analysis of the AVE–SESAME I period using statistical structure and correlation functions. Mon. Wea. Rev.,112, 1562–1576.

  • Gray, W. M., 1984: Atlantic seasonal hurricane frequency. Part I: El Niño and 30 mb quasi-biennial oscillation influences. Mon. Wea. Rev.,112, 1649–1668.

  • Hillger, D. W., and T. H. Vonder Haar, 1988: Estimating noise levels of remotely sensed measurements from satellites using spatial structure analysis. J. Atmos. Oceanic Technol.,5, 206–214.

  • Janowiak, J. E., 1988: The global climate for March–May 1988: The end of the 1986–87 Pacific warm episode and the onset of widespread drought in the United States. J. Climate,1, 1019–1040.

  • Jedlovec, G. J., and R. J. Atkinson, 1996: Quality and control of water vapor winds. Preprints, Eighth Conf. on Satellite Meteorology and Oceanography, Atlanta, GA, Amer. Meteor. Soc., 5–8.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year Reanalysis Project. Bull. Amer. Meteor. Soc.,77, 437–471.

  • Kousky, V. E., and A. Leetmaa, 1989: The 1986–87 Pacific warm episode: Evolution of oceanic and atmospheric anomaly fields. J. Climate,2, 254–267.

  • Laurent, H., 1993: Wind extraction from Meteosat water vapor channel image data. J. Appl. Meteor.,32, 1124–1133.

  • Lindzen, R. S., and A. Y. Hou, 1988: Hadley circulations for zonally averaged heating centered off the equator. J. Atmos. Sci.,45, 2416–2427.

  • Menzel, W. P., and J. F. W. Purdom, 1994: Introducing GOES-I: The first of a new generation of geostationary operational environmental satellites. Bull. Amer. Meteor. Soc.,75, 757–780.

  • Menzel, W. P., W. L. Smith, and T. R. Stewart, 1983: Improved cloud motion wind vector and altitude assignment using VAS. J. Climate Appl. Meteor.,22, 377–384.

  • Merrill, R. T., 1989: Advances in the automated production of wind estimates from geostationary satellite imagery. Preprints, Fourth Conf. on Satellite Meteorology and Oceanography, San Diego, CA, Amer. Meteor. Soc., 246–249.

  • Merrill, R. T., W. P. Menzel, W. Baker, J. Lynch, and E. Legg, 1991: A report on the recent demonstration of NOAA’s upgraded capability to derive cloud motion satellite winds. Bull. Amer. Meteor. Soc.,72, 372–376.

  • Montgomery, H. E., and L. W. Uccellini, 1985: VAS demonstration:(VISSR Atmospheric Sounder) description and final report. NASA Reference Publication 1151, 198 pp. [Available from NASA Center for Aerospace Information, P.O. Box 8757, Baltimore/Washington International Airport, Baltimore, MD 21240.].

  • Nieman, S. J., W. P. Menzel, C. M. Hayden, D. Gray, S. T. Wanzong, C. S. Velden, and J. Daniels, 1997: Fully automated cloud-drift winds in NESDIS operations. Bull. Amer. Meteor. Soc.,78, 1121–1133.

  • Oort, A. H., and E. M. Rasmusson, 1970: The annual variation of the monthly mean meridional circulation. Mon. Wea. Rev.,98, 423–442.

  • Peixoto, J. P., and A. H. Oort, 1996: The climatology of relative humidity in the atmosphere. J. Climate,9, 3443–3463.

  • Picon, L., and M. Desbois, 1990: Relation between METEOSAT water vapor radiance fields and large scale tropical circulation features. J. Climate,3, 865–876.

  • Rind, D., E. W. Chiou, W. Chu, J. Larsen, S. Oltmans, J. Lerner, M. P. McCormick, and L. McMaster, 1991: Positive water vapour feedback in climate models confirmed by satellite data. Nature,349, 500–503.

  • Ropelewski, C. F., 1988: The global climate for June–August 1988:A swing to the positive phase of the Southern Oscillation, drought in the United States, and abundant rain in the monsoon areas. J. Climate,1, 1153–1177.

  • Salathe, E. P., and D. L. Hartmann, 1997: A trajectory analysis of tropical upper-tropospheric moisture and convection. J. Climate,19, 2533–2547.

