Satellite and CALJET Aircraft Observations of Atmospheric ...

JULY 2004 1721R A L P H E T A L .

Satellite and CALJET Aircraft Observations of Atmospheric Rivers over the EasternNorth Pacific Ocean during the Winter of 1997/98

F. MARTIN RALPH, PAUL J. NEIMAN, AND GARY A. WICK

NOAA/Environmental Technology Laboratory, Boulder, Colorado

(Manuscript received 20 November 2003, in final form 16 January 2004)

ABSTRACT

This study uses a unique combination of airborne and satellite observations to characterize narrow regionsof strong horizontal water vapor flux associated with polar cold fronts that occurred over the eastern NorthPacific Ocean during the winter of 1997/98. Observations of these ‘‘atmospheric rivers’’ are compared with pastnumerical modeling studies to confirm that such narrow features account for most of the instantaneous meridionalwater vapor transport at midlatitudes.

Wind and water vapor profiles observed by dropsondes deployed on 25–26 January 1998 during the CaliforniaLand-falling Jets Experiment (CALJET) were used to document the structure of a modest frontal system. Thehorizontal water vapor flux was focused at low altitudes in a narrow region ahead of the cold front where thecombination of strong winds and large water vapor content were found as part of a low-level jet. A closecorrelation was found between these fluxes and the integrated water vapor (IWV) content. In this case, 75% ofthe observed flux through a 1000-km cross-front baseline was within a 565-km-wide zone roughly 4 km deep.This zone contained 1.5 3 108 kg s21 of meridional water vapor flux, the equivalent of ;20% of the globalaverage at 358N.

By compositing polar-orbiting satellite Special Sensor Microwave Imager (SSM/I) data from 46 dates con-taining long, narrow zones of large IWV, it was determined that the single detailed case was representative ofthe composite in terms of both the IWV amplitude (3.09 cm vs 2.81 cm) and the width of the area where IWV$ 2 cm (424 km vs 388 km). The SSM/I composites also showed that the width scales (defined by the 75%cumulative fraction along a 1500-km cross-plume baseline) for cloud liquid water and rain rate were 176 and141 km, respectively, which are narrower than the 417 km for IWV. Examination of coincident GeostationaryOperational Environmental Satellite (GOES) and SSM/I satellite data revealed that GOES cloud-top temperatureswere coldest and cloud-top pressures were lowest in the core of the IWV plumes, and that the core cloud topsbecame substantially colder and deeper for larger IWV. A strong latitudinal dependence of the satellite-derivedcross-river characteristics was also found.

Atmospheric rivers form a critical link between weather and climate scales. They strongly influence bothshort-term weather and flood prediction, as well as seasonal climate anomalies and the global water cycle,through their cumulative effects. However, the rivers remain poorly observed by the existing global atmosphericobserving system in terms of their horizontal water vapor fluxes.

1. Introduction

The distribution and transport of water vapor in theatmosphere is critical to a range of issues extending fromthe issuance of flood warnings to initialization of nu-merical weather prediction models and to understandingthe hydrological and global climate systems. From bothweather and climate perspectives it is well known thata key contributor to the advection of water vapor is thelow-level jet (LLJ), which is found in the warm sectorsof extratropical cyclones (e.g., Namias 1939; Bonner1968; Palmen and Newton 1969; Browning and Pardoe1973; Cotton and Anthes 1989; Bluestein 1993) and as

Corresponding author address: Dr. F. Martin Ralph, NOAA/En-vironmental Technology Laboratory, Mail Code R/ET7, 325 Broad-way, Boulder, CO 80305.E-mail: [email protected]

part of diurnal oscillations over the Great Plains (e.g.,Bonner 1968; Stensrud 1996; Whiteman et al. 1997).The type of LLJ that is the focus of this paper is partof a broader region of generally poleward heat transportwithin extratropical cyclones that is referred to as the‘‘warm conveyor belt’’ (e.g., Browning 1990; Carlson1991). The warm conveyor belt is an integral componentof extratropical cyclones that plays a key role in trans-porting sensible and latent heat poleward, balancing theequatorward transport of relatively cool, dry air in otherbranches of the extratropical cyclone’s circulation.

This paper focuses on a subset of the total heat trans-port, that is, the poleward meridional water vapor flux.For a variety of reasons its distribution can evolve dif-ferently from that of the sensible heat flux component,for example, the sources and sinks of water vapor differmarkedly from those of sensible heat. Recently, in a

1722 VOLUME 132M O N T H L Y W E A T H E R R E V I E W

numerical modeling study using European Centre forMedium-Range Weather Forecasts global analyses, Zhuand Newell (1998) explored the role of extratropicalcyclones in meridional water vapor transport. By estab-lishing a threshold value of meridional water vapor fluxthey segregated daily fluxes at each grid point as either‘‘broad’’ fluxes or ‘‘river’’ fluxes. Based on this seg-regation, it was possible to determine how much of thetotal meridional flux at a given latitude was attributableto atmospheric rivers and to diagnose the meridionalwidth scale of these features at a given latitude, at leastfrom the model perspective. They concluded that .90%of the total meridional water vapor flux in the midlat-itudes was accomplished by ‘‘atmospheric rivers’’ cov-ering ,10% of the total hemispheric circumference. Theterm ‘‘atmospheric river’’1 appropriately emphasizesboth their filamentary nature and the focus on watervapor. This also suitably distinguishes it from the broad-er region characterized as the warm conveyor belt withits mixture of both sensible and latent heat components.

Zhu and Newell (1998) also found that four to fivesuch rivers are usually present at any given time. Thiscorresponds well to planetary wavenumber 4–5, whichis often found during Northern Hemisphere cool sea-sons. This correspondence likely results from the factthat each planetary wave normally has associated withit a synoptic-scale extratropical cyclone, which in turnusually has large- and mesoscale frontal processes thatcan give rise to a narrow region of strong meridionalwater vapor flux. However, the position of the planetarywaves relative to the continents, major land surface fea-tures, and oceans can also strongly influence their im-pact on global meridional water vapor transport. Forexample, an extratropical cyclone entraining dry airfrom a continental region will not be effective at trans-porting large amounts of water vapor until it taps intoan area with greater moisture content. This is often seenwhen extratropical cyclones move from North Ameri-ca’s Rocky Mountains onto the Great Plains where theycan then entrain water vapor from the Gulf of Mexico(e.g., Palmen and Newton 1969). Although this paperdoes not address these issues of ocean versus land orof the planetary wave pattern, it does inform futureresearch on these topics.

It should be noted that a set of earlier studies focusedon middle- and upper-tropospheric cloud bands (andtheir implied regions of enhanced water vapor transport)using primarily infrared satellite images. Thepenier andCruette (1981) documented the evolution of distinctive,long bands of mid- and high clouds that could stretch

1 This term, or its related version ‘‘tropospheric rivers’’ was intro-duced by Newell et al. (1992) and has been used since then by Zhuand Newell (1994) in a study relating them to rapid extratropicalcyclogenesis, by Smirnov and Moore (1999, 2001) in studies of watervapor transport into the McKenzie River basin of northwestern Can-ada, and by Zhu et al. (2000) in a study of upper-tropospheric watervapor transport.

thousands of kilometers. This study was extended usingmore cases to document the origins of these cloud bandsthat mark a linkage between the Tropics and the extra-tropics (McGuirk et al. 1987; Kuhnel 1989; Iskenderian1995). While these studies revealed important aspectsof the cloud bands, including the role of tropical andextratropical triggers, they were unable to document theactual water vapor content within the bands (includingconditions at lower altitudes where the most water vaporexists), which is the focus of the study presented here.

The role of the LLJ (and of atmospheric rivers) inproducing heavy rainfall has been explored in complexterrain where orographic lifting of the moist, and oftenpotentially unstable, air mass can create heavy rainfall(Reynolds and Kuciauskas 1988; Heggli and Rauber1988; Neiman et al. 2002; Ralph et al. 2003). This re-lationship, along with limitations in observations of theLLJ offshore and along the U.S. West Coast, motivatedthe California Land-falling Jets Experiment (CALJET)that was conducted along and offshore of the Californiacoast in the winter of 1997/98 (Ralph et al. 1999). TheCALJET campaign, led by the National Oceanic andAtmospheric Administration’s (NOAA) EnvironmentalTechnology Laboratory (ETL), resulted in improved un-derstanding of the role of the LLJ in modulating oro-graphic rainfall (Neiman et al. 2002) and flooding(Ralph et al. 2003) and produced a dataset designed toexplore the connections between short-term climate var-iability and extreme weather events in a record wet win-ter. Atmospheric rivers form such a linkage and will bethe focus of the observational study presented here. Aprimary goal of this study is the use of observations toevaluate the model-based conclusions of Zhu and New-ell (1998), which used a model with effective grid sizeof 200 km that may not have adequately resolved therivers. Specific questions addressed in our paper include,what is the magnitude of the horizontal water vapor fluxin a well-observed moderate-intensity oceanic storm,and what are its characteristic width and depth scales?How do the case study results compare with compositesderived from satellite data? What are the capabilitiesand limitations of existing satellite-observing systemswith respect to detecting key aspects of atmosphericrivers over oceans for forecasting and climate monitor-ing?

The paper first describes the key ground-based, air-borne, and satellite-observing systems deployed by orutilized by CALJET. The synoptic and mesoscalecharacteristics of a well-observed atmospheric riverover the eastern Pacific are summarized, resulting ina detailed snapshot of an atmospheric river from drop-sonde data. The representativeness of this case studyis then examined using satellite data from the entirewinter season of 1997/98, which also are used to di-agnose key characteristics of the distributions of in-tegrated water vapor, cloud liquid water, rain rate,ocean-surface wind speed, cloud-top temperature, andcloud-top pressure.

JULY 2004 1723R A L P H E T A L .

2. Observing platforms

A suite of research observing platforms was deployedfor CALJET across coastal California and the data-sparse eastern Pacific (Ralph et al. 1999), comple-menting the established operational observing systems.A NOAA P-3 turboprop research aircraft measured stan-dard meteorological parameters every 1 s during a flightat its highest operating altitude (500–400 mb) and re-leased 29 Global Positioning System (GPS) dropsondesthrough a polar frontal zone offshore of California be-tween 2003 UTC 25 January and 0447 UTC 26 January1998. The dropsondes provided high-resolution verticalprofiles of wind velocity, temperature, and water vapor.A lower-fuselage scanning C-band (;5.6 cm wave-length) radar provided reflectivity measurements of pre-cipitation elements.

