Abstract

This study focuses on modulation of lightning and convective vertical structure in the southern Amazon as a function of the South American monsoon V index (VI). Four wet seasons (December–March 1998–2001) of Tropical Rainfall Measuring Mission (TRMM) Lightning Imaging Sensor (LIS) and Precipitation Radar (PR) data are examined together with two wet seasons (2000–01) of ground-based Brazilian Lightning Detection Network (BLDN) data. These observations are composited by VI phase (northerly or southerly) for a region of the southern Amazon and discussed relative to VI-regime environmental characteristics such as thermodynamic instability and wind shear.

Relative comparisons of VI-regime convective properties reveal 1) slightly larger (20%–25%) PR pixel-mean rainfall during periods of northerly VI due to increased stratiform precipitation, 2) a factor of 2 or more increase in lightning flash density and the lightning diurnal cycle amplitude during periods of southerly VI, 3) a factor of 1.5–2 increase in the conditional probability of any PR radar reflectivity pixel exceeding 30 dBZ above the −10°C level during periods of southerly VI, and 4) an associated factor of 2 or more increase in southerly VI pixel-mean ice water path, with the ice water path being highly correlated to trends in lightning activity. During periods of southerly VI, convection occurs in an environment of increased thermodynamic instability, weak southeasterly low-level, and deep upper-tropospheric easterly wind shear. During periods of northerly VI, low-level westerly shear opposes stronger deep tropospheric easterly shear in a relatively moist environment of weaker thermodynamic instability, consistent with the occurrence of more widespread stratiform precipitation.

The composite results of this study point to 1) regime differences in convective forcing that alter the prevalence of ice processes and, by inference, the vertical profile of latent heating and 2) the utility of lightning observations in delineating convective regime changes.

1. Introduction

Seasonal climatologies of tropical convection, precipitation, and lightning reflect the integrated effect of individual events that occur over intraseasonal time scales. These convective events and their associated forcing are often described or associated with “intraseasonal variability” and occur on time scales of days to tens of days. Well-known examples of intraseasonal variability in tropical convection include disturbances ranging from the 30–60 day time scale (Madden and Julian 1971) to disturbances occurring over much shorter time scales, such as the classical 3–5-day tropical easterly wave (Riehl 1954; Lau and Lau 1990; Petersen et al. 2003). Similar changes in precipitation, convective structure, and lightning occurring on intermediate scales of 10–20 days have also been documented, for example, during active and break periods of tropical monsoon systems over the Maritime Continent region of northern Australia (e.g., Rutledge et al. 1992; Williams et al. 1992). Previous research in the literature has established that wet season rainfall in South America (Nogues-Paegle and Mo 1997), as well as in other monsoon regions (Webster et al. 1998), is modulated primarily by intraseasonal variability of the large-scale monsoon circulation. However, the mechanism(s) by which tropical and/or subtropical-centered intraseasonal changes influence convective system type remains unclear.

For example, during the six-week Tropical Rainfall Measuring Mission—Large-scale Biosphere Atmosphere (TRMM–LBA) field campaign of January–February 1999 (Silva Dias et al. 2002) it was noted that systematic variations in convective vertical structure, lightning, and rainfall over the southwestern Amazon (SWAMZ) were highly correlated to variations (easterly or westerly) in the low-level zonal wind and other associated parameters such as convective available potential energy (CAPE) and even aerosol loading (Cifelli et al. 2002; Williams et al. 2002; Halverson et al. 2002; Rickenbach et al. 2002; Petersen et al. 2002; Jones and Carvalho 2002; Cifelli et al. 2004). Though the daily total rainfall occurring in each zonal wind regime (east and west) of TRMM–LBA was similar, the convection occurring in easterly (westerly) regimes was less (more) widespread but more (less) vertically developed and produced more (less) lightning in a regime of higher (lower) latent heat fluxes (Betts et al. 2002), higher (lower) CAPE (∼500 J kg−1 difference in the mean), and stronger easterly (westerly) low-level wind shear (Halverson et al. 2002).

Though the systematic variability in convection as a function of zonal wind direction was highly evident in the TRMM–LBA observations, the sampling domain was considerably limited in time and space. To extend the TRMM–LBA results and to examine the continental-scale variability of the convective signal, Petersen et al. (2002) used three years of TRMM PR, LIS, and passive microwave imager (TMI) data together with a continuous ground-based Brazilian Lightning Detection Network (BLDN) to further characterize the vertical structure, lightning, and rainfall characteristics of South American convection as a function of SWAMZ 850-mb zonal wind anomalies. Collectively, the TRMM satellite-based results showed that regional intraseasonal variability in wet-season convective structure 1) was evident over the southern Amazon, Altiplano, and southern Brazil; 2) varied in a consistent fashion over the southern Amazon with that observed during TRMM– LBA; and 3) occurred with composited synoptic signals over the South American continent during each mode of variability that were consistent with hypothesized extratropical forcing discussed in Garreaud and Wallace (1998) and Rickenbach et al. (2002).