  • Schmetz, J., and O. M. Turpeinen, 1988: Estimation of the upper tropospheric relative humidity field from METEOSAT water vapor image data. J. Appl. Meteor.,27, 889–899.

  • Schmetz, J., K. Holmlund, J. Hoffman, B. Strauss, B. Mason, V. Gaertner, A. Koch, and L. van de Berg, 1993: Operational cloud-motion winds from Meteosat infrared images. J. Appl. Meteor.,32, 1206–1225.

  • Schmetz, J., C. Geijo, W. P. Menzel, K. Strabala, L. van de Berg, K. Holmlund, and S. Tjemkes, 1995a: Satellite observations of upper tropospheric relative humidity, clouds and wind field divergence. Contrib. Atmos. Phys.,68, 345–357.

  • Schmetz, J., and Coauthors, 1995b: Monthly mean large-scale analyses of upper-tropospheric humidity and wind field divergence derived from three geostationary satellites. Bull. Amer. Meteor. Soc.,76, 1578–1584.

  • Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc.,74, 2331–2342.

  • Schwartz, B. E., and C. A. Doswell III, 1991: North American rawinsonde observations: Problems, concerns, and a call to action. Bull. Amer. Meteor. Soc.,72, 1885–1896.

  • Soden, B. J., 1998: Tracking upper tropospheric water vapor radiances from satellite. J. Geophys. Res.,103, 17 069–17 081.

  • Soden, B. J., and F. P. Bretherton, 1993: Upper tropospheric relative humidity from the GOES 6.7-μm channel: Method and climatology for July 1987. J. Geophys. Res.,96, 16 669–16 688.

  • Soden, B. J., and F. P. Bretherton, 1994: Evaluation of water vapor distribution in general circulation models using satellite observations. J. Geophys. Res.,99, 1187–1210.

  • Soden, B. J., and F. P. Bretherton, 1996: Interpretation of TOVS water vapor radiances in terms of layer-average relative humidities: Method and climatology for the upper, middle, and lower troposphere. J. Geophys. Res.,101, 9333–9343.

  • Soden, B. J., and J. R. Lanzante, 1996: An assessment of satellite and radiosonde climatologies of upper-tropospheric water vapor. J. Climate,9, 1235–1250.

  • Spencer, R. W., and W. D. Braswell, 1997: How dry is the tropical free troposphere? Implications for global warming theory. Bull. Amer. Meteor. Soc.,78, 1097–1106.

  • Stephens, G. L., D. L. Jackson, and I. Wittmeyer, 1996: Global observations of upper-tropospheric water vapor derived from TOVS radiance data. J. Climate,9, 305–326.

  • Stewart, T. R., C. M. Hayden, and W. L. Smith, 1985: A note on water-vapor wind tracking using VAS data on McIDAS. Bull. Amer. Meteor. Soc.,66, 1111–1115.

  • Sun, D. Z., and R. S. Lindzen, 1993: Distribution of tropical tropospheric water vapor. J. Atmos. Sci.,50, 1643–1660.

  • Trenberth, K. E., 1997: Short-term climate variations: Recent accomplishments and issues for future progress. Bull. Amer. Meteor. Soc.,78, 1081–1096.

  • van de Berg, L., A. Pyomjamsri, and J. Schmetz, 1991: Monthly mean upper tropospheric humidities in cloud-free areas from METEOSAT observations. Int. J. Climatol.,11, 819–826.

  • Velden, C. S., C. M. Hayden, S. J. Nieman, W. P. Menzel, S. Wanzong, and J. S. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc.,78, 173–195.

  • Velden, C. S., T. L. Olander, and S. Wanzong, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part I: Dataset methodology, description, and case analysis. Mon. Wea. Rev.,126, 1202–1218.

  • Weinreb, M. P., H. E. Fleming, L. M. McMillin, and A. C. Nuendorffer, 1981: Transmittances for the TIROS Operational Vertical Sounder. NOAA Tech. Rep. TR NESS 85, NOAA, Washington, DC, 60 pp. [NTIS PB82-155920.].

  • Weldon, R. B., and S. J. Holmes, 1991: Water vapor imagery: Interpretation and applications to weather analysis and forecasting. NOAA Tech. Rep. NESDIS 67, 213 pp. [Available from NOAA/NESDIS, 5200 Auth Rd., Washington, DC 20233.].