Research data collected from three 915-MHz windprofilers (e.g., Ecklund et al. 1988; Carter et al. 1995)operated by NOAA/ETL were utilized in this study.These profilers, located at Bodega Bay (BBY), PointArena (PAA), and Eureka (ERK), provided hourly av-eraged vertical profiles of horizontal wind velocity from;0.1 to 4.0 km above ground level (AGL) with #100m vertical resolution in clear, cloudy, and precipitatingconditions. The profiler winds were objectively editedusing the vertical–temporal continuity method of Weberet al. (1993). The height of the precipitation meltinglevel was determined objectively using the brightbanddetection method of White et al. (2002). At each profilersite a tipping-bucket rain gauge measured rainfall with0.01-in. (;0.25 mm) resolution, while observations ofwind, temperature, moisture, and pressure were col-lected from a 10-m tower, all with 2-min sampling.

Operational polar-orbiting and Geostationary Oper-ational Environmental Satellites (GOES) also furnishedobservations. A Special Sensor Microwave Imager(SSM/I) was carried on three Defense MeteorologicalSatellite Program polar orbiters (F11, F13, F14) thatcircled the globe once every ;102 min. Troposphericintegrated water vapor (IWV) (Schluessel and Emery1990), cloud liquid water (CLW) (Weng and Grody1994), rain rate (RR) (Ferriday and Avery 1994), andocean-surface wind speed (SPD) (Goodberlet et al.1990) were retrieved from each SSM/I sensor over a1400-km-wide swath and gridded at 25-km resolution.The SSM/I measurements are remarkably reliable(Wentz 1997), although two important exceptionsshould be noted: 1) IWV measurements can be degradedin heavy rain, and 2) useful retrievals of surface windspeed are not possible at all when rain is present. TheSSM/I sampling is asynoptic and somewhat irregular intiming and location, and retrievals are generally notavailable over land. A pair of GOES satellites (GOES-8/9) provided nearly continuous spatiotemporal obser-vations over the Pacific and adjacent landmasses withtheir infrared, visible, and upper-tropospheric water va-por sensors, and they also yielded 3-hourly measure-

ments of tropospheric feature-tracked winds when com-bined with operational numerical model forecasts (Vel-den et al. 1997; Nieman et al. 1997). The infrared sen-sors, in combination with operational model output,provided measurements of cloud-top temperature (CTT)and cloud-top pressure (CTP). A sounder package oneach GOES satellite furnished radiance observationsthat were ingested into a nonlinear physical retrievalalgorithm to generate IWV retrievals (Rao and Fuelberg1998; Menzel et al. 1998; Ma et al. 1999). Unlike thepolar-orbiting microwave sensors, the GOES sensorsprovide excellent areal and temporal IWV coverage; thatis, the same geographical area is scanned each hour andcoverage includes both ocean and land, and the effectiveresolution is ;14 km. However, also unlike the micro-wave data, clouds contaminate the GOES retrievals(e.g., Lipton 1998), which is disadvantageous over themoisture-laden Pacific. In addition, the GOES retrievalsdepend on numerical models for a first guess.

Ground-based operational observing systems werealso used in this study. California’s Automated LocalEvaluation in Real Time (ALERT; Mendell 1992) net-work of ;900 rain gauges measured rainfall with 0.04-in. (;1.0 mm) resolution as frequently as every 15 min.The rain gauge data were quality controlled, includingidentification of sites where missing data would havebiased storm-total precipitation analyses. Hourly surfacemeteorological observations were obtained from air-ports and moored and drifting buoys, while 3-h mete-orological data were acquired from ocean-borne ships.Rawinsondes from Oakland (OAK) provided 12-h ther-modynamic and wind soundings.

3. A case study of a polar frontal zone

This section utilizes the 29 GPS dropsondes deployedon 25–26 January 1998 to document the mesoscale spa-tial distribution of water vapor, and its horizontal trans-port, through a modest polar frontal zone over the east-ern Pacific. These dropsondes afforded a uniquely well-resolved (;100 km resolution) picture of the system asit approached the California coast. The dropsonde-basedmesoscale discussion is preceded by a brief descriptionof the synoptic-scale conditions.

a. Synoptic overview

The event studied here was associated with a polarcold-frontal zone that was laid out over the eastern Pa-cific by a major synoptic-scale cyclone with a centralpressure of ;962 mb near 488N and 1358W at 0000UTC 26 January 1998 (Fig. 1a). The storm’s cloud fieldpossessed a wrapped-up comma-cloud head spiralingoutward from the cyclone center and a comma-cloudtail extending offshore of California with the trailingcold front (Fig. 1b). The P-3 flight track and locationsof the 29 dropsondes are shown in Fig. 2, where it isevident that the aircraft transected the cold-frontal cloud


FIG. 1. (a) Sea level pressure (SLP) analysis at 0000 UTC 26 Jan1998. Wind flags 5 25 m s21, full barbs 5 5 m s21, and half barbs5 2.5 m s21. Stand-alone solid dots represent surface observationsof SLP with missing winds. The ocean-based observations are from3-h ship and 1-h buoy reports. Standard frontal notation is used. (b)NOAA GOES infrared satellite image and satellite-derived windsbelow 600 mb (color coded for altitude) at 0000 UTC 26 Jan 1998.Winds and fronts are as in (a). The dashed inset box defines thedomain for Fig. 2.

FIG. 2. Same satellite image as in Fig. 1b, except zoomed in. TheP-3 flight track (white) is time-to-space adjusted to 0000 UTC 26 Jan1998 using a phase velocity of 7.3 m s21 from 3328. The locationsof the 29 dropsondes along the flight track are portrayed with boldwhite dots. Key wind profiler sites are marked with an ‘‘x’’ andlabeled with their three-letter names, and the Oakland rawinsondesite is labeled ‘‘OAK.’’ The large dashed inset box defines the domainfor Figs. 5–7, and the small dashed inset box corresponds to thedomain in Fig. 3.

band three times within the inset mesoscale domain.This track was time-to-space adjusted to 0000 UTC 26January 1998 using a phase speed of the overall frontalsystem (7.3 6 2.0 m s21 from 3328) derived from se-quential satellite images. This slow phase speed meantthat individual dropsonde positions at the beginning andend of the flight were adjusted at most by 110 km.

The cold front produced significant rainfall in north-ern California, with up to 110 mm in 36 h along a 150-km stretch of coast (Fig. 3). The sharp along-coast gra-

dient north and south of this region resulted partiallyfrom the slow-moving nature of the frontal system re-sponsible for much of the rain and reflects the impor-tance of a narrow, but intense, plume of moisture foundoffshore (see sections 3b and 3c). The steady characterof the prefrontal southerly component wind flow, thebrightband height, and the rain rate was captured by thePAA wind profiler within the core of the heavy coastalrainfall (Fig. 4). Much less rain fell at the adjacent wind-profiler sites at BBY and ERK. Rainfall at PAA ceasedfollowing the cold-frontal passage between 2200 UTC25 January and 0000 UTC 26 January and the onset ofpostfrontal westerlies.

b. Mesoscale plan-view perspective

Mesoscale plan-view analyses were constructed fromthe dropsondes in the inset domain of Fig. 2. A sub-jective potential temperature (u) analysis at 1000 mb(Fig. 5) reveals two frontal zones rather than the singlezone depicted in the synoptic analysis (Fig. 1a). Thesingle zone in Fig. 1a is the leading—that is, the south-easternmost—of the multiple cold frontal zones shownin the mesoscale analyses. Each front exhibited a winddirection shift and a ;2 K decrease in u. The southernfront also corresponded with a rope cloud noted in vis-ible satellite imagery (not shown), which helped anchorthe analyzed near-surface position of this front. Theavailability of three cross-front dropsonde legs estab-

JULY 2004 1725R A L P H E T A L .

FIG. 3. Terrain base map with a 36-h accumulated rainfall analysis(mm) between 1800 UTC 25 Jan and 0600 UTC 27 Jan 1998. Thesmall dots show the ALERT rain gauge sites used for the analysis,and the three large dots mark the wind-profiler sites shown in Fig.2; the 36-h accumulated rainfall at these three sites are listed.

FIG. 4. Wind-profiler and rainfall data between 1800 UTC 25 Janand 0600 UTC 27 Jan 1998: (top) Time–height section of hourlywind profiles (flags and barbs are as in Fig. 1) and brightband height(gray-shaded line) measured by the PAA wind profiler. Standard fron-tal notation is used. (bottom) Time series of accumulated rainfall atthe three wind-profiler sites (e.g., PAA, BBY, and ERK).

FIG. 5. The 1000-mb potential temperature (u, K) and frontal anal-ysis at 0000 UTC 26 Jan 1998, based on the P-3 dropsondes (trianglevector heads) and the Oakland rawinsonde (square head); the nu-merical values of u are plotted. Surface winds from nearby windprofilers (star heads; labeled in Fig. 3), and ships and buoys (no heads)are also shown. Wind flags and barbs are as in Fig. 1. The bold gray-shaded line is a projection line for the cross section in Fig. 8; theeight dropsondes included in this cross section are marked with largetriangle vector heads.

lished that the fronts were nearly parallel to each otherand were separated by ;250 km. The northern frontcreated its own narrow, but well-defined, warm tongue,leaving a narrow zone of cooler air between the fronts.Other case studies with offshore aircraft data from theCoastal Observation and Simulation with Topography(COAST; Chien et al. 2001) and CALJET (Ralph et al.2003; Neiman et al. 2004) field studies in the regionhave shown that surface polar cold fronts have appearedas multiple transitions in other events as well. However,the reasons for this behavior are not yet well understood.

A mesoscale depiction of the horizontal IWV distri-bution across the fronts is shown in Fig. 6. An axis oflarge IWV was cradled between the 1000-mb fronts,illustrating the elongated and narrow nature of the IWVplume. Values of IWV exceeding 2 cm were confinedto a region only 424 km wide, and values .3 cmspanned only 70 km in the cross-front direction. Becausethe P-3 flew slightly above 500 mb, the dropsonde-basedIWV values include the contribution of water vapor onlybelow 500 mb. Based on the December–January–Feb-ruary mean vertical profile of water vapor specific hu-midity for the Northern Hemisphere by Peixoto and Oort(1992), the layer below 500 mb contributes ;98% tothe IWV on this seasonal hemispheric average. Simi-larly, rawinsonde launches from Oakland within the


FIG. 6. Vertically integrated water vapor (cm) analysis from thesurface to 500 mb at 0000 UTC 26 Jan 1998, based on the dropsondes(triangles) and Oakland rawinsonde; the numerical values of IWVare plotted. The fronts, bold gray-shaded line, and large triangles areas in Fig. 5.

FIG. 7. Water vapor specific humidity analysis (q; g kg21) at 0000 UTC 26 Jan 1998, based on the dropsondes (triangles) and Oaklandrawinsonde: (a) 1000 and (b) 700 mb. The numerical values of q are plotted. Wind flags and barbs at 1000 and 700 mb are as in Fig. 1.The fronts, bold gray-shaded line, and large triangles are as in Fig. 5. The offshore shading shows IWV . 3 cm.

moisture plume at 1200 UTC 26 January and 0000 UTC27 January documented a mean IWV of 2.64 cm, 96.3%of which was contained below 500 mb. Hence, the drop-sonde IWV values are representative of the full-tropo-spheric values to within ,5%.

