radar-observed characteristics of precipitating systems during

Radar-Observed Characteristics of Precipitating Systems during NAME 2004

TIMOTHY J. LANG

Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

DAVID A. AHIJEVYCH

National Center for Atmospheric Research, Boulder, Colorado

STEPHEN W. NESBITT*Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

RICHARD E. CARBONE

National Center for Atmospheric Research, Boulder, Colorado

STEVEN A. RUTLEDGE AND ROBERT CIFELLI

Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

(Manuscript received 25 October 2005, in final form 28 April 2006)

ABSTRACT

A multiradar network, operated in the southern Gulf of California (GoC) region during the 2004 NorthAmerican Monsoon Experiment, is used to analyze the spatial and temporal variabilities of local precipi-tation. Based on the initial findings of this analysis, it is found that terrain played a key role in thisvariability, as the diurnal cycle was dominated by convective triggering during the afternoon over the peaksand foothills of the Sierra Madre Occidental (SMO). Precipitating systems grew upscale and moved WNWtoward the gulf. Distinct precipitation regimes within the monsoon are identified. The first, regime A,corresponded to enhanced precipitation over the southern portions of the coast and GoC, typically duringthe overnight and early morning hours. This was due to precipitating systems surviving the westward trip(�7 m s�1; 3–4 m s�1 in excess of steering winds) from the SMO after sunset, likely because of enhancedenvironmental wind shear as diagnosed from local soundings. The second, regime B, corresponded to thesignificant northward/along-coast movement of systems (�10 m s�1; 4–5 m s�1 in excess of steering winds)and often overlapped with regime A. The weak propagation is explainable by shallow–weak cold pools.Reanalysis data suggest that tropical easterly waves were associated with the occurrence of disturbedregimes. Gulf surges occurred during a small subset of these regimes, so they played a minor role during2004. Mesoscale convective systems and other organized systems were responsible for most of the rainfallin this region, particularly during the disturbed regimes.

1. Introduction

The intensive observation period (IOP) of the NorthAmerican Monsoon Experiment (NAME; Higgins et

al. 2006) took place during July and August of 2004. Amajor component of the IOP was observations from amultiradar network placed in tier I, the core monsoonregion consisting of the Gulf of California (GoC) andthe Sierra Madre Occidental (SMO) in northwesternMexico (Higgins et al. 2006). The network, depicted inFig. 1a, consisted of three radars: the National Centerfor Atmospheric Research (NCAR) S-band dual-polarization Doppler radar (S-Pol), placed �100 kmnorth of Mazatlán, Mexico, on the coast west of theSMO; and two Servício Meteorológico Nacional (SMN)Doppler radars—one at Guasave farther north on the

* Current affiliation: Department of Atmospheric Sciences,University of Illinois at Urbana–Champaign, Urbana, Illinois.

Corresponding author address: Timothy J. Lang, Dept. of At-mospheric Science, Colorado State University, Fort Collins, CO80523.E-mail: [email protected]

1 MAY 2007 L A N G E T A L . 1713

DOI: 10.1175/JCLI4082.1

© 2007 American Meteorological Society

JCLI4082

coastal plain and one at Cabo San Lucas at the tip ofthe Baja California peninsula.

A central goal of NAME is to characterize and un-derstand convective and mesoscale processes in thecomplex terrain of the core monsoon region and theirinteraction within the context of the broader monsooncirculation. In particular, precipitation systems at thesescales affect the surrounding environment throughtransports of heat, moisture, and momentum. The re-sponse of the large-scale circulation is sensitive to thevertical distribution of latent heating in convective andmesoscale complexes (e.g., Hartmann et al. 1984).Therefore, radar data are needed to characterize andexamine the structure, kinematics, morphology, and di-urnal cycle of individual precipitating systems (includ-ing mesoscale convective systems, or MCSs) within themonsoon. In addition, the data are being used to studythe interaction of precipitating systems with the vari-able surface properties in the region over the SMO, theGoC, and the intervening coastal plain—as well as theinfluence of transient meteorological events such astropical easterly waves (e.g., Fuller and Stensrud 2000)and gulf surges (e.g., Hales 1972) on precipitation in theregion.

In this paper we will report initial findings on severaloutstanding research questions and hypotheses relevant

to the North American monsoon [NAM; see Higgins etal. (2006) for a complete summary of NAME scientificobjectives].

a. Diurnal cycle of precipitation

Ground-based radar assessment of the spatial andtemporal variability of precipitation in a tropical regionof significant orography (e.g., the SMO) provides in-sights into the physical characterization of precipitatingsystem life cycles through its ability to constantly moni-tor storm morphology. Thus, in this paper we will usethe NAME radar network to observe and describe sta-tistically the diurnal cycle of precipitating systems intier I.

Previous investigators (e.g., Negri et al. 1993) haveshown that the diurnal cycle of precipitation in the tierI domain has far-reaching implications for the mon-soon, but the details of the diurnal cycle are not wellunderstood. Better understanding is required of the lo-cation of convective development along the westernslopes of the SMO and the evolution and decay of thesystems as they move westward over the gulf.

It has been hypothesized that a polarization betweenlate afternoon and evening convection on the SMO andlate night/early morning convection over the GoC, seenby previous satellite investigations (e.g., Negri et al.

FIG. 1. (a) NAME radar composite domain with an inscribed subdomain used for the RDAs and precipitation feature analyses. Theorigin (0,0) corresponds to the center of the NRC. (b) The RDA subdomain and related terrain variations. Surface elevation is shadedand mean elevation profiles are shown along the sides. The x-axis segments corresponding to the GoC, coastal plain, SMO foothills,and SMO peaks are labeled.

1714 J O U R N A L O F C L I M A T E VOLUME 20

1993), is caused by either 1) gradual propagation of thedeveloping systems over the SMO to the west due toprevailing winds or 2) westward propagation of gravitywaves from the SMO convection, which then forcegulf convection later (e.g., Mapes et al. 2003). Thisstudy will establish the timing, evolution, and propaga-tion of convective systems, thus testing both of thesehypotheses.

b. Intraseasonal variability of precipitation

A major goal of NAME is to better understand re-gimes associated with intraseasonal variability of con-vection during July–August in the tier I region and itslinkages to precipitation in the southwestern UnitedStates, including the influences of surges, jets, easterlywaves, surface fluxes, and topographic blocking (Hig-gins et al. 2006). It has been hypothesized that gulfsurges and tropical easterly waves play major roles inorganizing convective activity throughout the tier I do-main. In this study, the radars will be used to charac-terize convective activity during gulf surges and duringother periods and, thus, statistically determine key con-vective structure differences (e.g., storm size and orga-nization) between different meteorological regimes. Inaddition, we will identify changes in the environment(e.g., wind shear) that could be relevant for observeddifferences in convection.

c. Relative importance of organized systems forprecipitation

It has been hypothesized that mesoscale convectivesystems and other modes of organized convection con-tribute significantly to total rainfall within the tier Idomain (Higgins et al. 2003). The radar data can testthis through the characterization of convective organi-zation and total rainfall by different precipitating sys-tems. In particular, we will examine the relative impor-tance of organized systems throughout their character-istic life cycles as a function of both the diurnal cycleand meteorological regime (i.e., intraseasonal variabil-ity).

d. Role of terrain in triggering and organizingconvection

It has been hypothesized that terrain plays a majorrole in organizing convection over the tier I domain(Higgins et al. 2003). S-Pol and SMN radar observa-tions can be used to assess the morphology and orga-nization of storms relative to major terrain features. Inparticular, we will identify preferred locations for con-vection along the SMO, as well as other locations, and

will identify the preferred timing for convection inthese regions.