Alternatively, Wang and Fu (2002, hereafter WF02) found that changes in the large-scale monsoon circulation and accompanying precipitation anomalies over South America were more closely linked to variations of the cross-equatorial meridional flow. Following the definition of a “universal monsoon index” (Lu and Chan 1999), WF02 found that the deviations in the 925-hPa meridional wind over the northwestern Amazon (AMZ: 5°S–5°N, 65°–75°W) best captured the seasonal reversal of the monsoon circulation (Figs. 1a,b). Subsequently, they defined a metric of area-averaged daily 925-hPa meridional winds over this spatial domain as the South American monsoon “V index.” In this context, a large positive (negative) V index (e.g., ≥2 m s−1, with seasonal mean removed) indicates a southerly (northerly) flow regime. WF02 further found that changes in the V index were well correlated to latitudinal shifts of precipitation and the associated upper-level circulation (Figs. 1c,d) on submonthly (5–14 days), seasonal, and interannual time scales over tropical and subtropical South America. When the V index was southerly (or less northerly) precipitation was located mainly on or to the north of the equator and the Bolivian high (200 mb) was shifted slightly southwestward. Conversely, when the V index was northerly (or less southerly), precipitation increased over the Amazon Basin and the subtropics and the Bolivian high moved northeastward (Figs. 1c,d).

Fig. 1.

Composites of ECMWF 925-hPa wind (m s−1; vectors) and precipitation (mm day−1; shaded) for periods of (a) southerly VI ≥ 2 m s−1 and (b) northerly VI ≤ 2 m s−1. (c) As in (a) but for 200 hPa; (d) as in (b) but for 200 hPa. Adapted from WF02. Boxes indicate approximate TRMM analysis domain.

Fig. 1.

Composites of ECMWF 925-hPa wind (m s−1; vectors) and precipitation (mm day−1; shaded) for periods of (a) southerly VI ≥ 2 m s−1 and (b) northerly VI ≤ 2 m s−1. (c) As in (a) but for 200 hPa; (d) as in (b) but for 200 hPa. Adapted from WF02. Boxes indicate approximate TRMM analysis domain.

Systematic changes in the large-scale monsoon circulation over South America and the spatial and temporal changes of precipitation pattern associated with those of the V index raise important questions as to whether and/or how convective system type might also change with the latter. Clarifying this question is important to determining the intraseasonal change of vertical structure of atmospheric heating (e.g., Hartmann et al. 1984; Tao et al. 1990; Mapes and Houze 1993; Schumacher et al. 2004; Cifelli et al. 2004), surface hydrological conditions (e.g., soil moisture, runoff, streamflow/flood), upper-tropospheric ice and water vapor content (e.g., Sun and Lindzen 1993; Price 2000), and also the changes of convective transport of chemical tracers (Scala et al. 1990; Pickering et al. 1995) associated with large-scale monsoon circulation. As a first step toward answering these questions, this study examines changes in convective system vertical structure as related to vertical profiles of buoyancy and wind shear associated with intraseasonal changes in the V index during the wet season over South America. In doing so, we aim to directly link the changes of convective structure, lightning, and microphysical properties to the dynamics of the monsoon circulation. Such links also serve as an observational basis for improving the space-borne detection, model simulation, and, ultimately, the prediction of precipitation intensity as a function of the large-scale circulation pattern over the South American region. Use of the universal monsoon index also allows us to potentially link our results for South America to a unified understanding of convective system type with monsoon circulation globally.

We use TRMM Precipitation Radar (PR), Lightning Imaging Sensor (LIS), and BLDN data to study the changes of convective precipitation structure with the V index. We contrast our results with the studies of Cifelli et al. (2002), Williams et al. (2002), and Petersen et al. (2002), who conducted similar studies based on transitions of zonal wind over the Amazon. We further consider the thermodynamic and dynamic environments of the two V index regimes by performing composite analyses on thermal buoyancy profile and tropospheric wind shear.

2. Data and methodology

Four years of wet season (December–March 1998–2001) TRMM LIS and PR data were partitioned by the V index (VI) regime to create domain maps of mean lightning flash density (FD) and PR-diagnosed ice water path (IWP), contoured frequency diagrams of PR reflectivity vertical structure, and diurnal cycles of rainfall rate, lightning, and IWP. Determination of VI regime days relied on the previous partitioning methods/data of WF02. TRMM VI regime means were then constructed by using the strongest 10 north and south VI days of each wet season month (40 days per season, 160 days total in each regime). Cross referencing the 160 sample days in each regime with TRMM orbits containing a part of the PR swath in at least some portion of the analysis domain (10° × 10° box discussed below) resulted in a TRMM sample dataset of approximately 400 orbits in each regime. Approximately 93% of these orbits possessed PR-detectable precipitation somewhere within the domain.

As a complement to the direct sampling of precipitation echo by radar, lightning observations provide an expedient method for detecting regions of cloud development associated with strong updrafts and associated ice processes [cf. MacGorman and Rust (1999) for a review). To this end, TRMM LIS lightning data were used to provide a rapid and reliable means for systematically identifying regions of, and/or tendencies for, differing convective vertical structure as a function of transitions in VI regime both spatially and diurnally. In particular, over south-central Amazonia (SCAMZ) the difference in lightning FD as a function of VI was found to be strong (Figs. 3a–c; 100%–300% increase in FD between regimes; significant at p values <5%) and to overlap substantially with a corresponding, but anticorrelated, change in precipitation (Fig. 1). Further, the diurnal cycle of LIS-detected FD (section 3b) also indicated a robust difference in VI-regime FD behavior over the SCAMZ. Given these large differences in FD behavior, the remainder of the analysis was focused primarily on the SCAMZ domain.