The water vapor specific humidity fields at 1000 and700 mb (Figs. 7a and 7b) reveal very different distri-butions of moisture in the lower and middle troposphere.At 1000 mb, the greatest moisture content (.10 g kg21)was ahead of the leading cold front, and a modest de-crease of 2 g kg21 was observed between the fronts. A

sharper moisture gradient resided behind the trailingcold front. In contrast, a plume of high moisture contentat 700 mb (.5 g kg21) was positioned between the1000-mb fronts, with lesser values on the warm side ofthe leading front. The moisture gradient on the polewardor cold side of the moisture plume at 700 mb was muchtighter than its counterpart at 1000 mb. The 700-mbspecific humidity field (Fig. 7b) bears striking resem-blance to the IWV analysis (Fig. 6), whereas the 1000-mb field (Fig. 7a) is decidedly different. The followingsubsection provides a meteorological explanation for thevertical variation in horizontal moisture distribution.

c. Anatomy of an atmospheric river

The vertical structure of the specific humidity (q) andalongfront wind (U) fields across the cold fronts is re-vealed in a cross section (Fig. 8a) analyzed from eightdropsondes along and near the easternmost north–southflight leg between 2106 UTC 25 January and 0021 UTC26 January. The U analysis, together with a companionu analysis (not shown), were used to diagnose the po-sition of the fronts. (The u analysis shows a 2.5-K con-trast in 75 km across the front at 700 mb.) The southernfront was marked by a sloping shear zone of U thatinitially extended upward to ;850 mb, where it becamehorizontal. Farther north, this zone extended up to 600mb, where a distinct region of midtropospheric dryingwas present north of a narrow but deep wedge of mois-ture situated at the northern end of a larger-scale plumeof enhanced prefrontal moisture. Flight-level observa-tions and airborne radar data from the P-3 reveal thatdeep convection was triggered in a line within this nar-row wedge, where a convective updraft and maximumcloud liquid water was also observed (see Fig. 8a). Sub-

JULY 2004 1727R A L P H E T A L .

FIG. 8. (a) Cross section of alongfront wind speed (solid contours, m s21; directed from 2258) and specific humidity (dotted contours, gkg21) along the gray-shaded projection line in Figs. 5–7. Frontal boundaries are marked with bold black lines. Wind flags and barbs are asin Fig. 1. The flight track is portrayed with a thin dotted line near 500 mb, and the maximum cloud liquid water (bracket) and peak convectiveupdraft (arrow) along this flight leg are labeled. (b) Traces of IWV (cm) from the dropsondes for three layers (surface–500 mb, surface–800 mb, and 800–500 mb). The times (UTC) of dropsonde deployments on 25 and 26 Jan 1998 are given.

sidence behind this deep convection is likely responsiblefor the drying ahead of the front above ;800 mbthrough the downward advection of drier air from higherelevations. An axis of maximum U coincided with the

warm side of the front, with evidence of an LLJ at 900mb in the vicinity of a second, shallower wedge of mois-ture at the southern end of the larger-scale plume. Thenorthern front was quite shallow but, as with its southern


FIG. 9. (a) Cross section of horizontal alongfront moisture flux Fh (3105 kg s21; directed from 2258) along the gray-shaded projectionline in Figs. 5–7. Fronts are as in Fig. 8. This section extends southeastward beyond the projection line, because an additional dropsondeat 2036 UTC 25 Jan and the rawinsonde at VBG at 0000 UTC 26 Jan 1998 were included in the analysis; 50-mb-averaged wind velocitiesare shown (flags and barbs are as in Fig. 1). (b) Traces of Fh (solid black; 3105 kg s21) and IWV (gray-shade dashed; cm) from the soundingsfor three layers (surface–500 mb, surface–800 mb, and 800–500 mb). The times (UTC) of dropsonde deployments on 25 and 26 Jan 1998are given.

partner, it too possessed a prefrontal LLJ and water va-por plume. The key features of this cross section werealso found in data from the two other cross-front drop-sonde legs (not shown), thus adding confidence to theanalyzed vertical structure.

Figure 8b reveals the IWV distribution along the crosssection. Although there is a significant northward in-crease in the surface–500-mb IWV toward the southernsurface front, the largest value, 3.09 cm, reflects the

deep moisture in the vicinity of the convection aheadof this front aloft. This is demonstrated by separatingthe IWV into contributions from two layers (surface to800 mb and 800 to 500 mb). The lower layer driedslightly northward from the surface front, but the upperlayer moistened significantly. The analysis of q in Fig.8a shows this well: for example, below 900 mb q de-creased northward from its prefrontal maximum valueof .10 g kg21, while near 600 mb q increased from

JULY 2004 1729R A L P H E T A L .

FIG. 10. Cumulative fraction (%) of total alongfront horizontalmoisture flux for three layers (surface–500 mb, surface–800 mb, and800–500 mb) centered on the core of the atmospheric river shownin Fig. 9.

FIG. 11. Representative dropsonde profiles of (a) specific humidity (g kg21) and meridional wind speed (m s21) for the modest frontalwave of 25–26 Jan 1998 (solid curves) and for a strong, deepening cyclone on 5 Feb 1998 (dashed curves), and (b) meridional moistureflux (3 105 kg s21) for the modest and strong events (solid and dashed curves, respectively), based on the curves in (a). The layer-integrated(1000–500 mb) meridional moisture flux values for the flux curves are shown in (b).

,1 g kg21 to .4 g kg21. The upper layer (800–500mb; see Fig. 8b) experienced a dramatic decrease inIWV behind the convection. Figure 8 clearly illustratesthat the elongated front-parallel plume of maximum sur-face–500-mb IWV (see also Fig. 6) was not associatedwith the position of the southern surface front but re-flected the deep moisture ahead of this front aloft. Thisaccounts for the structural similarity between the plan-view IWV and 700-mb q analyses (Figs. 6 and 7b) andthe difference between the plan-view q analyses at 700and 1000 mb (Figs. 7b and 7a).

The deep prefrontal moisture plume in conjunctionwith the strong front-parallel flow ahead of the southerncold front resulted in an atmospheric river of enhancedhorizontal water vapor transport oriented parallel to thefront. This atmospheric river is analogous to those dis-cussed by Zhu and Newell (1998), who concluded thatfour to five such rivers located around the hemisphereaccount for the total hemispheric meridional moistureflux. However, their analysis is based on global-scalemodel analyses that do not fully capture the narrowtransverse scale of moist plumes such as the one doc-umented here. The horizontal water vapor flux throughthe plane of the cross section in Fig. 8a was determinedby layer averaging the dropsondes into 50-mb incre-ments between 1000 and 500 mb before ingesting thesedata into a variant of the relationship used by Zhu andNewell (1998):

F 5 (rUq) 3 Dz 3 DL,h (1)


FIG. 12. Graphical depiction of the methodology employed to generate composite 1500-km-wide baselines of SSM/I-observed IWV, cloudliquid water, rain rate, and surface wind speed and GOES-observed cloud-top temperature and pressure across moisture plumes measuredby SSM/I satellites over the eastern Pacific during the CALJET winter of 1997/98: (a) length and width criteria of IWV plumes that exceeded2 cm, and (b) baseline geometry criteria relative to the SSM/I swaths for IWV plumes that exceeded 2 cm. The sample SSM/I image ofIWV is from 1610 UTC 23 Mar 1998 during a flood-producing storm in coastal California (White et al. 2003).

where Fh is the alongfront horizontal water vapor flux(kilograms per second) in each 50-mb increment, Dz isthe height range of each 50-mb increment, and r, U,and q are the layer-mean values of air density, alongfrontwind speed, and water vapor specific humidity withineach 50-mb bin. Because the mean spacing betweendropsondes in Fig. 8a is ;100 km, DL is also 100 km.The flux calculations for each dropsonde are assumedto be applicable in a 50-km window on either side ofeach dropsonde. An additional dropsonde at 2036 UTC25 January and a rawinsonde at Vandenberg (VBG) at0000 UTC 26 January 1998 were used to extend thecross section of Fh (Fig. 9a) southward beyond thatshown in Fig. 8a in an effort to document the completelateral extent of the atmospheric river.

Figure 9a shows a well-defined atmospheric river onthe warm side of the southern cold front. Water vaportransport within this river exceeded 80 3 105 kg s21

within individual 50-mb bins below 800 mb, and theriver extended up to ;600 mb in the vicinity of theconvection. A maximum value of .90 3 105 kg s21

was observed near the southern front’s LLJ. The north-ern cold front also possessed a local maximum of watervapor transport, though it was confined to a shallowlayer below 800 mb. Traces of Fh along the cross section

are presented in Fig. 9b for three bulk layers (surface–500 mb, surface–800 mb, 800–500 mb), together withcompanion traces of IWV (see also Fig. 8b). In eachbulk layer the character of the Fh trace mirrored that ofthe IWV. However, the Fh traces had a sharper peakassociated with the atmospheric river, including that as-sociated with the southern front’s LLJ below 800 mband that tied to the convection between 800 and 500mb.

The transverse scale of this atmospheric river wasassessed by determining the cumulative fraction of Fh

along a 1000-km distance of the cross section, centeredon the river core (Fig. 10). Between the surface and 500mb, 75% of Fh was contained within the river in a swathonly 565 km wide. In the upper portion of this layerbetween 800 and 500 mb, the same cumulative fractionwas only 440 km wide. Six dropsondes in the core ofthe river where 75% of the surface–500-mb alongfrontflux resided (2106 to 2253 UTC) possessed a meridionalflux of 1500 3 105 kg s21, or 20% of the global averageacross at 358N reported in Zhu and Newell (1998). Thesein situ observations support Zhu and Newell’s (1998)model-based conclusion that the hemispheric meridionalmoisture flux in the extratropics is accomplished almostentirely within narrow filaments. In our case the one

JULY 2004 1731R A L P H E T A L .

TABLE 1. A listing of the SSM/I and GOES baselines used in thesatellite compositing analysis.