2. Data and methodology

a. Network design, data quality control, andproduct generation

Figure 1a demonstrates the basic geometry of theNAME radar network. S-Pol was deployed in NAMEduring 8 July–21 August 2004 to a location 10 km westof La Cruz de Elota, Sinaloa, Mexico. Throughout theIOP, and among other types of scans, S-Pol providedone set of low-angle 360° surveillance scans (0.8°, 1.3°,and 1.8° elevation angles) for rain mapping, usually outto �210 km range. Another set of scans extending tohigher-elevation angles was used for the analysis of pre-cipitation vertical structure, but these data are not usedin this study. Both scan sets were routinely updatedevery 15 min.

The SMN radars are C-band Doppler radars. Theradars were operational prior to NAME, but did notdigitally record their data. Guasave was upgraded totemporarily record data on 10 June 2004. Cabo wassimilarly upgraded on 15 July. Guasave recorded datainto the fall, and Cabo recorded until 14 August. Gua-save data have been processed for 8 July–21 August.However, due to a disk failure, Guasave data aremostly missing during 22–31 July. During NAME, theSMN radars ran at a single elevation angle. For Cabothis angle was 0.6°. Guasave varied between 0.5°, 1.0°,and 1.5°. Both radars had data coverage out to at least�230 km.

Prior to Cartesian gridding and analysis, data from allthree radars were subjected to rigorous quality controlefforts, as outlined in the appendix. The version 1 re-gional composites were created on a 0.02° (�2 km)latitude–longitude grid. The files, available nearly every15 min during the NAME IOP, contain near-surfacereflectivity and near-surface rain rate. We did not in-terpolate to a fixed altitude, instead using the lowest-altitude data possible for each grid point.

Sweeps from the same time (within 2 min) and thelowest elevation angle were combined every 15 min toproduce network composites. Before converting to alatitude–longitude grid, the data along each ray weresmoothed using a 1000-m moving average, and resam-pled to a more sparse array of 1000-m “gates.” Whereradar gates from different radars overlapped, thelowest gate took precedence and higher gates wereeliminated. The remaining gates were combined andinterpolated to a regular latitude–longitude grid. Aninverse-distance weighting method was employed toproduce the interpolated values. In the final product,

1 MAY 2007 L A N G E T A L . 1715

the mean beam height of a grid point is about 3900 mMSL, with a standard deviation of 1900 m.

While the Baja peninsula was very dry, there oftenwas substantial echo south and west of the Cabo radar,over the ocean. After spot intercomparisons with sat-ellite observations [e.g., Geostationary Operational En-vironmental Satellite (GOES), Moderate ResolutionImaging Spectroradiometer (MODIS)], we suspect thatmuch of this echo is sea clutter. As a result, we limitanalyses to the eastern portion of our original domain,that is, the Gulf of California subdomain east to theSMO peaks subdomain (Fig. 1b).

No effort was made to validate rainfall estimatesfrom the radars. This will be addressed in the future.We examine rainfall and not radar reflectivity since theformer is of most interest to climate scientists. Whilethere are uncertainties in the rainfall estimates, we feelour quality control efforts (see the appendix) have beenextensive and justify the use of rainfall, particularlysince we interpret only relative differences and not ab-solute values of rainfall.

b. Reduced dimension methodology

To examine characteristics of the rainfall climatol-ogy, a reduced dimension analysis domain (RDA) wascreated. The RDA is aligned with the mean orienta-tion of the SMO and the GoC, as illustrated in Fig. 1.The grid is rotated 35° counterclockwise from truenorth. The data were bilinearly interpolated onto a2-km Cartesian grid from the NAME radar compositedomain (NRC), which is in latitude–longitude coordi-nates. The orientation of the RDA largely relegatesthe topographical and related surface variations to thex dimension. However, the coastal plain broadensconsiderably from the SE to the NW, thus increasingthe fraction of x distance over land versus sea from toSE to NW. The lower-right corner of the RDA is out-side of the NRC, thus rendering these grid locationsvacant.

The analyses from this domain are based on arith-metic averages of the radar-estimated rainfall ratealong either the x or y dimension and are subsequentlyplotted as a function of time (Hovmöller diagrams) at15-min intervals. Where data are incomplete (meaningno radar was scanning at a particular grid point at agiven time), at least 60 km of valid data (contiguous ornot) must be present to calculate an average along anygiven gridline at any given time.

c. Identification of precipitation features within theNAME radar composites

The methodology of identifying precipitation fea-tures (PFs) within the NRC is essentially identical to

the methodology first outlined in the Tropical RainfallMeasuring Mission (TRMM; Simpson et al. 1988) Pre-cipitation Radar and TRMM Microwave Imager datastudy of Nesbitt et al. (2000). Here, contiguous areas(including corner pixels) of composite equivalent radarreflectivity �15 dBZ (value chosen because of the sen-sitivities of the SMN radars) are considered as a PFwithin each composite. Features were identified withineach composite for the analysis period. The PFs wereidentified within the entire composite domain, but onlythose features with mass-weighted centroid locationswithin the rotated domain of Fig. 1b were analyzed inthis study.

Characteristics of each feature were recorded in adatabase based upon the reflectivity and rainfall struc-ture, location, and time of occurrence of each PF. Aconvective–stratiform separation algorithm is appliedto the reflectivity field [Yuter and Houze (1998)); withtunable parameters a � 8.0 and b � 55.0], and totalrainfall volume, convective and stratiform rainfall ar-eas, and rainfall fractions are recorded for each PF,along with its maximum reflectivity value. An ellipse-fitting procedure, developed by Nesbitt et al. (2006),was applied to each PF to objectively estimate themaximum dimension (i.e., twice the major axis lengthof each ellipse) of each feature. An example of thisanalysis is shown in Fig. 2, which shows the compositereflectivity from 5 August 2004 at 1715 local time (LT).The best-fit ellipse of each feature is plotted over eachidentified PF within the composite.

Features were stratified geographically, by regime(defined below), and by their characteristics. The dis-tances of each PF centroid relative to the axes of therotated coordinate system were calculated in kilome-ters (using nonspherical geometry as above). The fea-tures were assigned to four regions based on their dis-tance from the y axis to examine cross-coast variability(Fig. 1b) and two regions based on their distance fromthe x axis (to examine along-coast variability). Figure1b shows the x boundaries of each of the four cross-coast geographic regions in the rotated coordinate. Thetwo along-coast regions result from evenly splitting theFig. 1b RDA domain into north and south.