  • Young, J. T., J. W. Hagens, and D. M. Wade, 1995: GOES Pathfinder product generation system. Preprints, Ninth Conf. on Applied Climatology, Dallas, TX, Amer. Meteor. Soc., 176–180.

APPENDIX

Feature Tracking in Satellite Imagery

The basic approach to all satellite wind-tracking schemes is the identification of features common in time to a sequence of images, the determination of their relative positions with respect to some fixed coordinate system, and the assignment of that wind to a specific pressure level or height above the surface. The first step (feature identification) is not a trivial issue and is compounded both by relatively poor satellite resolution of clouds and water vapor features and by the changing dynamical environment, which governs cloud development and the movement of water vapor in the atmosphere. Originally this task was done using the concept of the “person in the loop” whereby the human, with superior ability visually to identify and to track features, views a sequence of images to identify accurately and to determine physical displacements of clouds in the imagery (Stewart et al. 1985). This manual tracking process is highly accurate because the humans also can use their experience and knowledge of atmospheric processes to determine the proper features to track in time. This process is very labor intensive and is not applicable to the generation of large wind datasets for research or operational applications. The second step is accomplished by using accurate navigational placement of the individual satellite images and a conversion of displacement in satellite coordinates to that of earth coordinates. This procedure is well established. The third step is a difficult one and often is the most critical in NWP and data assimilation studies.

There are a number of automated approaches to performing feature identification between a pair of images (e.g., Endlich and Wolf 1981; Eigenwillig and Fischer 1982). Soden (1998) has developed an approach that uses a time-lagged cross-correlation method on a sequence of two GOES water vapor images separated by 60 min. Reverse correlation [tracking features in the second (later in time) image to the first image] is used to provide confidence in the retrieved wind vector. Laurent (1993) uses a cross-correlation method with a sequence of three Meteosat images separated by 30 min, with vector pair checks used for determining good wind vectors. Velden et al. (1997, 1998) and Nieman et al. (1997) describe the approach used by the National Oceanic and Atmospheric Administration for both operational support and research applications. Their approach uses the sum of the squared differences between the target box and a search box in a sequence of three water vapor images separated by 30 min in time. Automated editing procedures consider vector-pair consistency checks, spatial buddy checks, and a comparison with model forecast fields to eliminate bad vectors and to adjust the height assignment of the wind vectors to a more appropriate pressure level (Velden et al. 1998). The MAW approach is a template-matching scheme in which the sum of the differences between pixels in a tracking template provides the sole quantitative measure of feature matching between three sequential water vapor images. These algorithms also differ in other ways, namely, in their use of template size, search area, target selection, and editing procedures, all of which can produce big differences in the accuracy and resolution of the derived wind vectors. A discussion of these differences is presented below.