Fig. 3.

TRMM LIS lightning flash density (Fl km−2 month−1) for (a) south, (b) total (all days), and (c) north VI regimes. Box indicates south-central Amazon domain.

Fig. 3.

TRMM LIS lightning flash density (Fl km−2 month−1) for (a) south, (b) total (all days), and (c) north VI regimes. Box indicates south-central Amazon domain.

To compute diurnal cycles of TRMM LIS lightning FD and PR rainfall over the SCAMZ for each VI regime, lightning flash data, and 2A25 stratiform and convective rainfall rates were accumulated and averaged unconditionally (i.e., flashing and nonflashing, raining and nonraining) over 4-h time bins for each individual hour of the day in order to mitigate sampling bias (Negri et al. 2002). Because it was not clear (a priori) which particular 4-h interval was appropriate to plot for examination of the diurnal cycle, the 4-h means centered on each individual hour were plotted. The resultant orbit sampling distribution (number of orbits occurring in four hour increments, centered on each hour, divided by the total number of orbits) varied by less than 2% from peak to trough over the diurnal cycle. Given the sampling distribution behavior and the number of days in the sample (e.g., Lin et al. 2000), the analysis should reflect the gross structure of the diurnal cycle of rainfall for each regime. Indeed, the diurnal cycle of rainfall compares reasonably well to the much larger sample reported in Nesbitt and Zipser (2003).

To further investigate VI-regime variability using a more temporally continuous proxy for convection we also analyzed intraseasonal time series and diurnal cycles of Brazilian Lightning Detection Network (Blakeslee et al. 2003) cloud-to-ground lightning (CG) flash counts for two wet seasons (1999–2000, 2000–01; Figs. 2a,b). The BLDN CG flash counts (collected on a 24/7 basis) were tabulated over a domain similar in size to the SCAMZ and over an area that encompassed the majority of the SCAMZ. The CG flash counts are most representative of the northwestern quarter of SCAMZ box where the center of the BLDN antenna network is located and best lightning detection efficiency is realized. Time series of BLDN CG flash counts and VI were compared for each of the two wet seasons to assess the temporal correlation in behavior between the two parameters. Diurnal cycles of BLDN lightning flash count were also computed (not shown) for each VI regime for both wet seasons in order to further validate the diurnal cycles computed using TRMM data.

Fig. 2.

(a) Five-day running means of BLDN lightning flash count anomaly (left ordinate; solid line) and VI anomaly (right ordinate; dashed line) for December 1999–March 2000. (b) As in (a), but for December–March 2000–01.

Fig. 2.

(a) Five-day running means of BLDN lightning flash count anomaly (left ordinate; solid line) and VI anomaly (right ordinate; dashed line) for December 1999–March 2000. (b) As in (a), but for December–March 2000–01.

As a direct measure of convective vertical structure variability in the SCAMZ box, TRMM PR reflectivity data (rain certain pixels; i.e., pixels existing in a given ray classified by the 2A25 algorithm as containing precipitation) were partitioned by VI regime and subsequently binned into 2-dBZ reflectivity and 1-km height intervals in order to construct contoured frequency diagrams as a function of altitude (i.e., CFADs; Yuter and Houze 1995). The CFADs provide the probability of sampling a given reflectivity at a given height conditioned on the presence of a measurable reflectivity being detected at that height. Note that the minimum detectable Z for the PR is ∼18 dBZ, so the Z distributions are truncated. The TRMM PR Z values were also used to construct maps and diurnal cycles of VI-regime mean ice water path IWP for precipitation-sized ice occurring at temperatures ≤ −10°C (Figs. 4c, 7). For calculation of IWP all TRMM PR-sampled pixels were used, precipitating or not, to create unconditional means. Total (all pixels regardless of type), convective, and stratiform IWPs were calculated using a ZM relationship to first compute ice water content at temperatures colder than −10°C (Petersen and Rutledge 2001), followed by an integration of the ice water contents to the 18-dBZ echo top. Ice particle size distributions were modeled as exponential with a fixed N0 of 4 × 106 m−4. Bulk ice particle densities in the ZM were varied as a function of reflectivity (e.g., Black 1990) and precipitation type (convective or stratiform) for each range gate. For example, PR pixels identified as convective and with Z ≥ 40 dBZ were assigned a density of 0.8 g cm−3. For convective Z, 30 dBZ < Z < 40 dBZ, the density decreased linearly to 0.4 g cm−3 and then was kept constant at 0.4 g cm−3 for all pixels with Z ≤ 30 dBZ. Stratiform reflectivities were assigned densities of 0.1 g cm−3 assuming the presence of lower density crystal aggregates/snow.

Fig. 4.