To identify features that have become organized onthe mesoscale, a subset of PFs has been identified asorganized features. These features meet the criteria thattheir maximum horizontal dimension is � 100 km, andthe storm contains at least 16 km2 of convective area.This definition attempts to match the Houze (1993,chapter 9) definition of a mesoscale convective systemwithin the framework of the NAME composites, butcould refer to any organized precipitating system.


3. Reduced dimension analysis

The following data are presented in Hovmöller dia-grams, one horizontal dimension (x or y) plottedagainst time, separately, for each horizontal axis. Twotypes of diagrams are shown: The first depicts the entiretime series over 42 days, and the second depicts diurnalvariability either for the period of record of portionsthereof.

a. Event data

Figure 3 shows the 42-day time series of average rain-fall rate in reduced dimension, both transverse (lhs)and parallel (rhs) to the GoC–SMO major axis. Severalcharacteristics are evident including the following:

• a pronounced diurnal cycle of rainfall over the periodof record;

• horizontal patterns, indicating little movement ofprecipitation regions;

• smoothly sloping patterns, indicative of a systematicphase speed;

• a tendency for major events to originate over SMOpeaks and foothills (lhs);

• phase speeds suggestive of slow movement from theSMO to the GoC (lhs);

• periods when precipitation progresses well into oracross the GoC (lhs);

• periods when phase speeds exhibit consistent north-ward along-coast movement (rhs); and

• periods when precipitation exhibits both along-coastand cross-coast (toward the gulf) components to theirmovement

The variability in precipitation patterns appearslargely systematic. For pragmatic reasons, we identifytwo regimes to help characterize this variability. Wehave set an arbitrary threshold of 0.17 mm h�1 for the24-h average rainfall over the GoC and coastal plain,above which regime A is designated. As a consequenceof this threshold, regime A occurs one-third of the time.Regime A is characterized by a coherent progression ofenhanced rainfall from the SMO to the GoC. Precipi-tation over the GoC is mainly nocturnal. A blue stripealong the right-hand side of Fig. 3 marks the periodsdesignated as regime A. It is evident from Fig. 3 thatother days exhibit a similar tendency for the coherentprogression of precipitation into the GoC under moresuppressed conditions.

FIG. 2. Composite equivalent radar reflectivity (dBZ ) shaded at 1715 LT 5 Aug 2005. Thebest-fit ellipse is shown for each feature in the PF database. The rotated domain used foranalysis is shown by the dashed line, and terrain is shaded (grayscale) in the background.

1 MAY 2007 L A N G E T A L . 1717

Fig 2 live 4/C

FIG. 3. Reduced-dimension time series of rainfall rate from [(a) 10–26 Jul, (b) 24 Jul–6 Aug, (c) 10–20Aug 2004]. Local time is shown on the y axis (Julian day and hour). Also shown are (left) the cross-coastdimension and (right) the along-coast dimension. Regimes A (blue) and B (pink) are denoted on therhs with colored stripes. For reference, the domain-averaged surface elevation is profiled at the top ofeach column and thin black lines mark the subregions described in Fig. 1 (lhs).


Fig 3a live 4/C

FIG. 3. (Continued)

1 MAY 2007 L A N G E T A L . 1719

Fig 3b live 4/C

FIG. 3. (Continued)


Fig 3c live 4/C

Regime B is defined when the along-coast progres-sion of precipitation is prominent. The along-coast pro-gression of precipitation manifests itself as sloping rain-fall streaks on the rhs of Fig. 3. From previous studies(Carbone et al. 2002) in other regions it has been de-termined that the sloping rainfall streaks are often as-sociated with organized mesoscale convection while thehorizontal structures are characterized as unorganizedconvection. The pink stripe along the right-hand side ofFig. 3 designates regime B and indicates along-coastmotion of 8–13 m s�1. Regime AB refers to the inter-section of regime A and regime B, that is, when pre-cipitation moves both along-coast and cross-coast.

Our identification of regimes is phenomenologicallybased. Further research is required to ascertain wheth-er there is any physical or dynamical basis to the ob-served precipitation patterns. However, as will be dem-onstrated below, our initial results suggest that regimesA, B, and AB are correlated with different wind shearenvironments, as well as with proximity to tropical east-erly wave passages.

b. Diurnal cycle

The percentage of time that rainfall meets or exceeds0.2 mm h�1 is shown in Fig. 4 as a function of x or ydistance and time of day (LT) for the entire period ofrecord. Note the maximum frequency of occurrencenear 1800 LT over the SMO peaks and foothills (lhs).

The persistent triggering of afternoon convection in thisregion is indicated by the rapid onset of high precipi-tation frequencies around 1400 LT. While the diurnalmaximum is broad, a progression toward the GoC isevident: precipitation moves off the SMO, arrives onthe coastal plain (x � �100 km) around local midnight,and often persists until sunrise. After 0300 LT, motionaway from the SMO and into the GoC decreases sig-nificantly as is evidenced by the frequency peak becom-ing diffuse with little appreciable change in the positionof the centroid.

The along-coast diurnal cycle is shown on the right-hand side of Fig. 4. Isolated maxima in precipitationfrequency appear to align with local peaks in the meanelevation profile (vertical dashed lines). As in midlati-tude North America, elevated heat sources in the SMOplay a dominant role in the excitation of deep moistconvection (Carbone et al. 2002; Ahijevych et al. 2004).Another feature of significance in Fig. 4 is the along-coast trend of increasing nocturnal precipitation in thesouthern portion of the domain, where the coastal plainis narrow and a larger fraction of the x dimension re-sides over the GoC.

Figure 5 is the diurnal cycle for the regime A periods(left panel) versus all other periods (right panel). Themost significant differences are in these cross-coast pat-terns. Rainfall is generally more frequent during regimeA; however, the difference in frequency is most promi-

FIG. 4. Percentage of time that the rainfall rate meets or exceeds 0.2 mm h�1 as a functionof local time (diurnal cycle repeated for clarity): (left) the cross-coast frequency and (right)the along-coast frequency. Mean surface elevation along each dimension is profiled at the top.The vertical black lines on the lhs correspond to the cross-coast zones identified in Fig. 1b.The vertical dashed lines on the rhs are aligned with local peaks in mean elevation and helpto identify the close relationship between elevated heat sources and rainfall frequency duringthe diurnal maximum.

1 MAY 2007 L A N G E T A L . 1721

Fig 4 live 4/C

nent at night. On non–regime A days, the precipitationlargely remains over the SMO foothills and peaks, anddissipates before local midnight. Regime A exhibits amuch higher frequency of occurrence throughout thenight along the coastline and into the GoC. This resultis not surprising since it is closely related to the metricfor stratification of the time series. In Fig. 5, a signifi-cant semidiurnal maximum is evident just offshore(�0800), which is not obvious from the daily time series(Fig. 3). The activity throughout the night and earlymorning hours is at a location likely to be coincidentwith the land-breeze front. Doppler radar velocityfields (not shown) exhibit a clear diurnal cycle of near-shore winds and the breeze convergence line over theGoC in the nocturnal hours, in close proximity to thestatistical morning rainfall maximum. Surprisingly,there is evidence of convective amplification and theappearance of cross-coast “propagation” from 0800through 1200 LT, resulting in precipitation across muchof the GoC. Phase speeds implied by the patterns inFig. 5 are approximately 7 m s�1, both for the eveningprogression of rainfall from the SMO and the nocturnalpropagation over the GoC.