Template size

One of the key parameters in the MAW approach is the template size. The template defines a region in the image that contains the pattern to be matched. Laurent (1993) uses a 32 × 32 pixel region in 5-km Meteosat data. Soden (1998) uses a region equivalent to 46 × 46 pixels in 8-km GOES-7 data. Nieman et al. (1997) use a 15 × 15 pixel region in 4-km GOES-8 data. These target area sizes are quite varied, with the smallest being 60 km × 34 km (Nieman et al. 1997; Velden et al. 1997) and the largest being 368 km × 322 km (Soden 1998) and 392 km × 350 km in the current study. (Note that a square array of pixels in GOES data actually corresponds to a rectangular area on the earth since the instantaneous field of view of each footprint is oversampled in the east–west direction.) If the template is too small, the pattern or structure of the water vapor field is uniform, making a successful match in the second or subsequent images difficult to obtain accurately. To understand what the best template size might be, one can examine the spatial structure in the GOES VAS water vapor imagery to infer proper template size, as in Jedlovec and Atkinson (1996). In the absence of clouds, the upper-tropospheric water vapor structure is small, with large gradients observed only at scales greater than several hundred kilometers. The presence of high clouds adds significant structure to the imagery at scales below this threshold. This structure is confirmed when looking at the water vapor imagery (e.g., the satellite image in Fig. 1). Another way to determine the appropriate template size for water vapor feature tracking is to determine the relationship between accuracy of the winds and template size. To examine this relationship, the MAW scheme was used to track features in water vapor imagery with varying template size. The results are shown in Fig. A1 for GOES VAS and GOES-8 imager data. The random noise present in the derived wind vectors for each run was determined with structure function analysis (Hillger and Vonder Haar 1988). It is obvious from the figure that small template sizes are associated with large wind errors for both GOES VAS and GOES-8 wind data. The noise is lower in the GOES-8 data because of better sensor resolution (nominally 16 km × 16 km for VAS and 8 km × 8 km for GOES-8 water vapor data) and better radiometric accuracy of the data. Large errors occur at small templates because there is ambiguity in matching small features (structure in small templates) over a given time interval. The random noise decreases with increasing template size as the larger template detects more image structure for both GOES VAS and GOES-8. The template size at which this rapid change in performance occurs varies somewhat between VAS and the GOES-8 imager for reasons already mentioned. The approximate node point in the curve where random wind errors no longer change rapidly with template size is about 49 for GOES VAS data. Based on these results, a template size of 49 × 49 8-km pixels was selected for the VAS data used in this study. Random wind errors for templates of this size range from 4 to 6 m s−1 for unedited data (with no postprocessing quality control). Smaller templates may be appropriate for cloud tracking where thermal (infrared) or reflectance (visible) image structure is greater than that of the water vapor imagery or in regions of the water vapor imagery where high clouds dominate. However, because of the greater spatial resolution and radiometric accuracy of the GOES-8 data, smaller water vapor discontinuities and features are identifiable at scales pertinent to NWP and mesoscale analysis. As template size is increased, the template may be more likely to cover regions of wind shear or features that do not represent the instantaneous wind motion. This likelihood may produce unrepresentative motion fields in some cases.

The use of quality control procedures to reduce the errors in the data has an interesting effect on the noise–template size relationship. Noise can be considerably reduced for all template sizes for both GOES VAS and GOES-8 data with use of appropriate quality control procedures. In Fig. A1, the thin curves corresponded to the use of 5 m s−1 acceleration criteria between two vector pairs. For this situation, random noise is estimated at between 2 and 4 m s−1 for all template sizes. Also, the performance difference between GOES VAS and GOES-8 is decreased; that is, the quality of the VAS-derived winds is now similar to that of GOES-8. As can be seen by the lines with the □ and ○ symbols in Fig. A1 and read off the right axis, the reduction in the number of good vectors incurred by imposing these quality control parameters is substantial.

Search area

Another parameter needed for the tracking algorithm is the size of the search area. The search area size (and shape) should be influenced by the expected magnitude and direction of the wind. Use of too large a search area may allow matches that produce unrealistic wind displacements (speeds). Very small search areas artificially constrain the winds and reduce the number of good matches in the search area. Velden et al. (1997, 1998), Nieman et al. (1997), and Schmetz et al. (1993) have used winds from a numerical model to provide a first guess of the image displacement (wind) to limit the search area and to reduce computation time. This method is helpful if computer time is important (operational constraints) and if the user is confident that the “guess” is reasonable (close to the correct speed and direction). In the performance analysis of the MAW algorithm, the rejection of a large number of wind vectors because the template difference is on the edge of the search area is an indication of too small a search area. In this application with GOES VAS data, the search distance was selected to be 31 (8 km × 7 km) pixels. This distance allows winds to be as strong as 70 m s−1 at the satellite subpoint and over 100 m s−1 in other regions (where the satellite zenith angle is greater than 30°).

Target spacing

Target selection or the spacing of the templates in the image is another consideration in retrieving satellite-derived wind vectors. Some approaches look for regions of large image structure and apply their tracking methodology to those regions. This method may optimize pattern matching for some targets and may minimize the need for editing the final product. This approach is dependent, however, on the threshold used to identify gradients that are large enough to be tracked and that also may produce clusters of winds, and on regions that are void of vectors altogether (fall below some subjective gradient in the image). This study’s approach places initial targets uniformly spaced in the image every 49 pixels in the line and element direction (equivalent to the template size of 49 × 49 8 km × 7 km pixels). In this way, every pixel is used to characterize the image structure and is used in the template-matching process. This approach produces a somewhat uniform distribution of winds (assuming most pass quality-filtering procedures) over the entire region.