Date/time ofSSM/I baseline

trios for1997/98 winter

(overpass at 308N)

MM/DDTime

(UTC)

Central SSM/Ibaseline information

Lat (8N)* Lon (8W)*Orienta-

tion (8)**

Timeoffset

betweenGOES

andSSM/I(min)

11/1911/2011/2311/2311/2311/2311/2411/2411/24

025504180224052314521608021214391554

33.5631.8825.1524.7725.6926.3427.1028.8430.09

125.48123.85132.59133.79131.40129.93128.63127.44124.89

325348331334334332330335339

—————————

12/0212/0312/0312/0312/0412/0512/0512/1012/3012/3012/3012/3112/3112/31

05130203050017281715013904350515032006111839030705591535

38.9929.9828.8928.8929.9228.8427.5940.4033.6228.5737.2526.2325.7525.26

136.99130.21131.18128.03124.61123.64125.70140.63149.04148.88143.12144.32143.99142.85

270314327311298303311302270290266310314318

47—603245

—8545

22021123927

1—

01/0701/1001/1101/1201/1501/1501/1601/1601/1601/16

0342171015020416174418430315060315421830

35.7331.2835.9029.3833.6237.0434.0534.7636.1134.92

132.81143.72131.89141.77151.38144.43147.79147.96141.60141.60

008323335333327327330331322323

242———16

24321523

242230

01/1701/1701/1701/2201/2301/2401/2501/2501/2501/2601/2601/2601/2801/3101/31

030205501818185703310315030615341820025404300540164301530504

35.1935.2437.9627.0524.6631.6129.1726.8926.6126.8924.9926.6136.0128.1332.42

142.64144.10137.26150.34149.37126.84145.02141.88141.22140.57143.07139.65131.40125.86152.03

336332291320329326317330331326329320292305331

2210

218257—

21526

234220

6—207767

—02/0102/0102/0102/0102/0102/0202/0202/0202/0402/0402/04

03220607155017171834031004130555151316371757

27.0526.8926.0225.3125.6424.8832.8623.8435.1934.5933.89

147.57146.54145.57146.43144.86143.45124.18142.58133.63133.30133.52

347346343336337322347319302307316

——————

273—

21383

3

TABLE 1. (Continued )

Date/time ofSSM/I baseline

trios for1997/98 winter

(overpass at 308N)

MM/DDTime

(UTC)

Central SSM/Ibaseline information

Lat (8N)* Lon (8W)*Orienta-

tion (8)**

Timeoffset

betweenGOES

andSSM/I(min)

02/0502/0502/0602/0602/0602/0802/0802/0802/1202/1202/1202/1202/12

0233051702210342050501550315044004030532151616301759

29.4929.9223.1423.2522.9828.3028.5128.2430.6330.4129.9830.4730.03

136.18135.04131.35130.86130.53127.49126.03125.97139.81138.39136.99133.08135.09

328326307311313328328329323327347344348

2743

———65

21580

26328

216290

102/1302/1302/1402/1402/2202/2202/2203/0103/0103/01

0237051914521734022805090509024403361512

29.7630.0929.3829.1130.2030.1430.0337.8037.3637.20

136.39134.60130.69128.85134.93132.65132.76137.43137.64133.08

356353333327333333334332333328

2341

82632515116

236212

03/0703/0703/0703/0703/0703/0803/0803/0803/08

031305501540162418170300034305371610

28.5730.2028.8435.2429.6529.1131.1229.7631.07

146.22143.40142.64130.69140.03142.04132.21139.22130.48

329328335318338339348347360

21310

240284217

0243

23—

03/1003/1003/1103/1203/1403/1403/1403/1503/2203/2203/2203/2303/24

1503174019221439040315561832035003320606162316101556

34.9237.5326.5633.8337.0431.8835.8436.0120.7522.7034.7033.4530.25

132.05128.47152.24126.62143.50147.03141.39138.40149.10145.24130.15127.00124.99

319316352302335333319330330320319308310

2320

——

263256232250——

283270256

* The latitude and longitude values are valid for the center gridpoint of each central baseline.

** The orientation represents the azimuth pointing angle on thecold side of each central baseline.

filament was roughly 600 km wide or ;1.8% of theearth’s circumference at 358 latitude.

Depending on the strength and orientation of the tro-pospheric winds within a region of similarly high mois-ture content, the meridional water vapor flux can varysignificantly. For example, Fig. 11a shows dropsonde


FIG. 13. The positions of all 312 baselines (thin lines) transecting104 moisture plumes measured by SSM/I satellites between 19 Nov1997 and 24 Mar 1998. These baselines were used to create thecomposite baseline (bold line) of SSM/I data. A 2000-km-long mois-ture plume (bold arrow) is depicted intersecting the composite base-line at a right angle. The tropical IWV reservoir (IWV $ 3 cm;shaded) is bounded by a solid gray-shaded line. The position of thisline (IWV 5 3 cm) was determined from an average of every tenthdaily IWV chart during the period of interest.

data from the cores of two atmospheric rivers, each withthe same IWV content of 2.7 cm. The dropsonde on 25January observed relatively weak meridional flow ofless than 18 m s21 within the modest frontal zone, whilethe dropsonde on 5 February measured strong meridi-onal flow approaching 40 m s21 within an intense cy-clone. The resulting meridional moisture-flux profileswere therefore quite different for these two cases (Fig.11b), assuming each sounding was representative ofconditions in a 100-km-wide swath. The total surface–500-mb meridional flux was 3 times greater for thestrong cyclone than for the modest front even thoughthe IWV was the same. Yet, the weaker of these systemswas still strong enough to account for a substantial frac-tion of the meridional fluxes across the earth’s midlat-itude belt (Zhu and Newell 1998).

4. Winter-season satellite analysis of moistureplumes over the eastern Pacific

The case study of the IWV plume (Fig. 6) and itsatmospheric river (Fig. 9) on 26 January highlights theriver’s transverse characteristics, vertical structure, andelongated nature. While the overall horizontal water va-por flux magnitude of this event matches the model-derived value that Zhu and Newell (1998) suggested istypical of an individual atmospheric river, it is possiblethat the detailed case study was not characteristic ofmost atmospheric rivers. Thus, it is important to assess

the representativeness of this single case using obser-vations that can quantitatively document at least onekey aspect of most atmospheric rivers. The combinationof polar-orbiting and geostationary satellite observa-tions provides the suitable spatiotemporal coverage forthis purpose. However, because satellites are unable todirectly observe Fh (sections 4b and 4c), the analysisfocuses on other key signatures of atmospheric rivers,that is, patterns of IWV and other satellite-derived prod-ucts such as CLW, RR, CTT, and CTP. The use of IWVas a proxy for determining the positions and widths ofatmospheric rivers is based upon the close correlationbetween IWV and vertically integrated horizontal watervapor fluxes shown in the case study and on the generalconnection that is to be expected between these variablesin extratropical cyclones. Henceforth, the term atmo-spheric rivers will be used when discussing the satellite-derived IWV results. The procedure used in section 4ato identify cases focuses on narrow, elongated IWV fea-tures exceeding 2-cm IWV. All SSM/I IWV images overthe eastern Pacific during the CALJET winter of 1997/98 were examined for the presence of narrow plumesof high IWV. Representative transects oriented orthog-onal to those plumes that met objective geometric cri-teria (section 4a) were then synthesized into a composite1500-km-wide transverse baseline of SSM/I-observedIWV, CLW, RR, and SPD. GOES-observed CTT, CTP,and IWV along these same baselines were then alsocalculated. Events meeting the objective criteria wereidentified on 46 dates during the same winter as the casestudy.

a. Compositing methodology using SSM/I and GOESobservations

The winter-season satellite analysis of elongated tro-pospheric moisture plumes over the eastern Pacific dur-ing CALJET encompassed a 137-day period between15 November 1997 and 31 March 1998 in a domainbounded by 08–458N latitude and 1058–1608W longi-tude. The initial analysis step for generating a compositecross-plume baseline required the perusal of daily av-eraged plan-view images of SSM/I-observed IWV formoisture plumes whose core values exceeded 2 cm for$2000 km in the along-plume direction and #1000 kmin the cross-plume direction. Because the daily averagedimages covered the entire eastern Pacific, the selectionmethodology allowed identification of IWV plumes ofany orientation. The threshold of 2 cm was chosen be-cause the atmospheric river in Fig. 9a was roughlybounded by IWV 5 2 cm. For those days when moistureplumes conformed to the aforementioned dimensions,the individual SSM/I IWV orbital swaths that composedthe daily averages were then inspected to ascertain ifthese plumes were authentic (e.g., Fig. 12a). (Readilyidentifiable artifacts in these daily maps can arise be-cause of averaging that was applied when there wasspatial overlap from overpasses that occurred at differ-

JULY 2004 1733R A L P H E T A L .

ent times during the same day. However, the analysespresented henceforth are based entirely on the individualorbital swaths and thus are not affected by this aver-aging.) Using this methodology, 175 individual SSM/Iimages possessed an elongated ($2000 km 3 #1000km) moisture plume exceeding 2 cm on 58 out of 137days, that is, on 42% of the days.

Figure 12b highlights the next set of steps that wasimplemented to create a composite transverse baselineorthogonal to the IWV plumes. To ensure that the com-posite baseline extended well beyond the narrow(#1000 km) corridor of IWV $ 2 cm, individual base-lines that made up the composite were required to span750 km on each side of the plume core. For each of the175 SSM/I images, a pair of 1500-km-long ‘‘bookend’’baselines were first determined, based on the orientationof the ;1400-km-wide satellite swath relative to theIWV plume. The bookend baselines were positionedsuch that their centers resided in the core of the IWVplume, they were oriented perpendicular to the plume,and one end of each baseline terminated at the edge ofthe satellite swath. If the distance between the bookendbaselines in the core of the IWV plume exceeded 500km, then a center baseline was cut orthogonal to theIWV plume midway between the bookend baselines.Then, two more baselines were cut orthogonal to theplume 250 km on each side of the center baseline. Eachtrio of baselines enclosed by the bookend baselines werethen used in the subsequent SSM/I compositing of IWV,CLW, RR, and SPD. These baselines possessed a 25-km grid spacing that is only slightly smaller than thenative resolution of the SSM/I observations. Using triosof baselines in a 500-km along-plume segment ensuresthat representative sampling across each moisture plumewas obtained. These additional criteria reduced the num-ber of days and images with elongated IWV plumes to46 and 104, respectively. Hence, the composite SSM/Ibaselines of IWV, CLW, RR, and SPD are each com-posed of 312 individual transects. The time, position,and orientation of the central baseline of each triplet areprovided in Table 1.

The locations of all 312 baselines are shown in Fig.13 within the satellite analysis domain. These baselinesreside in the midlatitudes north of the tropical moisturereservoir where IWV $ 3 cm. The reservoir’s northwall migrated systematically southward during the win-ter. The position of the 1500-km-wide composite base-line is also shown in Fig. 13. The composite possessesa northwest-to-southeast orientation of 3258 (cold) to1458 (warm), thus revealing that the mean moistureplume was directed from southwest to northeast towardCalifornia.