Figure 6 shows the diurnal frequency associated withregime B and non–regime B periods. Unlike regime A,the strongest signals appear in the along-coast struc-tures. Regime B exhibits pronounced along-coast

movement that is phase locked with the diurnal cycle.The slope of this pattern suggests an along-coast phasespeed of 8.5–12.5 m s�1. Other periods exhibit little evi-dence of systematic movement, reflecting a simple di-urnal maximum closely tied to elevated heat sources.

4. Precipitation feature analysis

a. NAME IOP time series

Time series of various precipitation feature statistics,covering the entire domain and NAME IOP, are shownin Fig. 7. A 24-h running mean filter has been applied toall time series, for clarity. Also shown are shaded barsindicating the occurrence of regimes A (dark shading)and B (light shading). Overall, regimes A and B corre-spond to relative maxima in mean volumetric rainfallper feature (Fig. 7a), mean feature maximum dimen-sion (Fig. 7c), and fraction of rainfall contributed byorganized features (Fig. 7d). In fact, larger mean valuesof these parameters occurred during regimes A, B, andAB relative to nonregime periods, as shown in Table 1.Though not obvious from Fig. 7b, the fraction of rainproduced by convective (as opposed to stratiform) pix-els actually shows a slight increase during nonregimeperiods (Table 1). Overall, during regimes A, B, andAB, features tend to be larger and produce more rain-

FIG. 5. Same as in Fig. 4 (left column) except for regime A vs other periods.


Fig 5 live 4/C

fall, and the relative importance of organized featuresincreases.

b. Diurnal cycle

The diurnal cycle for various precipitation featurestatistics was determined and broken down by both re-gime and location. A common theme in all the plotsthat will be shown here is the occurrence of two relativemaxima across various PF statistics: a smaller one in themorning hours, and a larger one in the afternoon/evening. In addition, there is a general tendency for theconvective fraction to peak first, followed sequentiallyby feature rainfall, mean feature maximum dimension,and finally fraction of rain from organized features.This reflects a general tendency for features to growupscale from small unorganized cumulonimbi to largerand more organized systems.

Features were assigned as members of the A, B, andAB regimes based upon their time of occurrence. Fig-ure 8 shows the diurnal cycle of precipitation featurestatistics for regime AB. Given the time overlap be-tween A and B (e.g., Fig. 7), plots for regime A onlyand regime B only give very similar results, so we con-sider only AB here. Non-AB periods show a maximumin mean volumetric rainfall per feature near 1900 LT(Fig. 8a). The regime AB peak comes later in time, near2100 LT. There is more rainfall per feature throughoutthe day during regime AB.

As expected from Table 1, the diurnal cycle of theconvective area fraction tends to overlap between AB

and non-AB periods (Fig. 8b). They both show twodiurnal peaks, occurring during a late afternoon peakand a morning peak. However, the non-AB morningpeak is a couple of hours earlier than the AB peak.During the AB period, the mean feature maximum di-mension (Fig. 8c) is larger than that for the non-ABperiod, and the shapes of the diurnal cycles are aboutthe same. The diurnal cycle of the organized featurerainfall fraction (Fig. 8d) shows that in both regime ABand the non-AB regime, there is a relative maximumnear midnight, and another morning peak. However, asin the convective area fraction, the morning AB peakcomes 2 h later than for non-AB. Also, the organizedfeature rainfall fraction during AB is larger than non-AB throughout the day.

Figure 9 shows the diurnal cycle of PF statistics bygulf-normal location, that is, mean behavior for eachsubdomain in Fig. 1b (SMO peaks, SMO foothills,coastal plain, GoC), regardless of regime. Most rain perfeature falls over the SMO foothills (Fig. 9a), but gen-erally only during the afternoon and evening hours.The foothills also receive the greatest rainfall 2 h laterthan the SMO peaks. The coastal plain and GoC tendto have broader peaks in their feature rainfall diurnalcycle. Coastal plain rainfall peaks at the same time asthat for the foothills, but extends much further into theovernight and morning hours than the SMO. The peaktime for the coastal plain probably reflects the combi-nation of afternoon convection triggered along thecoastline (e.g., sea-breeze front) and late evening con-vection moving off the SMO. The GoC diurnal cycle is

FIG. 6. Same as in Fig. 4 (right column) except for regime B vs other periods.

1 MAY 2007 L A N G E T A L . 1723

Fig 6 live 4/C

roughly 12 h out of phase with the SMO, with rainfallmainly occurring over a broad time period includingovernight and in the morning. The diurnal cycles in therest of Fig. 9 mostly conform to expectations derivedfrom Fig. 9a.

Splitting the entire domain (Fig. 1b) equally into anorthern half and southern half (i.e., gulf-parallel sub-

domains) also reveals interesting characteristics of thediurnal cycle, as seen in Fig. 10. The main result is thatthe morning secondary peak in the statistics is moresignificant in the south than the north. In addition, theamplitude of the afternoon/evening peak in the PF sta-tistics often is smaller in the south. As will be shown inthe next section, this is due to a greater number of early

FIG. 7. Time series of precipitation feature statistics for the entire NAME IOP. Dark-shaded bars denote time periods for regime A;light-shaded bars denote time periods for regime B. (a) Feature volumetric rainfall. (b) Fraction of volumetric rainfall produced byconvective pixels. (c) Mean feature maximum dimension (i.e., ellipse major axis length). (d) Fraction of volumetric rainfall producedby organized features.

TABLE 1. Mean values of precipitation feature statistics during and outside of disturbed regimes. All differences in means aresignificant at the 95% confidence level (Student’s t test).

Statistic Mean (A/�A) Mean (B/�B) Mean (AB/�AB)

Volumetric rainfall per feature (mm h�1 km2) 1494/868 1517/913 1693/858Convective rainfall fraction (%) 52.8/55.0 53.3/54.2 52.9/54.8Feature max dimension (km) 20.3/18.0 20.3/18.2 20.9/17.9Fraction of rain from organized features (%) 75.5/60.7 74.9/63.3 76.7/60.0


morning organized systems in the south when com-pared to the north. The environmental reasons for thisphenomenon will be explored in section 5.

c. Organized features and the diurnal cycle

To highlight the spatial coherence of organizedmodes of convection within the diurnal cycle, Fig. 11shows the location of all organized features during theday. Each panel in the figure shows the centroid loca-tion of each organized feature, and the color of thesymbol indicates the regime assigned to each PF.