Diurnal cycles of NVI and SVI over the SCAMZ for (a) TRMM LIS total lightning flash rate; (b) TRMM PR total, convective, and stratiform rainfall rate; (c) as in (b) but ice water path (kg m−2).

Fig. 4.

Diurnal cycles of NVI and SVI over the SCAMZ for (a) TRMM LIS total lightning flash rate; (b) TRMM PR total, convective, and stratiform rainfall rate; (c) as in (b) but ice water path (kg m−2).

Fig. 7.

As in Fig. 3, but for mean convective ice water path (kg m−2) for (a) SVI, (b) total, and (c) NVI regimes.

Fig. 7.

As in Fig. 3, but for mean convective ice water path (kg m−2) for (a) SVI, (b) total, and (c) NVI regimes.

To place the precipitation and lightning measurements within the broader context of the VI north/south environments, we also analyzed available sounding data for three rawinsonde stations located on the western side of the SCAMZ box. The stations were located at Alta Floresta: 9.87°S, 56.1°W, Vilhena: 12.73°S, 60.13°W, and Cuiaba: 15.6°S, 56.1°W. Standard thermodynamic and wind data for all mandatory and significant levels from the surface to tropopause for 0000 UTC soundings were collected and composited (i.e., averaged) by VI regime for the 10 strongest VI days in each regime for each month. The 0000 UTC soundings were selected for analysis because the actual launch time (1–1.5 h prior to the standard time) would have been approximately 1830 LT, a period closer to the expected peak in the diurnal cycle of convection (e.g., Rickenbach et al. 2002; Nesbitt and Zipser 2003; and consistent with the results presented herein).

Prior to averaging the sounding data, the data were quality controlled to exclude the presence of frontal boundaries and direct precipitation influences on the sounding measurement. Frontal boundaries were removed by manual/visual inspection of National Centers for Environmental Prediction (NCEP) reanalysis vector winds and temperatures at the 850-hPa level. For a given sounding, if the vector winds indicated a possible frontal zone (e.g., obvious shift in the wind across the box) and the temperature data indicated a gradient across the analysis box of 2°C or greater, the sounding was discarded. To remove soundings influenced by nearby precipitation-driven outflows, etc., rainfall data from the TRMM 3B42 product (3-h resolution, 0.25° grid) were examined for grid boxes corresponding to each sounding location. When 3-h-averaged rainfall rates of ≥1 mm h−1 were observed over sounding grid boxes in the time window preceding the sounding standard time, the soundings were discarded. As a result of the sounding quality control, approximately 60–65 days were eliminated from the 160 possible days in each VI category. For the remaining 95–100 days, approximately 60 soundings in each regime collected at 0000 UTC were available for averaging.

The available sounding data were subsequently interpolated to intervals of 10 hPa below 700 hPa and to intervals of 50 hPa above for heights above the 700-hPa level. From the composited sounding data, joint distributions of thermodynamic and dynamic parameters such as temperatures, relative humidity, parcel buoyancy (mixed layer parcel from 930-mb level), deep tropospheric wind shear (layer mean wind 200–250 hPa − 925–850-hPa layer mean), and lower-tropospheric wind shear (700–925 hPa) were computed (Figs. 9 –11).

Fig. 9.

(a) Relative frequency (%) of thermal buoyancy for the SVI regime at 0000 UTC. Departure from moist neutrality (°C) plotted on x axis, height (m) on y axis, and relative fraction contoured at intervals of 5, 10, 15, and 20 (values >10% shaded). A gray-shaded triangle is used to highlight the difference in slope of the lower-level outer buoyancy envelope (5% contour) with height. (b) As in (a) but for NVI.

Fig. 9.

(a) Relative frequency (%) of thermal buoyancy for the SVI regime at 0000 UTC. Departure from moist neutrality (°C) plotted on x axis, height (m) on y axis, and relative fraction contoured at intervals of 5, 10, 15, and 20 (values >10% shaded). A gray-shaded triangle is used to highlight the difference in slope of the lower-level outer buoyancy envelope (5% contour) with height. (b) As in (a) but for NVI.

Fig. 11.

Distributions of (a) low-level and (b) deep tropospheric wind shear for the (top) NVE and (bottom) SVI regimes. The speed of the wind shear vector is noted for each regime on the ordinate and shear direction is indicated on the abscissa. The relative frequencies (%) for each wind shear bin are color shaded with values indicated in the legend.

Fig. 11.

Distributions of (a) low-level and (b) deep tropospheric wind shear for the (top) NVE and (bottom) SVI regimes. The speed of the wind shear vector is noted for each regime on the ordinate and shear direction is indicated on the abscissa. The relative frequencies (%) for each wind shear bin are color shaded with values indicated in the legend.