In an effort to place the composite baseline into syn-optic-scale context, composite plan-view analyses weregenerated using daily global gridded data from the Na-tional Centers for Environmental Prediction–NationalCenter for Atmospheric Research (NCEP–NCAR) re-analysis project (e.g., Kalnay et al. 1996). Those days

with at least one suitable SSM/I IWV swath from whichbaseline trios were obtained (46 days total) were in-cluded in the synoptic compositing (Fig. 14). Althoughthe coarse grid spacing of the NCEP–NCAR fields (2.58latitude by 2.58 longitude) could not resolve the narrowatmospheric rivers themselves, the synoptic compositesdo provide the large-scale context. Specifically, thecomposite baseline resided south of a deeply occludedcyclone over the Gulf of Alaska, crossed a warm andmoist axis at 925 mb, transected an isolated and elon-gated region of large specific humidity at 700 mb, wasdownstream of the mean 500-mb trough position (with-out exception, the SSM/I IWV plumes that were thebasis for the synoptic composites were situated on thedownstream side of 500-mb extratropical trough axes),and resided in the right exit region of a 300-mb jet. Tofurther define the synoptic context, approximate frontalpositions (shown in Fig. 14) were derived from distinctfeatures in the 925-mb thermal and specific humidityfields. This synoptic compositing and frontal analysisplaces the composite baseline nearly orthogonal to the925-mb polar cold-frontal zone and indicates that thecomposite SSM/I IWV maximum at the midpoint of thebaseline was located in the cyclone warm sector, as wasalso true in the detailed case study of 26 January 1998(section 3). Based on the composite analysis results, theIWV plumes were found within warm conveyor beltsof extratropical cyclones (e.g., Browning 1990; Carlson1991). Although it is theoretically possible for an elon-gated IWV plume to exist without it marking an at-mospheric river, such as in a col region of weak wind,the synoptic compositing places the events studied herein a large-scale setting where these conditions are ex-tremely unlikely to occur.

We assessed the cloud-top temperature and pressurecharacteristics across these moisture plumes by creatingcomposite baselines of CTT and CTP from GOES datathat were available north of ;228N latitude on a 3-hbasis from 1 December 1997 to 31 March 1998. A com-posite baseline of GOES IWV was also synthesized.The positions of the individual baselines used to createthe SSM/I cross-plume composites were also used tocreate the companion GOES composites, provided thetime difference between SSM/I and correspondingGOES images was less than 1.5 h. Because of the small-er spatial window of GOES data availability relative tothe SSM/I observations, the GOES composites of CTT,CTP, and IWV contain only 196 samples rather than312. The average time difference between the 196GOES images and companion SSM/I images is only23.2 min, with a standard deviation of 43.1 min. Table1 indicates which samples were used to create the GOEScomposites. To make meaningful comparisons betweenthe SSM/I and GOES composites, 196-sample subsetcomposites of the SSM/I data were also constructed.

b. SSM/I composite resultsThe SSM/I-observed composite baseline of IWV con-

structed from the 312 transects with IWV $ 2 cm is


JULY 2004 1735R A L P H E T A L .

←

FIG. 14. Composite analyses based on the NCEP–NCAR daily reanalysis global gridded dataset for the 46 days with baselines shown inFig. 13: (a) SLP (mb), (b) 925-mb temperature (T; 8C), (c) 925-mb specific humidity (q; g kg21), (d) 700-mb specific humidity (q; g kg21),(e) 500-mb geopotential height (Z; m), and (f ) 300-mb isotachs (m s21) and wind vectors. Frontal interpretations are based on the compositeanalyses. Standard frontal notation is used. The bold gray line transecting the cold front in each panel is the composite baseline shown inFig. 13.

shown in Fig. 15a. The mean and standard deviation ofthe width scale of IWV that exceeds 2 cm is 388 6 121km, which is similar to the 424-km width scale observedin the case study and much narrower than the 1000-kmcross-plume criterion that was employed. Two subsetcomposite baselines of IWV were synthesized from the312 transects and are also presented in Fig. 15a. Thesesubset composites are composed of transects whose coreregions of IWV exceeded 3 and 4 cm, and their re-spective sample sizes are 84 and 9 (i.e., 27% and 9%of 312, respectively). The 21, 31, and 41 cm tracescontain peak values of 2.81, 3.41, and 4.19 cm, re-spectively (Table 2), and the peak value of the 41 cmtrace is 49% larger than that of the 21 cm trace. How-ever, .200 km beyond the core region, the IWV valuesof these composite traces are quite similar. Thus, themagnitude of IWV within the core of the moistureplumes is not well correlated with IWV content outsidethe plume cores. As expected, the endpoints or wingsof these composite traces show a poleward decrease inthe background moisture content.

Companion composite traces of CLW and RR (Figs.15b and 15c) show well-defined peak values in the coreof the 21, 31, and 41 cm composite IWV plumes(Table 2 lists these values). The peak value of CLW inthe 21 and 41 cm IWV traces increases by 192%, eventhough the corresponding IWV peak increases by only49%. The peak value of RR increases by an even larger463%. Hence, from a statistical perspective, a relativelymodest increase of IWV within the core of a moistureplume yields a much more substantial increase in thecorresponding cloud liquid water and rain-rate peaks.This suggests that nonlinear processes, such as thosedriving vertical circulations associated with fronts and/or convection, are enhanced in IWV plumes with greatermoisture content. The wings of the CLW and RR tracespossess a background trend opposite to that observedin the IWV traces. The larger values of CLW and RRon the northern wing quite likely reflect the presence ofslantwise baroclinic ascent on the cold side of the IWVplumes.

The narrow transverse character of the compositeIWV traces was quantified by calculating the cumulativefraction of the IWV perturbation centered on the plumecore.2 This analysis was repeated for the composite

2 The total perturbation is defined by the area bounded beneath theIWV curve and above a linear-trend line connecting the endpoints ofthis curve. Fractional perturbation elements are 50 km wide centeredon the plume core, and they are also bounded by the curve and trendline.

CLW and RR traces. For the average of the three com-posite IWV traces, ;75% of the perturbation to thebackground IWV is contained in a narrow region 417.3km wide (Fig. 15d; Table 3). It is noteworthy that thiswidth is independent of the core value of IWV (Table3). Roughly 75% of the perturbation to the backgroundCLW and RR is contained in even narrower ribbonsonly 175.7 and 140.7 km wide, respectively (Figs. 15eand 15f; Table 3). The difference in width scale betweenthe IWV traces and the CLW and RR traces suggeststhat processes leading to the development of moistureplumes operate on a much larger scale than those re-sponsible for the generation of cloud liquid water andrainfall. Based on preliminary numerical model results(not shown), the primarily geostrophic horizontal frontalconfluence contributes significantly to the developmentof IWV plumes, and ageostrophic secondary verticalfrontal circulations focus the narrower CLW and RRpeaks within these plumes. A detailed examination ofthese numerical results will be the focus of a futurestudy.

Because rainfall is often observed within elongatedIWV plumes but interferes with SSM/I-based wind re-trievals, the SPD coverage in this important region isdeficient (Fig. 16a). Only 25% of SPD measurementsin the core of the 21 cm IWV composite trace weredeemed reliable by the SSM/I retrieval algorithm. A310-km window of ,80% SPD coverage straddles theplume core (Fig. 16b), corresponding to a compositeCLW of $0.08 kg m22. The asymmetric position of thiswindow relative to the IWV core reflects the prevalenceof rainfall on the cold side of the core and the likelihoodthat slantwise baroclinic ascent played a role in thiscold-side bias. The percentage of successful SPD re-trievals decreased to 18% and 0% in the core of the 31and 41 cm plumes, respectively, thus revealing thatrainfall and its detrimental impact on the SPD retrievalsare more prevalent in moisture plumes with greater IWVcontent.

Composite traces of IWV and CLW composed of in-dividual transects where no rain was observed (i.e.,those traces composed entirely of valid SPD observa-tions—a subsample of 35) are compared to their 312-sample counterparts (i.e., the 21 cm IWV plumes; Fig.17a). The composite IWV trace with valid SPD obser-vations mirrors the 21 cm trace, but the composite CLWtrace composed of valid SPD data is quite different.Namely, the peak value of ;0.08 kg m22 within theCLW subsample composite is 1/6 the magnitude ofCLW corresponding to the 21 cm trace, and this value


FIG. 15. Composite SSM/I-observed baselines of (a) IWV (cm), (b) CLW (kg m22), and (c) RR (mm h21) across those moisture plumescontaining $2 cm (solid line), $3 cm (dashed line), and $4 cm (dotted line) of IWV observed by SSM/I. Cumulative fraction (%) ofperturbation (d) IWV, (e) CLW, and (f ) RR, centered on the core of the composite of SSM/I-observed moisture plumes containing $2, $3,and $4 cm of IWV.

JULY 2004 1737R A L P H E T A L .

TABLE 2. Maximum mean values of SSM/I-observed IWV (cm), CLW (kg m22), and RR (mm h21), and minimum mean values of GOES-observed CTT (K) and CTP (mb) in the core of the composite moisture plumes containing greater than 2, 3, and 4 cm of IWV observedby SSM/I. The sample sizes used to obtain the mean values are also provided.

Composite IWVcore magnitude

SSM/I observations

IWV(cm)

CLW(kg m22)

RR(mm h21)

Samplesize

GOES observations

CTT(K)

CTP(mb)

Samplesize

21 cm plumes31 cm plumes41 cm plumes

2.813.414.19

0.480.771.40

1.303.387.32

31284

9

251.4244.1235.3

481.2392.1321.7

19652

9

TABLE 3. Width scales (km) of the 75% cumulative fraction ofperturbation IWV, CLW, and RR across those composite moistureplumes containing greater than 2, 3, and 4 cm of IWV observed bySSM/I. The average width scales are also shown.