Considering early afternoon (1200–1600 LT; Fig.11a), organized features mainly appear over land; how-ever, there is a spatial separation in the locations wherefeatures form during this time period based on the re-gime(s) under consideration. These features are likelytriggered by solar insolation over the elevated terrain ofthe SMO. During “no regime” periods, organized fea-tures were mainly confined to the high terrain. Duringdisturbed periods, particularly regime AB, organizedfeatures tended to be located in the foothills. Duringthe late afternoon (1600–2000 LT; Fig. 11b), there isevidence that the bulk of the organized features haveadvanced toward the coast, with organized features

during disturbed periods having locations closer to thecoast than during nondisturbed times. In Fig. 11c(2000–0000 LT), there is generally more organized fea-ture activity in the northern portion of the domain dur-ing the evening hours than to the south, similar to thebulk results of Fig. 10. These organized features in thenorthern part of the domain either dissipate as theyreach the coast by midnight (Fig. 11d), or exit the do-main.

The few features in the southern foothills of the do-main in Fig. 11c, along with features that likely propa-gate in from outside the southern edge of the domain,appear to have long lifetimes as they propagate into theGoC during the early morning hours (0000–0800 LT;Figs. 11d and 11e). The majority of these overgulf earlymorning organized features occur during regime AB.During the late morning (Fig. 11f), there is some evi-dence of organized features forming again along theSMO high terrain as the daily cycle of solar insolationleads to convective development.

5. Environmental influences

The sounding data collected during the NAME IOPat Los Mochis and Mazatlán (see Fig. 1) allow an ex-

FIG. 8. Diurnal cycle of precipitation feature statistics, broken down by regime AB (inter-section of A and B; dashed) and non–regime AB (solid). (a) Total volumetric rainfall. (b)Fraction of volumetric rainfall produced by convective pixels. (c) Mean feature maximumdimension (i.e., ellipse major axis length). (d) Fraction of volumetric rainfall produced byorganized features (as defined in text).

1 MAY 2007 L A N G E T A L . 1725

amination of the environmental influences (thermody-namic and shear profiles) associated with the precipi-tation regimes identified above. Sounding data fromthe Los Mochis Integrated Sounding System (ISS) siteand Mazatlán International Airport are compositedduring the same time periods as the radar data analysis.While thermodynamic data and winds are available atLos Mochis, currently only wind data are availablefrom Mazatlán due to a dry bias that has been identifiedin the operational soundings at Mazatlán (P. Cieselski2005, personal communication).

To examine thermodynamic profiles as a function ofregime, Fig. 12 shows regime-composite skew T–logpdiagrams at Los Mochis, showing temperature, dew-point, and pseudoadiabatic surface-based parcel ascentpath. Regime AB soundings are the moistest on aver-age, especially above the level of free convection (LFC,which is around 700 hPa in all regimes). Relativelysmall differences in the soundings are present below thelifting condensation level (LCL, which was located atroughly 900 hPa in all cases), and all soundings displaya weak conditionally unstable layer just above the LFCto near 550 hPa, with a pseudoadiabatic layer aboveuntil nearly the tropopause. This unstable layer (possi-

bly due to elevated sensible and latent heating over thenearby SMO) contributes to significant overall condi-tional instability; all regime-averaged soundings con-tain at least 1500 J kg�1 of CAPE [CAPE and convec-tive inhibition (CIN) were calculated using the averagetemperature and dewpoint of the lowest 50 hPa in thesounding], with slightly higher CAPE values during theno-regime periods, and the least during regime A (seeTable 2). However, there is a significant cap in the av-eraged soundings, with 40–70 J kg�1 of CIN present inthe averaged soundings, with the largest (smallest)cap present in the AB (no regime) composite. In addi-tion, the no-regime periods are drier near midlevels(i.e., 450 hPa).

The significant amount of CAPE available above theLFC (at around 750 hPa in each regime; Fig. 12) sug-gests that, provided the necessary lifting mechanismsexist (e.g., sea-breeze convergence zone, or long-livedconvective systems with established cold-pool or grav-ity wave dynamics), systems could potentially tap thepositive energy aloft if they are able to lift parcelsabove the LFC. It is apparent that thermodynamic pro-files are sufficient for deep convective triggering if thecap could be broken by appropriate forcing mechanisms.

FIG. 9. Diurnal cycle of precipitation feature statistics, broken down by E–W subdomain(Fig. 1b): SMO peaks (gray dashed), SMO foothills (gray solid), coastal plain (black dashed),and GoC (black solid). (a) Feature volumetric rainfall. (b) Fraction of volumetric rainfallproduced by convective pixels. (c) Mean feature maximum dimension (i.e., ellipse major axislength). (d) Fraction of volumetric rainfall produced by organized features (as defined in text).


Figure 13 shows regime-averaged wind profiles com-posited as above, and rotated 35° in a mathematicalsense as was done for the radar composites to look atcross-coast and along-coast shear profiles. Profiles ofregime-composited cross-coast and along-coast wind di-rections, as well as wind speeds, are shown for LosMochis and Mazatlán. Note that positive cross-coastwind is away from the gulf toward the SMO, and posi-tive along-coast wind means away from the mouth ofthe gulf toward, say, Arizona.

Below 850 hPa, very different characteristic wind di-rections and speeds exist between the two sites duringthe disturbed regimes (A, B, AB). At Los Mochis, thereis a low-level wind maximum on average of about 3–4m s�1 in these regimes, which has a large southerlycomponent (in unrotated space). This is likely associ-ated with the GoC low-level jet, which is a climatologi-cal feature of the NAM (Douglas 1995; Stensrud et al.1997). This feature is not present in terms of southerlyflow at Mazatlán (which is south and east of the clima-tological position of the jet), where light northwesterlywinds are present at low levels (in unrotated space).The difference between the Mazatlán and Los Mochislow-level flows is consistent with mean wind patternsobserved during the Southwest Area Monsoon Project

(SWAMP) and reflect contrasts in thermal structureacross the northern and southern portions of the GoCregion (Douglas 1995).

Above 850 hPa, the wind directions become moresimilar between the two sites (southeasterly); however,along-coast component speeds are generally stronger atMazatlán than at Los Mochis. This is especially true ataround 650 hPa during the disturbed regimes, wherelow-level shear (which is largely unidirectional) be-tween the surface and roughly 4 km is more pro-nounced at Mazatlán. Table 3 shows that values of 0–4-km shear are larger at Mazatlán, nearly by a factor of 2during disturbed regimes. This increased shear may ex-plain the prevalence of the longer-lived convective sys-tems over the southern portion of the radar domain,which were observed to last into the early morninghours (e.g., Rotunno et al. 1988).

To examine cell propagation modes between the re-gimes (recall the definitional differences in terms of cellpropagation in regimes A, B, and their intersectionAB), cross-coast and along-coast winds at 700 hPa aredisplayed in Table 3 as an indicator of prevalent steer-ing flows. Recall that the cross-coast precipitating sys-tem movement was �7 m s�1 during regime A, and thealong-coast movement was �10 m s�1 during regime B.