3. Results

a. Intraseasonal variability of lightning and the V index

Owing to a lack of continuous, broad-area radar coverage over much of the Amazon, area-integrated lightning flash counts are used as a proxy for the identification of marked changes in convective vertical structure, and in particular in situ ice processes, that may occur with transitions in the VI (and associated meteorological regime; Williams et al. 1992; Rutledge et al. 1992; Petersen and Rutledge 2001; Petersen et al. 2002, 2003; Cifelli et al. 2002, 2004). In this study, BLDN CG lightning data collected during the 1999–2000 and 2000–01 wet seasons (December–March) provide the requisite data continuity, temporal resolution, and area coverage to quickly infer intraseasonal changes in convective structure as a function of VI regime (Figs. 2a,b). In Fig. 2 it is clear that intraseasonal variability in both VI and BLDN CG flash counts is common, as indicated by the presence of periodic peaks and troughs in the anomaly values (anomalies relative to the monthly means). In fact, Fig. 2 also suggests that there is coherence in the two signals with a tendency for relative maximums and minimums in lightning to occur nearly coincident with, or slightly lag, southerly VI (SVI) and northerly VI (NVI) regimes respectively (lag correlation 0.6 at 1–2 days). The lag between lightning and VI seems most pronounced during the heart of the wet season (January–February). The relative coherence between the two variables is interesting given the exclusivity of the two datasets being compared [one via European Centre for Medium-Range Weather Forecasts (ECMWF) model analyses and the other large-area observations not assimilated into the model].

The trends in lightning data suggest that periods of deep convection in SVI (NVI) regimes are characterized by a subpopulation of cells that are more (less) vertically developed; that is, exhibiting stronger (weaker) updrafts just below and above the freezing level and more (less) robust mixed phase ice processes. Importantly, precipitation trends reported in WF02 suggest more extensive precipitation across the domain sampled by the BLDN during NVI periods. Hence, the SVI peaks in the lightning data shown in Figs. 2a,b do not necessarily reflect increases in area-mean precipitation, but do emphasize the occurrence of systematic changes in the overall convective vertical structure above the freezing level (e.g., prominence of the ice phase). This is an important distinction because VI-regime differences in rainfall reflect differences in column integrated latent heating, whereas the inferred changes in convective vertical structure reflect changes in vertical mass flux and heating profile (Cifelli et al. 2004), especially latent heating in the upper troposphere.

To significantly expand the spatial coverage of our analysis we turn to TRMM LIS observations of total lightning. Here we rely on composited LIS data and VI values collected over a total of four separate wet seasons occurring between December and March for the years 1997 to 2001. Maps showing the resultant difference in total lightning flash density between NVI regime, overall mean, and SVI regime are plotted in Figs. 3a–c. Large positive flash density differences (1–4 flashes km−2 month−1; significant at a p value <0.05), occur over the SCAMZ and are associated with the SVI regime. The lightning flash density differences depicted in Figs. 3a–c are similar to those observed in easterly and westerly wind regimes over the southern Amazon (cf. Petersen et al. 2002, their Fig. 4). Indeed, WF02 noted that, when NVI (SVI) regimes were present, there were general increases (decreases) in westerly zonal flow over the Amazon Basin and subtropics and even some similarity in the large-scale flow patterns of both regimes. This similarity may reflect a possible connection between intraseasonal monsoon circulations and equatorward incursion of extratropical synoptic systems, as reported in Garreaud and Wallace (1998), Liebmann et al. (1999), Rickenbach et al. (2002), and Li and Fu (2006). It may also be the result of similar underlying forcing mechanisms reflected in buoyancy and shear instability changes between westerly (easterly) and northerly (southerly) wind regimes.

b. Diurnal cycles of lightning, clouds, and precipitation between the two VI regimes

Diurnal variation offers important information about the forcing of convection. More specifically, by plotting diurnal cycles of parameters such as total lightning, rainfall, and IWP (Figs. 4a–c) we can isolate the response of both precipitation and convective structure to changes in diurnal forcing associated with each VI regime. Consider first the diurnal cycle of lightning flash density (Fig. 4a). To facilitate direct comparison to the PR data, these diurnal cycles were created by counting lightning flashes occurring only within the 215-km swath of the PR. The lightning diurnal cycles suggest that there is more (less) lightning activity present during the SVI (NVI) regime and that the diurnal cycle of lightning during the SVI is more (less) amplified by a factor of 4 at the diurnal cycle peak. The relative trends in the diurnal cycles of lightning for both regimes suggests that convection in the SVI regime experiences a more “explosive” growth phase in the late afternoon in terms of activity and/or intensity, exhibiting the classic diurnally forced continental convective signal (e.g., Christian et al. 2003; Rickenbach 2004). Note that lightning diurnal cycles computed from both the full field-of-view LIS (600 × 600 km), regardless of PR coverage, and continuous BLDN lightning datasets (not shown) yield nearly identical results.