Composite IWVcore magnitude

Width scale (km) for 75% cumulative fraction

IWV CLW RR

21 cm plumes31 cm plumes41 cm plumes

415.6408.9427.3

184.3164.5178.4

133.4136.6152.0

Average 417.3 175.7 140.7

FIG. 16. (a) Percent wind speed coverage (observed by SSM/I)across those moisture plumes containing $2 cm (solid line), $3 cm(dashed line), and $4 cm (dotted line) of SSM/I-observed IWV. (b)Composite baselines of IWV (solid; cm) and CLW (dashed; kg m22)across those moisture plumes containing $2 cm of SSM/I-observedIWV. The gray-shaded bars correspond to the width scale of SSM/Iobservations with ,80% wind speed coverage for IWV plumes $ 2cm.

is consistent with the CLW threshold shown in Fig. 16b.A composite trace of SPD derived from the 35-transectsubsample (Fig. 17b) contains a modest wind speedmaximum of 6.4 m s21 within the core of the compositeIWV plume. Using the case-study results as a guide,this maximum represents a pre-cold-frontal LLJ. A sec-ond composite SPD trace, composed of individual tran-sects whose core IWV values exceed 4 cm (Fig. 17b),contains a stronger LLJ perturbation and stronger windsoverall, but the SPD values are missing in the core. Acomparison of these traces suggests that moistureplumes with a larger IWV content are accompanied bystronger winds and, therefore, a nonlinear increase inhorizontal moisture flux. However, because the SSM/ISPD retrievals are degraded in rainfall, and rainfall iscommonly observed in elongated moisture plumes (es-pecially in plumes accompanied by stronger surfacewinds), the combined SSM/I dataset of IWV and SPDcannot provide reliable low-level horizontal water vaporflux estimates within these plumes.

c. GOES composite results

To assess the cloud-top temperature and pressurecharacteristics across the moisture plumes describedabove, composite baselines of GOES CTT and CTPwere constructed when concurrent GOES and SSM/Idata were available, that is, from 196 transects north of228N latitude between 1 December 1997 and 31 March1998. Composite traces of SSM/I-observed IWV andCLW for this 196-transect subsample (where the corevalue of IWV in each transect exceeded ,2 cm) beara striking resemblance to their 312-transect counterparts(Fig. 18a), except for slightly lower values of IWV atthe wings and a slightly higher value of CLW in the


FIG. 17. (a) Composite baselines of IWV (solid lines, cm; observedby SSM/I) and CLW (dashed lines, kg m22; observed by SSM/I)across all moisture plumes containing $2 cm of SSM/I-observedIWV (black lines), and across those moisture plumes containing $2cm of SSM/I-observed IWV with valid SSM/I-observed wind speedobservations (gray shaded lines). (b) Composite baselines of windspeed (m s21; observed by SSM/I) across those moisture plumescontaining $2 cm of SSM/I-observed IWV with valid wind speedobservations (solid), and across those moisture plumes containing $4cm of SSM/I-observed IWV (dashed; the broken portion of the curveis where no wind speed data were available from any of the individualtransects).

FIG. 18. (a) Composite baselines of IWV (solid lines, cm; observedby SSM/I) and CLW (dashed lines, kg m22; observed by SSM/I)across all moisture plumes containing $2 cm of SSM/I-observedIWV (black lines), and across those moisture plumes containing $2cm of SSM/I-observed IWV when GOES data were also available(gray shaded lines). (b), (c) Composite baselines of CTT (K; observedby GOES) and CTP (mb; observed by GOES) across those moistureplumes containing $2 cm (solid line), $3 cm (dashed line), and $4cm (dotted line) of SSM/I-observed IWV.

core. These minor differences arose due to the latitudinaldependence of the compositing results (see section 4d),since the southernmost transects that make up the 312-sample composites were not included in the 196-samplecomposites because of the absence of GOES data southof 228N. Because the 196- and 312-sample compositesare so similar (as are the composites corresponding tothe 31 and 41 cm IWV plumes; not shown), mean-ingful comparisons can be made between the GOES andSSM/I results.

JULY 2004 1739R A L P H E T A L .

FIG. 19. GOES satellite image of IWV (cm; color coded) and cloudsat 0000 UTC 26 Jan 1998. The dropsonde-derived IWV analysiswithin the dashed inset box (see Fig. 6) is overlaid.

Composite traces of GOES CTT and CTP (Figs. 18band 18c) show that the coldest and deepest cloud topsare situated in the core of the IWV plumes. In addition,the cloud tops become substantially colder and deeperwithin the core of the IWV plumes when the moisturecontent within these plumes increases (see also Table2). Specifically, the core value of CTT decreases from251.4 K in the 21 cm composite IWV trace to 235.3K in the 41 cm trace, and the corresponding core valueof CTP decreases from 481.2 to 321.7 mb between the21 and 41 cm composite traces. The anticorrelationbetween the core moisture content in the composite IWVplumes and the embedded cloud-top core values mayarise, at least in part, because of latent heating: as themoisture content within IWV plumes increases, latentheating triggered by forced ascent will be more vigor-ous, resulting in enhanced upward motion and saturatedconditions deeper in the troposphere. Those plumes withgreater moisture content may also be indicative of stron-ger and deeper baroclinic zones and attendant verticalcirculations that cause saturated conditions deeper intothe troposphere. The wings of the composite CTT tracecontain a poleward decrease in cloud-top temperatureof ;17 K, quite likely reflecting the background de-crease of temperature with latitude.

Because elongated moisture plumes represent regionsof deep cloudiness, the GOES satellites do not providereliable IWV retrievals there, despite the excellent spa-tiotemporal coverage of the GOES IWV data in cloud-free regions. For example, the GOES satellite image andsuperimposed dropsonde-based IWV analysis at 0000UTC 26 January 1998 (Fig. 19) reveal a narrow plumeof enhanced IWV (.3 cm in a swath ,100 km wide)

associated with an atmospheric river residing beneaththe cold-frontal cloud band. These clouds hide the coreof the moisture plume from the GOES retrievals, whichsee at most 2.2 cm of IWV at the periphery of thisregion. Thus, in this case the core of the moisture plumehad IWV values 40% greater than those seen by theGOES retrieval. From a statistical perspective, the com-posite baselines of GOES IWV data for the 21, 31,and 41 cm SSM/I-observed IWV plumes have 0%GOES coverage in the core of the composite moistureplumes (Fig. 20a), and the width scale of the data voidin the core region expands with increasing moisture con-tent. Hence, there is no GOES IWV coverage withinthe core region of the narrow moisture plumes or at-mospheric rivers, and even the wings of the compositetraces are characterized by GOES IWV coverage of only;25%–55%.

To explore if the GOES IWV retrievals at the pe-riphery of moisture plumes are able to help determinethe moisture content within the core of those plumes,an autocorrelation analysis was performed on the 312SSM/I transects that make up the 21 cm composite IWVtrace. Specifically, a correlation analysis was carried outbetween the 312 IWV observations at the center or coregrid point at X 5 750 km and the 312 IWV measure-ments at each of the 61 grid points from X 5 0 km toX 5 1500 km. The results are presented in Fig. 20b andshow a narrow region of large autocorrelation cradledwithin the envelope of poor GOES IWV coverage (Fig.20a). Significantly, the maximum IWV value that char-acterizes a moisture plume or atmospheric river is nearlyuncorrelated with values just 200 km away from thecore. This has important implications not only for theoperational nowcasting use of GOES IWV in the vi-cinity of moisture plumes, but also for utilizing GOESIWV data in numerical model initializations and dataassimilation. However, the GOES CTT and CTP resultssuggest that the correlations between CTT, CTP, andIWV could prove to be operationally valuable. In aneffort to help fill this gap in GOES IWV coverage, Sco-field et al. (2000) developed a composite precipitablewater vapor product that takes advantage of both GOESand SSM/I satellite datasets as well as operational glob-al-scale numerical model analyses of water vapor. How-ever, this composite product has not been evaluated here.

d. Characteristics of atmospheric rivers as a functionof latitude

The winter-season satellite analysis and case studyhave documented fundamental transverse characteristicsof elongated moisture plumes and associated atmo-spheric rivers. In an effort to also ascertain key along-plume characteristics, the impact of latitude on moistureplumes was evaluated by generating composite SSM/Iand GOES baselines based on the latitude of the centergrid point of each of the 312 transects. The individualtransects were grouped into four 58 wide latitude bins


FIG. 20. (a) Percent IWV coverage (observed by GOES) acrossthose moisture plumes containing $2 cm (solid line), $3 cm (dashedline), and $4 cm (dotted line) of SSM/I-observed IWV. (b) Auto-correlation between the peak SSM/I-observed IWV value (i.e., thecenter grid point) from each of the 312 baselines and the IWV valueat each 25-km resolution grid point from each of these baselines.

FIG. 21. Composite baselines of SSM/I-observed (a) IWV (cm),(b) CLW (kg m22), and (c) RR (mm h21), across all moisture plumescontaining $2 cm of SSM/I-observed IWV within four discrete lat-itude bins (latitudes $358N, solid; 308–358N, dashed; 258–308N, dot-ted; ,258N, dot–dash).

(i.e., 208–258N, 258–308N, 308–358N, and 358–408N),and the composite baselines were then created fromthese transects for each bin. The mean orientation ofthese baselines does not change appreciably with in-creasing latitude (ranging from 3168–1368 to 3298–1498), and they mirror the orientation of the 312-transectcomposite baseline for SSM/I IWV $ 2 cm (3258–1458).Because GOES data were not available south of 228N,we were unable to construct GOES composites for thesouthernmost bin.

JULY 2004 1741R A L P H E T A L .

TABLE 4. Maximum mean values of SSM/I-observed IWV (cm), CLW (kg m22), and RR (mm h21), and minimum mean values of GOES-observed CTT (K) and CTP (mb) as a function of latitude in the core of the composite moisture plumes containing greater than 2 cm ofIWV observed by SSM/I. The sample sizes used to obtain the mean values are also provided.

Latitude bin

SSM/I observations

IWV(cm)

CLW(kg m22)

RR(mm h21)

Samplesize

GOES observations

CTT(K)

CTP(mb)

Samplesize

,258N258–308N308–358N358–408N

2.812.812.852.76

0.340.410.520.63

0.260.591.722.52

30121

9962

—259.5251.5240.2

—579.2475.1379.6

0617659

TABLE 5. Width scales (km) of SM/I-observed IWV $ 2 cm, CLW $ 0.2 kg m22, and RR # 0.5 mm h21, and GOES-observed CTT #260K and CTP $ 600 mb as a function of latitude across those composite moisture plumes containing greater than 2 cm of IWV observedby SSM/I.

Latitude binIWV

$2 cmCLW

$0.2 kg m22

RR$0.5 mm h21

CTT#255 K

CTP#600 mb

,258N258–308N308–358N358–408N

449.8370.7351.3298.0

114.8121.1182.2267.2

0.030.7

159.3252.8

—21.1

410.3885.6

—118.0499.1768.0

The peak moisture content within the core of theSSM/I IWV composites (Fig. 21a) does not vary ap-preciably between 208 and 408N; an average peak valueof 2.81 6 0.04 cm is observed (see also Table 4). How-ever, the width scale of IWV that exceeds 2 cm in thesecomposites decreases systematically from south to northfrom 449.8 to 298.0 km (Table 5). Because the casestudy and synoptic compositing imply that elongatedmoisture plumes coincide with frontal zones, the north-ward decrease in width scale suggests that frontal con-fluence may act to contract the lateral scale of the IWVplumes with increasing distance downstream of thesouthern terminus of the fronts. The IWV content ineach wing of these composites decreases systematicallywith increasing latitude, thus documenting the north-ward decrease in background water vapor. The CLWand RR composite traces (Figs. 21b and 21c) show lat-itudinal variations primarily within the core of the mois-ture plumes. Specifically, the peak value of CLW in thecore region increases by nearly a factor of 2 betweenthe southernmost and northernmost baselines, while thepeak value of RR increases by an order of magnitudefrom south to north (see also Table 4). In addition, thewidth scale of CLW and RR increases with increasinglatitude (see also Table 5), opposite that of the compositeIWV traces. These CLW and RR characteristics reflectthe fact that frontal circulations are generally strongerand broader in the midlatitude storm track than they arein the subtropics. Jet dynamics may have played a rolein the latitudinal dependence of CLW and RR, giventhat the more southern moist-plume transects were quitelikely located in the often subsident right exit region(Keyser and Shapiro 1986) of quasi-linear jet streams(see Fig. 14f) where cloud formation and rainfall were

suppressed. Finally, the latitudinal variation of CLW andRR may be partly attributable to Hadley cell subsidence,which can extend northward to ;308N in the NorthernHemispheric winter mean (Peixoto and Oort 1992, 157–159), thus preferentially impacting the southernmostmoist-plume transects.