FIG. 10. Diurnal cycle of precipitation feature statistics, broken down by N–S subdomain:northern half of analysis domain (dashed) and southern half (solid). (a) Feature volumetricrainfall. (b) Fraction of volumetric rainfall produced by convective pixels. (c) Mean featuremaximum dimension (i.e., ellipse major axis length). (d) Fraction of volumetric rainfall pro-duced by organized features (as defined in text).

1 MAY 2007 L A N G E T A L . 1727

At both Los Mochis and Mazatlán, slightly strongerregime-composited positive along-coast winds are ob-served in regime B compared with regimes A and AB.Note that precipitating system phase speeds slightly ex-ceed the cross-coast and along-coast winds, in regimesA and B, respectively, suggesting weak propagation.

Given the large time overlap with the A and B re-gimes, periods where the regime was classifed as B butnot A were also examined to highlight B periods exclu-sively. This yielded average along-coast winds at 700hPa of 6.8 (2.7) m s�1 at Mazatlán (Los Mochis) thatdiffer by a greater margin than the differences betweenB and A regimes alone, which largely overlap in time.This indicates that southerly flow is 2 m s�1 stronger, onaverage, when convective systems propagate with acoast-parallel component relative to times when thesteering flow is more coast-normal.

North American Regional Reanalysis data (NARR;Mesinger et al. 2006) are used to examine large-scaleinfluences on the aforementioned precipitation regimesidentified during NAME. The NARR reanalyses are a

subset of the National Centers for Environmental Pre-diction–National Center for Atmospheric Research(NCEP–NCAR) global reanalysis and, since 2003, havebeen run in near–real time as a forecasting tool. For thisanalysis, 3-hourly data were averaged at 700 hPa duringthe regime periods in order to ascertain correlationsbetween radar-observed precipitation features and thelarge-scale circulation.

Figure 14 shows the composite 700-hPa wind fieldand relative humidity fields for regime AB (Fig. 14a)and no regime (Fig. 14b). During regime AB (the 700-hPa characteristics of regime A—not shown—are simi-lar to regime AB), the radar domain is located in closeproximity to a tropical easterly wave (EW) trough (in acomposite sense) with a predominant southeasterlyflow component and high relative humidity advectinginto the tier I domain. In contrast, the composite EWposition during no regime is farther east and has aweaker intensity (in terms of the 700-hPa height gradi-ent). During these periods, the 700-hPa flow over theradar domain is easterly, and is more heavily influenced

FIG. 11. Centroid locations of organized features as a function of regime (see legends) during the following 4-h periods: (a) 1200–1600,(b) 1600–2000, (c) 2000–0000, (d) 0000–0400, (e) 0400–0800, and (f) 0800–1200 LT. Topography is shaded in the background (grayscale).


Fig 11 live 4/C

by the subtropical ridge centered over northernMexico. The composite pattern during regime B (notshown) is similar to the no-regime pattern except thatthere is slightly more moisture over the radar domain.

The increase in relative moistness and strength ofsoutheasterly flow in the NARR composite at 700 hPaduring regime AB is consistent with the analysis of thefield campaign soundings presented herein. The in-

creased shear and moisture associated with an EW pas-sage preferentially influences the southern portion ofthe radar domain, and is linked with the radar obser-vations of more numerous longer-lived organized con-vective features to the south. In addition, this suggeststhat the juxtaposition of the flow patterns associatedwith EWs in relation to the southern SMO is an impor-tant factor in determining convective system lifetimesand propagation modes in these regimes. However, it isclear that the signals are relatively subtle in the ob-served sounding data, and more detailed analyses arenecessary to quantify the role of EW forcing on theprecipitation regimes in the region.

6. Discussion and conclusions

The initial findings presented in this study demon-strate that the SMO plays an important role in the trig-

FIG. 12. Regime-composited skew T–logp diagrams from Los Mochis during (a) regime A, (b) regime B, (c)regime AB, and (d) no regime. Temperature (solid line), dewpoint (thin dashed line), and surface-based parcelascent path (thick dashed) are shown.

TABLE 2. Regime-averaged CAPE and CIN values at LosMochis.

Regime CAPE (J kg�1) CIN (J kg�1)

A 1539 68B 1672 75AB 1560 79No regime 1687 44

1 MAY 2007 L A N G E T A L . 1729

gering of precipitating systems in the core NAM region.In the along-coast reduced-dimension analysis, isolatedmaxima in precipitation frequency were found to alignroughly with local peaks in the mean along-coast eleva-

tion profile. Precipitation begins along the peaks andfoothills of the SMO in the late afternoon and movesWNW toward the GoC. During undisturbed periods,SMO convection does not survive the trip to the gulfafter sunset.

Intraseasonal variability in the study region is con-siderable. We identified two major disturbed regimes:A and B. During regime A there is enhanced precipi-tation over the coast and GoC, especially overnight andin the early morning. During regime B there is signifi-cant along-coast movement of precipitating systems.There is considerable—though not complete—overlapbetween these two regimes, such that when A was oc-curring, often so was B (and vice versa). This led to therecognition of a third disturbed regime: AB. The oc-currence of these regimes appears to be related to theenhancement of low-level shear in the environment,particularly near the southern gulf, allowing precipitat-

FIG. 13. Regime-averaged profiles of cross-coast (thin solid) and along-coast (thin dashed) wind components areplotted from Los Mochis (black) and Mazatlán (blue) for (a) regime A, (b) regime B, (c) regime AB, and (d) noregime. Los Mochis (green solid) and Mazatlán (green dashed) profiles of wind speed are also displayed.

TABLE 3. Regime-averaged 0–4-km wind shear and 700-hParotated wind components at Los Mochis and Mazatlán.

Site Regime

0–4-kmshear

(m s�1)

Cross-coast700 hPa(m s�1)

Along-coast700 hPa(m s�1)

Los Mochis A 2.6 �1.8 1.9B 3.0 �1.6 2.0AB 3.0 �1.6 1.8No regime 2.8 �1.2 2.1

Mazatlán A 5.7 �1.9 4.0B 6.5 �2.3 4.6AB 6.7 �2.7 4.2No regime 2.6 �1.2 2.4


Fig 13 live 4/C

ing systems to scale upward and have longer life cycles(e.g., Rotunno et al. 1988).

During regime AB, organized convection seems tobe weakly propagating in excess of steering winds: 3–4m s�1 from the SMO to the sea and 4–5 m s�1 NWalong the SMO–GoC major axis. The cold pool re-quired to effect this implied propagation is 500–1000 mdeep and has 1–2°C negative buoyancy (Keenan andCarbone 1992), so either very shallow or extremelyweak cold pools are all that is needed to explain thisbehavior. More elaborate mechanisms, such the Mapeset al. (2003) gravity wave hypothesis originally used toexplain observations of convection moving at 15 m s�1

up to several hundred kilometers offshore of Colombia,are not needed.

During the NAME IOP, there were at least two sig-nificant gulf surges (Higgins et al. 2006). These surgestended to overlap with disturbed precipitation regimes(both A and B). For example, the first identified gulfsurge overlapped with the first appearances of regimesA and B, around 12–13 July 2004. Overall, time periodsassociated with gulf surges form a small subset of re-gime A and B periods. However, there were many dis-turbed regime periods not associated with a gulf surge,particularly during August. Thus, during 2004 gulfsurges could have played only a small role in tier Iprecipitation. When they occurred, they appeared to beimportant (i.e., were associated with disturbed re-gimes), but they did not occur often enough to be amajor factor.