Despite the factor of 3–4 increase in lightning flash rates observed during the SVI relative to the NVI (Fig. 4a), TRMM PR-diagnosed mean daily rainfall rates over the SCAMZ differ by only 20%–25% between the two regimes (cf. section 3c; NVI regime exhibits a higher mean daily rainfall). Analysis of the diurnal cycle of rainfall (Fig. 4b) reveals 1) a slightly broader diurnal cycle of total rain rate (convective + stratiform) during the NVI; 2) surprisingly similar diurnal cycles of convective rainfall in the SVI and NVI (compared to the lightning trends); and 3) a uniformly larger contribution to NVI total rain volume by stratiform rain rates (Fig. 4b), especially in the early morning. From a sampling perspective, accumulated NVI raining pixels (summed over all 160 days) outnumbered SVI raining pixels by approximately 50% and this difference was primarily due to increased stratiform rain coverage during the NVI. Both regime diurnal cycles suggest the presence of a “noon balloon” associated with the growth of convection, but the NVI diurnal cycle also seems to suggest the presence of a mesoscale convective system (MCS) mode in the nocturnal hours (Nesbitt and Zipser 2003; Rickenbach 2004). If we now compare the diurnal cycle of precipitation ice water path (Fig. 4c) to that of lightning (Fig. 4a) and rainfall (Fig. 4b), we find a much better correlation between regime differences in IWP and lightning relative to those of either rainfall and lightning or rainfall and IWP. The more highly correlated trends between IWP and lightning are a strong indication that deep convective vertical development (i.e., updrafts extending well above the 0°C level) in the SVI synoptic regime responds more strongly to diurnal forcing. Considering the relatively small difference between the SVI and NVI diurnal cycles of rainfall, but pronounced difference between the diurnal cycles of lightning and IWP, it is reasonable to conclude that the precipitation water budget, vertical structures of latent heating, and mass flux are also likely to be quite different between the two VI regimes (e.g., as in the case of zonal wind regimes; e.g., Petersen et al. 2002; Cifelli et al. 2002, 2004).

At the most basic level, the different diurnal patterns are forced in part by, and feed back to, differences in diurnal surface forcing between the two VI regimes; that is, surface sensible and latent heat fluxes, which are in turn dominated by surface solar fluxes, soil moisture, and variability in large-scale moisture convergence (WF02). The surface solar flux is modulated by the presence of clouds, and soil moisture by rainfall. Figure 5 shows the diurnal change in cloud cover for SVI and NVI regimes derived from the International Satellite Cloud Climatology Project (ISCCP) dataset. Comparing cloudiness between the SVI (Fig. 5a) and NVI (Fig. 5b) regimes, we find little difference in low-level cloudiness; however the coverage of midlevel to upper-level cloudiness in the NVI is ∼10%–20% greater in the SVI with the largest relative differences occurring between 0900 and 1500 UTC (0500–1000 local time). This difference in upper-level cloudiness suggests the potential for larger fluxes of surface sensible and latent heat from sunrise to noon in the SVI, more rapid convective boundary layer growth, and an attendant increase in thermodynamically driven early afternoon bursts in convection. The relatively reduced high and midlevel cloudiness in the SVI regime could be indicative of a drier free troposphere throughout most of the diurnal cycle except for local afternoon. A relatively drier troposphere in the SVI would also be consistent with the enhanced SVI column-integrated moisture flux divergence noted in this region by WF02. In comparison, more extensive high and midlevel cloudiness occurs at night and in the morning hours of the NVI regime, perhaps indicating the presence of a moister, more moist-neutral free troposphere (consistent with enhanced NVI moisture flux convergence; WF02). The greater total cloudiness of the NVI regime should result in weaker diurnal surface heating and, consequently, a reduction in conditional instability and atmospheric boundary layer height in the early afternoon compared to the SVI. In turn, this should result in a relative reduction in afternoon lightning activity. Last, the larger ice water path in Fig. 4 and lower fraction of high clouds in Fig. 5 seem to imply that the column-integrated heating occurring in the SVI regime may take place in more vertically developed cloud systems that cover a smaller horizontal area.

Fig. 5.

Mean diurnal cycles of low (solid) and mid- to high-level cloud fraction plus standard deviations (±− 1σ) for the period January–March 1998–2000 for (a) NVI and (b) SVI regimes over the SCAMZ

Fig. 5.

Mean diurnal cycles of low (solid) and mid- to high-level cloud fraction plus standard deviations (±− 1σ) for the period January–March 1998–2000 for (a) NVI and (b) SVI regimes over the SCAMZ

c. Convective structure and rainfall

Relative to the results discussed in sections 3a and 3b (Figs. 2 –5), we expect to see marked differences between VI-regime distributions of radar echo structure in the vertical. These differences are displayed in Figs. 6a–d by creating 3D-contoured relative frequency histograms of TRMM PR reflectivity as a function of height and convective type (convective or stratiform) for each regime (e.g., Yuter and Houze 1995). Comparing the two VI regimes in Figs. 6a–c, we find that the largest differences in precipitation echo intensity are manifested at heights above ∼5 km (temperature below 0°C) in the convective echo. That is, when PR reflectivities are detected above the freezing level in convection by the TRMM PR in the SVI regime (Fig. 6a), there is a greater probability (by a relative factor of 1.5–2.0) that they will exceed 30 dBZ relative to the NVI (Fig. 6c). Further, the decrease in reflectivity with height above the 5-km level is larger in the NVI then the SVI. These results are consistent with the aforementioned lightning and IWP trends (Figs. 4a–c) and suggest the presence of a more robust ice process in the convective echo occurring in the SVI regime. Though the regime diurnal cycles of precipitation (Fig. 4b) suggest that stratiform rainfall contributes a larger percentage of water mass to NVI total rainfall, vertical profiles of stratiform precipitation structure (Figs. 6b–d) in the NVI and SVI exhibit very little relative difference, beyond a slight relative enhancement in the median strength of the melting layer signature of the SVI profile.