Wind errors

In the application of tracking algorithms to sequential satellite imagery for wind determination, it is assumed that the clouds or water vapor features are conservative, passive tracers of the wind field. However, this assumption is not always the case, and changes in shapes and radiative characteristics of clouds and water vapor features may be misinterpreted and lead to wind errors. Velden et al. (1997) recognized this and provide an explicit correction to the wind vectors to account for a consistently slow bias when compared with the model forecast of winds. Wind vectors can be in error as a result of misidentification of targets from scene to scene, from improper determination of correct image displacements, and from poor height assignment.

Most errors in wind vectors occur as a result of poor target matches. All tracking schemes have to deal with this problem. These types of errors can be minimized by the selection of appropriate tracking criteria (discussed above) and postprocessing quality control procedures. For the application here, a large template (49 × 49 8 km × 7 km pixels) and search distance (31–8 km × 7 km–pixel radius) and hourly GOES VAS data are used to attempt to keep these errors to a minimum. While the minimum difference pattern-matching approach eliminates some of the more common target identification problems, the relatively coarse image resolution (subsampled 16 km × 14 km VAS data) and time interval of the GOES VAS Pathfinder dataset do permit changes in cloud features that can blur the matching process. To address the possible errors introduced by the image resolution, two quality control parameters are used in postprocessing (editing) of the wind data. Since two separate wind vectors are determined for each target in a sequence of three images, consistency between these vectors should provide a greater level of confidence in the individual winds. Others have used a vector pair difference between the individual u and υ wind components in their postprocessing procedures with good success (Merrill et al. 1991; Velden et al. 1997; Laurent 1993). Additional automated and manual editing often is imposed to constrain the results to be consistent with a model forecast wind field (Velden et al. 1997, 1998; Laurent 1993). The rms error of the water vapor–based winds with this approach ranges from roughly 2 to 10 m s−1, depending on the application. The current approach is similar and applies a vector-pair consistency check with separate thresholds on the wind direction and speed. In a related research effort, the authors found that a 10 m s−1 speed and 25° direction threshold produced an error reduction similar to that for a 5 m s−1 u and υ threshold. For this application, differences between the magnitude of the winds (speed) of vector 1 and vector 2 of greater than 15 m s−1 or direction differences greater than 30° for a given location are flagged as bad. This cutoff is less stringent and is justified given the analysis procedures used in this climate analysis research. Retrieved winds are analyzed objectively before being combined with the moisture to produce second-order parameters. This gridding process smooths the winds and reduces the impact of any remaining error on the analysis fields.

The magnitude of the error implicit in satellite-derived winds before and after quality filter procedures are applied is, at times, difficult to assess because of the lack of ground-truth or verification data. There is a tendency to compare satellite-derived winds with rawinsonde winds to assess their accuracy. While rawinsondes provide a traditional standard for ground-truth comparisons of many satellite-derived products (temperature and moisture profiles, total precipitable water vapor, winds, etc.), care must be used in interpreting the results of such comparisons for winds. The rawinsondes are essentially a point measurement at a specific time. Satellite-derived winds are a volumetric (vertical and horizontal) estimate of the flow characteristics averaged or sampled over a period of time (anywhere from a few minutes to hours). These comparisons can only provide limited guidance on the accuracy of satellite-derived winds. A unique approach to error assessment in satellite-derived winds is the use of statistical structure functions to quantify independently the random error associated with the wind dataset without reference to rawinsonde or modeled wind data. Hillger and Vonder Haar (1988), Fuelberg and Meyer (1984), and others have shown that structure function analysis can be used to estimate the magnitude of mean nondirectional gradients (structure) in data fields. The slope of the structure curves at small separation intervals can be used in the error estimation. The reduction in this random error associated with the use of editing procedures is used as a measure of success for the quality control parameters. The current quality control approach used with the MAW algorithm for this application (vector pair differences of 15 m s−1 in speed or 30° direction) reduces random errors in GOES VAS data to less than 4 m s−1 (Jedlovec and Atkinson 1996).

Fig. 1.
 Fig. 1.

WVTI from (a) NCEP–NCAR Reanalysis at 400 hPa and (b) GOES VAS for 1200 UTC on 14 Jun 1988. The GOES VAS data are shown with streamlines in blue. WVTI units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 2.
Fig. 2.

The MAW algorithm feature-identification and -tracking procedure. A sequence of three images is used to determine image displacements used for estimates of the wind.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 3.
 Fig. 3.