This study also suggests a possible link betweentropical EWs and precipitation regimes, which couldhave important implications for the initiation of mois-

ture surges in the GoC. Although a number of previousinvestigators have identified relationships betweentropical EW and moisture surges in the GoC (Hales1972; Brenner 1974; Stensrud et al. 1997; Fuller andStensrud 2000; Anderson et al. 2000), the understand-ing of mechanism(s) whereby EWs trigger surges re-mains elusive. It may be that the precipitation featuresidentified herein, through the action of cold pool and/orgravity wave dynamics, play a role in surge initiation.This is an important topic for future research.

Our hypothesis of the importance of the rainfall con-tribution of organized rainfall modes is supported bythe evidence. The development and propagation of or-ganized systems (e.g., MCSs and other large precipita-tion features) are key components of the diurnal cycleof precipitation in this region, particularly during dis-turbed regimes. During disturbed periods, organizedfeatures account for �75% of all feature-producedrainfall (Table 1). This fraction is still large during un-disturbed periods, �60%. Organized systems are par-ticularly important for rainfall in the evening along theSMO, and during the early morning along the southernportions of the coast and GoC.

Acknowledgments. The lead S-Pol engineer was DonFerraro of NCAR’s Earth Observing Laboratory(EOL). Jon Lutz, of NCAR/EOL, provided engineer-ing oversight for the S-Pol operation and led the SMNradar upgrade and data collection efforts. He was as-sisted in the latter by Arturo Valdez-Manzanilla of Jua-rez University and by Armando Rodriguez Davila ofSMN. We thank all the other NCAR engineers, tech-nicians, and field scientists, as well as Robert Bowie of

FIG. 14. Regime-averaged NARR 700-hPa geopotential height contours (dam), relativehumidity (%, shaded), and wind vectors (see vector scale at upper right) for (a) regime ABand (b) no regime. The white dashed line in the left panel shows the approximate location ofan easterly wave trough. Approximate location of S-Pol is also shown.

1 MAY 2007 L A N G E T A L . 1731

the CSU-CHILL radar staff, for contributing to the suc-cessful S-Pol operation. Bob Rilling of NCAR/EOL ledthe initial quality control and distribution of the radardata. Further data quality control efforts, as outlined inthe appendix, were assisted by Chad Chriestenson, LeeNelson, and Gustavo Pereira of Colorado State Univer-sity (CSU). Sounding data were obtained from RichardJohnson and Paul Ciesielski of CSU. We thank SMNfor their cooperation in making the S-Pol deploymentand the SMN upgrades possible. Radar observationsand analyses were funded by the National Oceanic andAtmospheric Administration (NOAA; SMN radars)and the National Science Foundation (NSF; S-Pol).NCAR is funded by NSF.

APPENDIX

Data Quality Control

a. Quality control of S-Pol radar data

S-Pol data were corrected for attenuation as well asclutter, insect, and second-trip contamination. Nonme-teorological echo was removed via thresholds on vari-ous polarimetric fields. Any remaining spurious echowas removed by hand using NCAR’s soloii software.Differential phase (�DP) was filtered using a 21-gate(3.15 km based on 150-m gate spacing) finite-impulseresponse filter (developed by J. Hubbert of NCAR andV. N. Bringi of Colorado State University). Specificdifferential phase (KDP) was calculated from the slopeof a line fitted to the filtered �DP field. The window ofthe fitted line varied from 31 to 11 gates (4.65–1.65 km)as reflectivity (ZH) increased. The rainfall attenuationcorrection methodology was based on Carey et al.(2000b). The value of ZH for all radars (including SMN)was further corrected for gaseous attenuation followingBattan (1973, chapter 6).

Significant amounts of beam blockage due to terrainoccurred in S-Pol’s northeast sector (351°–105° azi-muth). Correction of this blockage followed the basicmethodology of Carey et al. (2000a). The locations ofthe blocks were determined to the nearest degree inazimuth and nearest kilometer in range by visual in-spection of clear-air radar sweeps. Then, in the blockedregions, we examined the behavior of ZH as a functionof azimuth for a given range of KDP. Due to the self-consistency between polarimetric variables in rain(Scarchilli et al. 1996), for a given range of KDP,ZH should vary only over a small range as well. Thedifference in the median ZH values in unblocked re-gions, and median ZH values in a blocked ray, is thepositive dBZ correction that needs to be applied to ZH.We only used this methodology to correct small blocks

(down to �5 dBZ). For larger blocks, we used the fol-lowing methodology. If 0.8° had a severe block in aparticular ray (reduction �5 dBZ), then we used infor-mation from 1.3° at all ranges greater than that of theblock. If the 1.3° ray itself was severely blocked, thenwe resorted to 1.8° (which was never blocked morethan 5 dBZ). In addition, we filled in low-level gapscaused by clutter removal (0.8° and 1.3°) using infor-mation from higher sweeps (1.3° and 1.8°). As this cor-rection technique is experimental, S-Pol data within the351°–105° azimuths are expected to be of inferior qual-ity to the data outside these boundaries. If we correctedblocked ZH by applying a positive dBZ offset, then weset the differential reflectivity to a missing data value.

Rain rates were calculated using the Colorado StateUniversity blended rainfall algorithm (Cifelli et al.2002). This algorithm varies between various polari-metric rainfall estimators depending on the values ofthe polarimetric variables and the presence of mixed-phase precipitation. The Z–R relationship used as partof the algorithm was Z�221R1.25. This Z–R, usedmainly in light rain, was determined via intercompari-sons of reflectivity with gauge rain rates at the NOAAprofiler site �45 km northwest of S-Pol.

b. Quality control of SMN radar data

We used the most complete Cabo and Guasavesweeps closest in time to each 15-min mark. We appliedautomated filters on ZH, noise-corrected power, and ontotal power. Due to antenna backlash (a lag betweenradar gears, the servo mechanism, and encoders thatmanifests itself as an offset between azimuths obtainedduring clockwise and counterclockwise motions of theantenna) when Guasave changed spin direction everyfew days, that radar required a small correction to themeasured azimuths. We then applied an automatedclutter filter to the Guasave data. This clutter filter que-ried a clutter map created from clear-air Guasavesweeps taken over several days. Clutter-affected gateswere removed. Cabo data did not have many stormsoverrun its clutter, so its clutter was hand edited only.We hand edited the filtered datasets for any remainingclutter, noise, second-trip contamination, and insectsusing soloii.

A reflectivity offset was then applied to the databased on visual and statistical intercomparisons withS-Pol reflectivities. The statistical evaluation comparedthe closest gates within 500 m in the horizontal and200 m in the vertical distance. Histograms of reflectivitydifferences were obtained from this statistical intercom-parison. In addition, visual intercomparison of well-placed echoes was done. Based on these methods, a


reflectivity correction was applied to the SMN radardata.