The GOES composite traces of CTT (Fig. 22a) showa systematic northward decrease in cloud-top temper-ature across these traces southeast of X ø 250 km thatpartly reflects the poleward decrease in temperature.Significantly, a northward increase in the cold cloud-top perturbation within the core of the moisture plumesis also observed across these traces. The northernmostcomposite, which possesses a core CTT value of 240.2K, is characterized by a well-defined cold perturbationof 15–20 K. In contrast, the core of the southernmostcomposite is nearly 20 K warmer (see also Table 4) anddoes not exhibit a substantive cold perturbation. Hence,elongated moisture plumes are more easily distinguish-able using GOES CTT observations at the more northernlatitudes. The composite width scale of CTT # 260 Kincreases systematically from south to north (Table 5).

Companion composite traces of CTP (Fig. 22b) showthat clouds within the moisture core of the northernmostcomposite penetrate much deeper into the troposphere(379.6 mb) than clouds within the core of the centraland southern composites (475.1 and 579.2 mb, respec-tively; see also Table 4), and the CTP perturbation andits width scale (see also Table 5) increase with increasinglatitude. The connection between both the CTP and CTTcharacteristics and latitude in the core of the moistureplumes was quite likely driven by physical processesthat also impacted the latitudinal dependence of theCLW and RR fields.


FIG. 22. Composite baselines of GOES-observed (a) CTT (K) and(b) CTP (mb) across all moisture plumes containing $2 cm of SSM/I-observed IWV within four discrete latitude bins (latitudes $358N,solid; 308–358N, dashed; 258–308N, dotted; ,258N, no data).

5. Conclusions

Observations presented in this paper document thenarrow nature of the regions responsible for most of themeridional water vapor flux in the midlatitudes. Theseconcentrated regions of enhanced flux, also known asatmospheric rivers, are typically coupled with polar coldfronts. In addition to observationally documenting sev-eral key attributes of atmospheric rivers, this study alsoconfirms the model-based conclusions presented by Zhuand Newell (1998) that four to five rivers (within theearth’s circumference at midlatitudes) accomplish.90% of the total instantaneous meridional water vaportransport. The detailed CALJET case study of one rivershown here revealed that the horizontal vapor flux wasfocused at low levels in the region of a moisture-laden

prefrontal LLJ, and that 75% of the total observed fluxthrough a 1000-km cross-front baseline was within a565-km-wide zone ;4 km deep. This zone represented1.5 3 108 kg s21 of meridional flux, the equivalent ofroughly 20% of the global average at 358N.

To evaluate the representativeness of this single case,a method was developed to objectively identify atmo-spheric rivers using polar-orbiting (SSM/I) and geosta-tionary (GOES) satellite data from the 1997/98 winterseason. Because the case study revealed a close corre-lation between the horizontal distribution of IWV andhorizontal water vapor flux, and because IWV is ac-curately measured by satellite, IWV was used as a proxyto identify atmospheric rivers. The technique yielded312 baselines, representing 104 sets of observationsthrough many rivers that each showed a narrow, elon-gated plume of IWV $ 2 cm. Composite traces of SSM/I-observed IWV, CLW, and RR and GOES-observedCTT and CTP were then created and used to quantifythe characteristics of satellite-observed rivers for theentire winter season (summarized in Tables 2–5).

The single detailed case was found to be represen-tative of the 312-case composite for IWV $ 2 cm interms of both the IWV amplitude (3.09 cm vs 2.81 cm)and the width of the area where IWV $ 2 cm (424 kmvs 388 km). The composite maximum SSM/I-observedIWV was only 10% greater than the mean of 2.56 cmfound by McMurdie and Katsaros (1991) using earlierversions of SSM/I data in a study of 27 cold-frontalcases in the North Pacific. The SSM/I composites alsoshowed that the mean width scales (defined by the 75%cumulative fraction along a 1500-km composite cross-plume baseline) for CLW and RR were 176 and 141km, respectively, which are both much narrower thanthe 417 km for IWV. The much narrower width scalesfor CLW and RR indicate that vertical circulations re-sponsible for cloud and rain formation occurred on ascale less than half as wide as the physical mechanism(s)responsible for the formation of the IWV plumes. Com-posite traces of GOES CTT and CTP possessed the cold-est and deepest cloud tops in the core of the IWVplumes, and the core cloud tops became substantiallycolder and deeper when the IWV content within theseplumes increased. The anticorrelation between the coremoisture content and the cloud-top core values may havearisen partly due to latent heating: as the moisture con-tent within IWV plumes increases, latent heating trig-gered by forced ascent will be more vigorous, resultingin enhanced upward motion and saturated conditionsdeeper in the troposphere. Those plumes with greatermoisture content may also be indicative of stronger anddeeper baroclinic zones and attendant vertical circula-tions that cause saturated conditions deeper into the tro-posphere.

A latitudinal dependence of the satellite-derivedcross-plume river characteristics was established to as-certain key along-plume attributes. The core IWV con-tent did not vary appreciably with latitude. However,

JULY 2004 1743R A L P H E T A L .

FIG. 23. Conceptual representation of an atmospheric river over the northeastern Pacific Ocean. (a) Plan-view schematic of concentratedIWV (IWV $ 2 cm; dark green) and associated rain-rate enhancement (RR $ 0.5 mm h21; red) along a polar cold front. The tropical IWVreservoir (.3 cm; light green) is also shown. The bold line AA9 is a cross-section projection for (b). (b) Cross-section schematic throughan atmospheric river [along AA9 in (a)] highlighting the vertical structure of the alongfront isotachs (blue contours; m s 21), water vaporspecific humidity (dotted green contours; g kg21), and horizontal alongfront moisture flux (red contours and shading; 3105 kg s21). Schematicclouds and precipitation are also shown, as are the locations of the mean width scales of the 75% cumulative fraction of perturbation IWV(widest), CLW, and RR (narrowest) across the 1500-km cross-section baseline (bottom).

the width scale of IWV $ 2 cm decreased systematicallyfrom south to north, thus suggesting cold-frontal con-fluence may act to contract the lateral scale of the IWVplumes with increasing distance downstream of thesouthern terminus of the fronts. In contrast, the peakvalues of CLW and RR in the core region increased bynearly a factor of 2 and an order of magnitude, respec-tively, from south to north, while their width scales alsoincreased with increasing latitude. The latitudinal var-iation of CLW and RR reflects the fact that frontal cir-culations are generally stronger and broader in the mid-latitude storm track than they are in the subtropics. Thisvariation may also reflect Hadley cell subsidence, whichcan extend northward to ;308N in the Northern Hemi-spheric winter mean. It is intriguing to note that thenorthward decrease of total IWV (factoring in that thearea of IWV $ 2 cm narrows yet the core value staysconstant) appears to be offset by the increasing CLWand RR. While this is suggestive of a conservation ofwater mass in a Lagrangian sense, future efforts focusedon a full water budget study are required to demonstratethis.

The case study and composite results are summarizedschematically in Fig. 23. In Fig. 23a the mean positionof the composite baseline is shown with respect to thecomposite cold front and atmospheric river. The verticalstructure is shown in Fig. 23b, including the key featuresidentified in the case study and in the satellite and syn-optic composites at the position of the composite base-

line where RR . 0. Overall, Fig. 23 highlights the nar-row and shallow nature of atmospheric rivers as wellas the elevated position of their core near 1 km MSL.Their relationship to upper- and lower-level jets andvertical frontal circulations and their associated con-vection, are also shown.

An additional goal of the study was to objectivelyevaluate the capabilities and limitations of current ob-serving systems to detect atmospheric rivers over theeastern Pacific for operational weather prediction andclimate monitoring. SSM/I-observed IWV data are re-liable in clouds and light to moderate rain, while GOESIWV measurements cannot be obtained in cloudy re-gions. The GOES IWV data on the edges of cloud bandsassociated with atmospheric rivers were not correlatedwith the SSM/I-observed IWV within these bands. Infact, the IWV values outside a 300-km-wide swath cen-tered on the IWV core are nearly uncorrelated with thecore values (i.e., r2 ø 0.1), which suggests difficultiesin using current data assimilation systems applied toGOES IWV data to try to fill the gap. Nonetheless, newensemble-based data assimilation systems hold promisefor using the GOES data outside the rivers to refine thepositioning (but not amplitude) of IWV plumes throughuse of an appropriate flow-dependent model of back-ground error covariances (Evenson 2003). Additionally,Scofield et al. (2000) have developed a promising com-positing technique using both SSM/I and GOES IWVdata in combination with operational numerical model


analyses. The analysis of SSM/I data also showed thatthe surface wind speed measurements are not reliablein the core of the rivers because of frequent blockageby rain. While the newer satellite-borne Quick Scatter-ometer (QuikScat) sensor offers the capability to ob-serve surface wind speed and direction even under mod-erate rainfall conditions, there remains substantial un-certainty about how to infer wind conditions aloft wherethe greatest horizontal water vapor fluxes occur. Thus,current satellite systems are unable to adequately mon-itor atmospheric river fluxes for operational forecastingor climate monitoring, although the best approach maybe to blend the strengths of SSM/I and QuikScat polarorbiters with those of GOES.

Because the characteristics of atmospheric rivers aredirectly tied to the storm track, and thus to short-termclimate variability that influences storm tracks, and atthe same time are critical in determining precipitationdistributions and amounts, they form an opportunity forfurther study of the connections between short-term cli-mate variability and the occurrence of extreme precip-itation and flooding. For climate models, the filamentarynature, or skewness, of this key phenomenon suggestspotential problems with aliasing these narrow featuresinto larger scales. Future studies will investigate boththe moisture sources of atmospheric rivers (e.g., frontalconfluence, upward moisture fluxes from the sea sur-face, direct tapping of tropical moisture), as well as thedepletion processes (e.g., rainout). This water budgetwork is especially intriguing in terms of the influenceof passage over complex coastal terrain. Examinationof variations of stable isotopes of rain and water vapor(e.g., Gedzelman et al. 2003) could provide a valuabletechnique in such a budget study.