Fig. 6.

Contoured relative frequency histograms of PR pixel reflectivity as a function of height for (a) SVI convective and (b) SVI stratiform echo; (c) NVI convective and (d) NVI stratiform echo. Reflectivity (dBZ) is indicated on the x axis, height (km) on the y axis. The relative frequency of occurrence (%) at each height and reflectivity is contoured using values of 0.1, 0.5, 1, 2, 4, 6, 10, 14, 18, 24, 30, 36, 42, and 48. The thin dotted line indicates the median of the distribution.

Fig. 6.

Contoured relative frequency histograms of PR pixel reflectivity as a function of height for (a) SVI convective and (b) SVI stratiform echo; (c) NVI convective and (d) NVI stratiform echo. Reflectivity (dBZ) is indicated on the x axis, height (km) on the y axis. The relative frequency of occurrence (%) at each height and reflectivity is contoured using values of 0.1, 0.5, 1, 2, 4, 6, 10, 14, 18, 24, 30, 36, 42, and 48. The thin dotted line indicates the median of the distribution.

The reflection of VI-regime convective structure differences as related to precipitation microphysics is further illustrated in spatial distributions of TRMM PR-diagnosed convective IWP for temperatures colder than −10°C (Figs. 7a–c). Though we focus on the convective IWP, correspondence between total IWP and lightning (not shown) was found to be nearly identical; this is because the convective IWP values dominate the much smaller stratiform contributions. As expected, based on Figs. 3, 4 and 6, there is a large decrease in the convective IWP when moving from SVI to NVI regimes (Figs. 7a,c, respectively). As in the case of the IWP diurnal cycle (Figs. 4a,c), note that the spatial trends in IWP are also well reflected in those of lightning (Figs. 3 and 7).

In contrast to the IWP trends, rainfall rates derived from the TRMM dataset for the subset of strongest NVI and SVI days reveal small, but identifiable, differences between the two regimes. During NVI (SVI) periods, the average rain rate over the SCAMZ taken from the near-surface rain product of the 2A25 PR algorithm (Iguchi et al. 2000) was 8 (6) mm day−1, consistent with the ECMWF estimates shown in Fig. 1. Approximately 50% (60%) of the NVI (SVI) PR near-surface rain volume was classified as convective. The strong differences in convective vertical development and lightning activity, but smaller differences in area-average rainfall point to a more active ice phase and perhaps more intense and spatially confined precipitation production in the SVI compared to the NVI (or more frequent precipitation that is similarly isolated, but weaker, in the NVI).

d. Environment

Causes for the aforementioned changes in convective structure and lightning associated with intraseasonal variations of monsoonal flow are investigated in this section. Because of the lower concentration of biomass burning aerosols present during the wet season, we focus on differences in thermodynamic and dynamic parameters over the SCAMZ as a function of VI. Two parameters known to exert a direct influence on convective organization, frequency, and intensity (here the word intensity refers to vertical development) include ambient thermodynamic buoyancy and tropospheric shear.

1) Thermodynamic environment

When the SVI regime prevails, air in SCAMZ tends to come from the east or southeast (WF02), originating over the drier Nordeste or southeastern regions of Brazil. During periods of NVI, a northerly or northwesterly fetch advects humid air into the region from the Amazon. The thermodynamic properties of the air in the SCAMZ are expected to reflect those of their source regions. Within the framework of ambient thermodynamics and the influence on cloud scales, we specifically consider differences in the vertical profiles of temperature and moisture as manifested in the amount of thermal buoyancy in each VI regime (e.g., CAPE and its structure). Though not in phase with the peak of diurnal heating, for the purposes of our analysis we focus on sounding data collected near 0000 UTC (∼1900 LT) as the 0000 UTC launches occur nearest to the diurnal cycle peak in lightning (and by proxy convective activity; Fig. 4) as compared to the early morning 1200 UTC launch.

Rather than simply focus on the mean profiles of temperature and moisture, it is more useful to consider differences in the temperature and moisture profiles as a function of pressure in each VI regime (Fig. 8). This approach will prove fruitful in the following discussions of regime buoyancy profiles. Figure 8 clearly illustrates the following two characteristics with regard to differences in the ambient thermodynamics of the two environments (beginning near the top of the mixed layer at 930 hPa): 1) the SVI mean sounding is warmer in the first 100 hPa of the troposphere and becomes approximately 0.5° cooler through the depth of the troposphere with the exception of a small layer at midlevels and near the upper troposphere where the SVI tropopause occurs at a lower level and 2) moisture differences in the two soundings are for the most part small in the lower and upper 100 hPa of the sounding, but the SVI sounding is systematically dryer by ∼0.5 g m−3 in the midlevels of the troposphere. These changes are expected based on the changes in origin of air arriving in the SCAMZ region. The resultant systematic thermodynamic differences imply steeper virtual temperature lapse rates in the mid- and lower troposphere of the SVI sounding, which in turn will influence the distribution of thermal buoyancy encountered for air parcels originating in the mixed layer.