An attenuation correction by rain was based on thePatterson et al. (1979) algorithm, which uses a Z–Rrelationship to estimate rainfall, then iteratively cor-rects ZH at a gate based on the theoretical treatment ofattenuation by all the rainfall up to the given gate. TheZ–R scheme used was the same as the S-Pol Z–Rscheme. Based on intercomparions with S-Pol andTRMM precipitation radar data, we believe that thecorrected SMN ZH measurements are accurate towithin 1–2 dBZ. No blockage correction was attemptedat Cabo or Guasave, but blockage was only a minorproblem for these radars, either due to a lack of blocks(Guasave) or a lack of storms in blocked areas (Cabo).The SMN radar rainfall rates were determined from theaforementioned Z–R relationship, with capping at 53dBZ (231.5 mm h�1) to reduce ice contamination. Moreinformation on NAME radar quality control can befound in Lang et al. (2005) and online (http://radarmet.atmos.colostate.edu/~tlang/readme_NAME_regional_radar_composites_v2.pdf).

REFERENCES

Ahijevych, D. A., C. A. Davis, R. E. Carbone, and J. D. Tuttle,2004: Initiation of precipitation episodes relative to elevatedterrain. J. Atmos. Sci., 61, 2763–2769.

Anderson, B. T., J. O. Roads, and S.-C. Chen, 2000: Large-scaleforcing of summertime monsoon surges over the Gulf of Cali-fornia and the southwestern United States. J. Geophys. Res.,105, 24 455–24 468.

Battan, L. J., 1973: Radar Observation of the Atmosphere. Uni-versity of Chicago Press, 324 pp.

Brenner, I. S., 1974: A surge of maritime tropical air—Gulf ofCalifornia to the southwestern United States. Mon. Wea.Rev., 102, 375–389.

Carbone, R. E., J. D. Tuttle, D. Ahijevych, and S. B. Trier, 2002:Inferences of predictability associated with warm season pre-cipitation episodes. J. Atmos. Sci., 59, 2033–2056.

Carey, L. D., R. Cifelli, W. A. Petersen, and S. A. Rutledge,2000a: Preliminary report on TRMM-LBA rainfall estima-tion using the S-Pol radar. Department of Atmospheric Sci-ence Paper 697, Colorado State University, 19 pp.

——, S. A. Rutledge, D. A. Ahijevych, and T. D. Keenan, 2000b:Correcting propagation effects in C-band polarimetric radarobservations of tropical convection using differential propa-gation phase. J. Appl. Meteor., 39, 1405–1433.

Cifelli, R., W. A. Petersen, L. D. Carey, S. A. Rutledge, and M. A.F. da Silva Dias, 2002: Radar observations of the kinematic,microphysical, and precipitation characteristics of two MCSsin TRMM LBA. J. Geophys. Res., 107, 8077, doi:10.1029/2000JD000264.

Douglas, M. W., 1995: The summertime low-level jet over theGulf of California. Mon. Wea. Rev., 123, 2334–2347.

Fuller, R. D., and D. J. Stensrud, 2000: The relationship between

tropical easterly waves and surges over the Gulf of Californiaduring the North American monsoon. Mon. Wea. Rev., 128,2983–2989.

Hales, J. E., Jr., 1972: Surges of maritime tropical air northwardover the Gulf of California. Mon. Wea. Rev., 100, 298–306.

Hartmann, D. L., H. H. Hendon, and R. A. Houze Jr., 1984: Someimplications of the mesoscale circulations in tropical cloudclusters for large-scale dynamics and climate. J. Atmos. Sci.,41, 113–121.

Higgins, R. W., and Coauthors, 2003: Progress in pan AmericanCLIVAR research: The North American monsoon system.Atmosfera, 16, 29–65.

——, and Coauthors, 2006: The NAME 2004 field campaign andmodeling strategy. Bull. Amer. Meteor. Soc., 87, 79–94.

Houze, R. A., Jr., 1993: Cloud Dynamics. Academic Press, 573 pp.Keenan, T. D., and R. E. Carbone, 1992: A preliminary morphol-

ogy of precipitation systems in tropical northern Australia.Quart. J. Roy. Meteor. Soc., 118, 283–326.

Lang, T. J., R. Cifelli, L. Nelson, S. W. Nesbitt, G. Pereira, S. A.Rutledge, D. Ahijevych, and R. Carbone, 2005: Radar obser-vations during NAME 2004. Part I: Data products and qualitycontrol. Preprints, 32d Conf. on Radar Meteorology and 11thConf. on Mesoscale Processes, Albuquerque, NM, Amer. Me-teor. Soc., CD-ROM, JP3J.5.

Mapes, B., T. T. Warner, and M. Xu, 2003: Diurnal patterns ofrainfall in northwestern South America. Part III: Diurnalgravity waves and nocturnal convection offshore. Mon. Wea.Rev., 131, 830–844.

Mesinger, F., and Coauthors, 2006: North American Regional Re-analysis. Bull. Amer. Meteor. Soc., 87, 343–360.

Negri, A. J., R. F. Adler, R. A. Maddox, K. W. Howard, and P. R.Keehn, 1993: A regional rainfall climatology over Mexico andthe southwest United States derived from passive microwaveand geosynchronous infrared data. J. Climate, 6, 2144–2161.

Nesbitt, S. W., E. J. Zipser, and D. J. Cecil, 2000: A census ofprecipitation features in the Tropics using TRMM: Radar, icescattering, and lightning observations. J. Climate, 13, 4087–4106.

——, R. Cifelli, and S. A. Rutledge, 2006: Storm morphology andrainfall characteristics of TRMM precipitation features. Mon.Wea. Rev., 134, 2702–2721.

Patterson, V. L., M. D. Hudlow, P. J. Pytlowany, F. P. Richards,and J. D. Hoff, 1979: GATE radar rainfall processing system.NOAA Tech. Memo. EDIS 26, Washington, DC, 34 pp.[Available from National Technical Information Service,5285 Port Royal Rd., Springfield, VA 22161.]

Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory forstrong, long-lived squall lines. J. Atmos. Sci., 45, 463–485.

Scarchilli, G., E. Gorgucci, V. Chandrasekar, and A. Dobaie,1996: Self-consistency of polarization diversity measurementof rainfall. IEEE Trans. Geosci. Remote Sens., 34, 22–26.

Simpson, J. R., R. F. Adler, and G. R. North, 1988: A proposedTropical Rainfall Measuring Mission (TRMM) satellite. Bull.Amer. Meteor. Soc., 69, 278–295.

Stensrud, D. J., R. L. Gall, and M. K. Nordquist, 1997: Surges overthe Gulf of California. Mon. Wea. Rev., 125, 417–437.

Yuter, S. E., and R. A. Houze Jr., 1998: The natural variability ofprecipitating clouds over the western Pacific warm pool.Quart. J. Roy. Meteor. Soc., 124, 53–99.

1 MAY 2007 L A N G E T A L . 1733

radar-observed characteristics of precipitating systems during

Documents