An example of products derived from GOES data on 14 Jun 1988: (a) water vapor–based winds (knots), unedited; (b) edited and cloud-filtered winds interpolated to a grid (knots); (c) upper-tropospheric relative humidity (%) derived at wind vector locations and interpolated to a grid; and (d) assigned pressure (hPa) of the winds and humidity. In (a), the red wind flags are those with large accelerations or directional deviations between vector pairs, and blue wind flags correspond to winds in cloudy regions.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 4.
 Fig. 4.

Relationship between the logarithm of layer relative humidity [left side of (4)] and the GOES water vapor brightness temperature (simulated from the NCEP–NCAR Reanalysis Jan 1988 monthly mean data). Red dots represent points where satellite zenith angle θ is greater than 75°. Least squares fit line and regression statistics displayed in upper right corner of figure are computed from θ less than 75°.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 5.
Fig. 5.

Scatterplot of relative humidity showing the bias that results from the use of retrieval coefficients not representative of the retrieval period. (a) Jul coefficients applied to data from 5 Jan 1988; (b) Jan coefficients applied to data from 14 Jun 1988. Deviations from the one-to-one fit line represent biases associated with applying the inappropriate monthly coefficients.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 6.
Fig. 6.

Water vapor weighting functions for different moisture profile configurations: (top) AFGL tropical;(middle) AFGL tropical but with relative humidity configured as in Profile A; (bottom) AFGL tropical but with relative humidity configured as in Profile B. Note that for nearly the same absolute moisture amount q the shape and peak of the weighting functions change because of the different vertical distribution of moisture.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 7.
Fig. 7.

Upper-level specific humidity q and wind speed SPD derived from (a), (c) GOES and (b), (d) NCEP–NCAR Reanalysis data for 14 Jun 1988 at 1200 UTC. Model data are at 400 hPa. Units of specific humidity are 10−1 g kg−1 and wind speeds are m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 8.
Fig. 8.

GOES-derived WVTI for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 9.
Fig. 9.

GOES-derived mean meridional transport of specific humidity for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 10.
Fig. 10.

Mean monthly pressure assigned to the layer wind and humidity fields for Jun, Jul, and Aug of (a), (c), (e) 1987 and (b), (d), (f) 1988. Units are hPa.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 11.
Fig. 11.

Jun 1988 mean vertical velocity estimates from NCEP–NCAR Reanalysis at 400 hPa. Units are 10−2 Pa s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 12.
Fig. 12.

Zonally averaged (30°–120°W) (a), (b) WVTI and (c), (d) meridional transport of specific humidity for Jun, Jul, and Aug 1987 and 1988. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 13.
Fig. 13.

Layer estimates of WVTI and derived from the NCEP–NCAR Reanalysis monthly mean data using three different methods: (a), (b) simulated brightness temperature; (c), (d) observed brightness temperature; (e), (f) full weighting function approach. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 14.
Fig. 14.

WVTI simulated from the NCEP–NCAR Reanalysis monthly mean temperature and moisture profile for Jun, Jul, and Aug of 1987 and 1988. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Fig. 15.
Fig. 15.

Meridional transport of moisture simulated from the NCEP–NCAR reanalysis monthly mean temperature and moisture profile for Jun, Jul, and Aug of 1987 and 1988. Units are g kg−1 m s−1.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

i1520-0450-39-1-15-fa1

Fig. A1. Random wind error associated with size of tracking template. The bold lines correspond to unedited wind data and the thin lines correspond to edited winds using a 5 m s−1 acceleration threshold. The medium thickness lines with values read from the right axis show the reduction in good wind vectors because of the editing.

Citation: Journal of Applied Meteorology 39, 1; 10.1175/1520-0450(2000)039<0015:ASDUTW>2.0.CO;2

Table 1.

Relative humidity retrieval coefficients derived from NCEP–NCAR monthly mean data for 1988 for satellite zenith angles less than 75° (about 2300 points). The correlation values show the strong relationship for each month.

Table 1.
Table 2.

Total monthly mean values for GOES layer-specific humidity q, pressure P, meridional flux of specific humidity , and water vapor transport index WVTI. Corresponding values derived from the NCEP–NCAR simulated water vapor channel brightness temperatures are shown in parentheses.

Table 2.
Save