Acknowledgments. The CALJET field program wasmade possible by the dedicated participation of indi-viduals from numerous organizations. We are gratefulto all of them, particularly the staff of NOAA’s AircraftOperations Center without whom the invaluable NOAAP-3 aircraft data would not have been possible. We arealso indebted to Rolf Langland of the Naval ResearchLaboratory for helping encourage the CALJET team totarget this event for data collection with the P-3. Wethank Chris Velden of the University of Wisconsin—Madison Cooperative Institute for Meteorological Sat-ellite Studies (CIMSS) for collecting and providing theGOES satellite data, Gary Wade (also of CIMSS) forits processing, and Allen White of NOAA/CooperativeInstitute for Research in Environmental Sciences forgenerating the satellite-based composite figures. JimAdams provided exceptional drafting services. The col-lection of data and subsequent study presented here weresupported by NOAA Research (as part of NOAA’s RapidResponse to the strong El Nino of 1997/98), the En-vironmental Technology Laboratory, and the U.S.Weather Research Program. We also thank three anon-

ymous reviewers at MWR, and two internal reviewersat NOAA, for their helpful comments.

REFERENCES

Bluestein, H. B., 1993: Synoptic-Dynamic Meteorology in Midlati-tudes. Vol. 2, Observations and Theory of Weather Systems.Oxford University Press, 594 pp.

Bonner, W. D., 1968: Climatology of the low-level jet. Mon. Wea.Rev., 96, 833–850.

Browning, K. A., 1990: Organization of clouds and precipitation inextratropical cyclones. Extratropical Cyclones: The Erik PalmenMemorial Volume, C. W. Newton and E. Holopainen, Eds., Amer.Meteor. Soc., 129–153.

——, and C. W. Pardoe, 1973: Structure of low-level jet streamsahead of midlatitude cold fronts. Quart. J. Roy. Meteor. Soc.,99, 619–638.

Carlson, T. N., 1991: Mid-latitude Weather Systems. Harper-Collins,507 pp.

Carter, D. A., K. S. Gage, W. L. Ecklund, W. M. Angevine, P. E.Johnston, A. C. Riddle, J. S. Wilson, and C. R. Williams, 1995:Developments in UHF lower tropospheric wind profiling atNOAA’s Aeronomy Laboratory. Radio Sci., 30, 997–1001.

Chien, F.-C., C. F. Mass, and P. J. Neiman, 2001: An observationaland numerical study of an intense landfalling cold front alongthe northwest coast of the United States during COAST IOP2.Mon. Wea. Rev., 129, 934–955.

Cotton, W. R., and R. A. Anthes, 1989: Storm and Cloud Dynamics.Academic Press, 883 pp.

Ecklund, W. L., D. A. Carter, and B. B. Balsley, 1988: A UHF windprofiler for the boundary layer: Brief description and initial re-sults. J. Atmos. Oceanic Technol., 5, 432–441.

Evenson, G., 2003: The ensemble Kalman filter: Theoretical for-mulation and practical implementation. Ocean Dyn., 53, 343–367.

Ferriday, J. G., and S. K. Avery, 1994: Passive microwave remotesensing of rainfall with SSM/I: Algorithm development and im-plementation. J. Appl. Meteor., 33, 1587–1596.

Gedzelman, S., E. Hindman, and X. Zhang, 2003: Probing hurricaneswith stable isotopes of rain and water vapor. Mon. Wea. Rev.,131, 1112–1127.

Goodberlet, M. A., C. T. Swift, and J. C. Wilkerson, 1990: Oceansurface wind speed measurements of the special sensor micro-wave/imager (SSM/I). IEEE Trans. Geosci. Remote Sens., 28,823–828.

Heggli, M. F., and R. M. Rauber, 1988: The characteristics and evo-lution of supercooled water in wintertime storms over the SierraNevada: A summary of microwave radiometric measurementstaken during the Sierra Cooperative Pilot Project. J. Appl. Me-teor., 27, 989–1015.

Iskenderian, H., 1995: A 10-year climatology of Northern Hemi-spheric tropical cloud plumes and their composite flow patterns.J. Climate, 8, 1630–1637.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Keyser, D., and M. A. Shapiro, 1986: A review of the structure anddynamics of upper-level frontal zones. Mon. Wea. Rev., 114,452–499.

Kuhnel, I., 1989: Tropical–extratropical cloudband climatology basedon satellite data. Int. J. Climatol., 9, 441–463.

Lipton, A. E., 1998: Improved GOES sounder coverage of cloud-broken data fields. J. Appl. Meteor., 37, 441–446.

Ma, X. L., T. J. Schmit, and W. L. Smith, 1999: A nonlinear physicalretrieval algorithm—Its application to the GOES-8/9 sounder. J.Appl. Meteor., 38, 501–513.

McGuirk, J. P., A. H. Thompson, and N. R. Smith, 1987: Moisturebursts over the tropical Pacific Ocean. Mon. Wea. Rev., 115,787–798.

McMurdie, L. A., and K. B. Katsaros, 1991: Satellite-derived inte-

JULY 2004 1745R A L P H E T A L .

grated water-vapor distribution in oceanic midlatitude storms:Variation with region and season. Mon. Wea. Rev., 119, 589–605.

Mendell, T., 1992: Integration of automated hydrological data. Pre-prints, Conf. on Interdisciplinary Approaches in Hydrology andHydrogeology, Portland, OR, American Society of Civil Engi-neers. [Available from NOAA/California–Nevada River Fore-cast Center, 3310 El Camino Ave., Sacramento, CA 95821-6308.]

Menzel, W. P., F. C. Holt, T. J. Schmit, R. M. Aune, A. J. Schreiner,G. S. Wade, and D. G. Gray, 1998: Applications of GOES-8/9soundings to weather forecasting and nowcasting. Bull. Amer.Meteor. Soc., 79, 2059–2077.

Namias, J., 1939: The use of isentropic analysis in short term fore-casting. J. Aeronaut. Sci., 6, 295–298.

Neiman, P. J., F. M. Ralph, A. B. White, D. E. Kingsmill, and P. O.G. Persson, 2002: The statistical relationship between upslopeflow and rainfall in California’s coastal mountains: Observationsduring CALJET. Mon. Wea. Rev., 130, 1468–1492.

——, P. O. G. Persson, F. M. Ralph, D. P. Jorgensen, A. B. White,and D. E. Kingsmill, 2004: Modification of fronts and precipi-tation by coastal blocking during an intense landfalling winterstorm in southern california: Observations during CALJET. Mon.Wea. Rev., 132, 242–273.

Newell, R. E., N. E. Newell, Y. Zhu, and C. Scott, 1992: Troposphericrivers? A pilot study. Geophys. Res. Lett., 19, 2401–2404.

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

Palmen, E., and C. W. Newton, 1969: Atmospheric Circulation Sys-tems: Their Structure and Physical Interpretation. AcademicPress, 603 pp.

Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. AmericanInstitute of Physics, 520 pp.

Ralph, F. M., and Coauthors, 1999: The California Land-falling JetsExperiment (CALJET): Objectives and design of a coastal at-mosphere–ocean observing system deployed during a strong ElNino. Preprints, Third Symp. on Integrated Observing Systems,Dallas, TX, Amer. Meteor. Soc., 78–81.

——, and Coauthors, 2003: The impact of a prominent rain shadowon flooding in California’s Santa Cruz Mountains: A CALJETcase study and sensitivity to the ENSO cycle. J. Hydrometeor.,4, 1243–1264.

Rao, P. A., and H. E. Fuelberg, 1998: An evaluation of GOES-8retrievals. J. Appl. Meteor., 37, 1577–1587.

Reynolds, D. W., and A. P. Kuciauskas, 1988: Remote and in situobservations of Sierra Nevada winter mountain clouds: Rela-tionships between mesoscale structure, precipitation, and liquidwater. J. Appl. Meteor., 27, 140–156.

Schluessel, P., and W. J. Emery, 1990: Atmospheric water vapourover oceans from SSM/I measurements. Int. J. Remote Sens.,11, 753–766.

Scofield, R., G. Vicente, and M. Hodges, 2000: The use of watervapor for detecting environments that lead to convectively pro-duced heavy precipitation and flash floods. NOAA Tech. Rep.NESDIS 99, 64 pp. [Available from the National Technical In-formation Service, 5285 Point Royal Rd, Springfield, VA,22161.]

Smirnov, V. V., and G. W. K. Moore, 1999: Spatial and temporalstructure of atmospheric water vapor transport in the MackenzieRiver basin. J. Climate, 12, 681–696.

——, and G. W. K. Moore, 2001: Short-term and seasonal variabilityof the atmospheric water vapor transport through the MacKenzieRiver basin. J. Hydrometeor., 2, 441–452.

Stensrud, D. J., 1996: Importance of low-level jets to climate: Areview. J. Climate, 9, 1698–1711.

Thepenier, R.-M., and D. Cruette, 1981: Formation of cloud bandsassociated with the American subtropical jet stream and theirinteraction with midlatitude synoptic disturbances reaching Eu-rope. Mon. Wea. Rev., 109, 2209–2220.

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

Weber, B. L., D. B. Wuertz, D. C. Welsh, and R. McPeek, 1993:Quality controls for profiler measurements of winds and RASStemperatures. J. Atmos. Oceanic Technol., 10, 452–464.

Weng, F., and N. C. Grody, 1994: Retrieval of cloud liquid waterusing the Special Sensor Microwave Imager. J. Geophys. Res.,99, 25 535–25 551.

Wentz, F. J., 1997: A well-calibrated ocean algorithm for SpecialSensor Microwave/Imager. J. Geophys. Res., 102, 8703–8718.

White, A. B., D. J. Gottas, E. Strem, F. M. Ralph, and P. J. Neiman,2002: An automated brightband height detection algorithm foruse with Doppler radar vertical spectral moments. J. Atmos. Oce-anic Technol., 19, 687–697.

——, P. J. Neiman, F. M. Ralph, D. E. Kingsmill, and P. O. G. Persson,2003: Coastal orographic rainfall processes observed by radarduring the California Land-falling Jets Experiment. J. Hydro-meteor., 4, 264–282.

Whiteman, C. D., X. Bian, and S. Zhong, 1997: Low-level jet cli-matology from enhanced rawinsonde observations at a site inthe southern Great Plains. J. Appl. Meteor., 36, 1363–1376.

Zhu, Y., and R. E. Newell, 1994: Atmospheric rivers and bombs.Geophys. Res. Lett., 21, 1999–2002.

——, and ——, 1998: A proposed algorithm for moisture fluxes fromatmospheric rivers. Mon. Wea. Rev., 126, 725–735.

——, ——, and W. G. Read, 2000: Factors controlling upper-tro-posphere water vapor. J. Climate, 13, 836–848.

Satellite and CALJET Aircraft Observations of Atmospheric ...

Documents