Fig. 8.

Vertical profiles of absolute temperature difference (ΔT, bold black; SVIT − NVIT) and specific humidity difference (ΔQ, gray dash; SVIQ − NVIQ) between the SVI and NVI regimes. Differences plotted on abscissa in °C and g m−3; pressure plotted on ordinate.

Fig. 8.

Vertical profiles of absolute temperature difference (ΔT, bold black; SVIT − NVIT) and specific humidity difference (ΔQ, gray dash; SVIQ − NVIQ) between the SVI and NVI regimes. Differences plotted on abscissa in °C and g m−3; pressure plotted on ordinate.

To examine the thermal buoyancy differences, we chose to compute profiles of thermal buoyancy using a parcel originating at the 930-hPa level of the composited sounding. This layer was selected in order to normalize the statistics for varying altitudes of the sounding locations (several hundred meters), to account for any erroneous surface data in the soundings, and to simultaneously enable selection a parcel that still existed in the mixed layer of the transitioning convective boundary layer (CBL). Profiles of θ were used to verify that the 930-mb level occurred within or near the top of the mixed layer. Following Williams and Satori (2004) and Williams and Stanfill (2002), and within the framework of pseudoadiabatic parcel thermodynamics, we will also consider potential differences in cloud-base height as a possible influence on updraft width and perhaps precipitation and cloud electrical processes.

CFADs of thermal buoyancy as a function of height in the SVI and NVI regimes are shown in Figs. 9a,b, respectively. In general, the SVI regime is associated with more buoyant energy in the mean as reflected in the depth of the positive buoyancy trend. Furthermore, there is a tendency for larger positive buoyancy to occur in the lower troposphere of the SVI regime relative to the NVI regime (consistent with Fig. 8) in addition to a greater increase in the buoyancy with height below 4 km in the SVI. The greater low-level buoyancy in the SVI regime should help to offset updraft buoyancy losses due to entrainment and water loading (for equivalent updraft diameters). Thus, for similar values of integrated buoyant energy, greater low-level buoyancy should result in more intense convection (for the same shear profile; e.g., Lucas et al. 1994; McCaul and Cohen 2002). Mean relative humidities (RH) at 930 hPa exhibited values of approximately 78% (mixing ratio = 14.3 g kg−1, mean θ = 302.1 K) in the NVI regime and 75% (mean mixing ratio = 14.3 g kg−1, mean θ = 302.7 K) in the SVI regime. In the mean, the SVI convective boundary layer (CBL) is slightly warmer and drier than the NVI CBL (cf. Fig. 8). The θe of the parcel at 930 hPa in the SVI regime is 345.2 K, approximately 0.7° warmer than that of the NVI regime (344.5 K). The maximum θe in the SVI (NVI) layer below 930 hPa was 348.3 (347.4) K. Relative to cloud-base heights, the 3% increase in NVI RH equates to a relatively small mean difference in base height relative to the SVI (<100 m: Williams and Satori 2004). The absolute maximum regime difference in RH at any level below 930 hPa was ∼14%, suggesting a maximum difference in mean cloud-base height on the order of 300 m. Note that these cloud-base height differences are consistent with those measured by tethersonde ascents during TRMM–LBA by Betts et al. (2002). Importantly, the approximately exponential relationship between cloud-base height and lightning flash rate presented in Williams and Stanfill (2002) and Williams and Satori (2004) suggests that the VI regime difference in cloud-base height inferred herein could account for ∼25% of the increase in SVI lightning flash rate relative to the NVI regime. On the other hand, the inferred change in FD based on cloud-base height alone is still considerably smaller than the observed 100%–200% increases in lightning activity observed during SVI periods (Figs. 3 –4).

As previously noted, the NVI-regime virtual temperature lapse rates for height levels above the 930 parcel level are not as steep as those of the SVI, resulting in the disparity in both negative and positive sides of the buoyancy distributions shown in Fig. 9. Collectively, the aforementioned thermodynamic features, but most importantly the relative differences in VI environment parcel instability, are consistent with the presence of more vigorous, vertically developed convection, possessing a more vigorous ice phase and more lightning during the SVI regime.

2) Wind profiles and shear distribution

Changes in wind shear and buoyancy work in tandem to influence both convective organization and vertical development (cf. Houze 1993; McCaul and Cohen 2002). Following this line of reasoning, Halverson et al. (2002) correlated shifts in ambient wind shear and convective structure to changes in the SWAMZ zonal wind regime during TRMM–LBA. Since basinwide circulation changes in the troposphere over the Amazon are accompanied by changes in near-equatorial VI (e.g., Figs. 1a,b), it is reasonable to expect changes in VI to be associated and/or correlated to changes in wind shear over the SCAMZ.

Analysis of SCAMZ sounding winds (Fig. 10) indicates a marked difference in low to midlevel (e.g., 930–300 hPa) wind structure with similar upper-tropospheric winds. There is a pronounced shift in the zonal wind below 700 hPa [similar to that noted for easterly–westerly regimes in the SWAMZ by Cifelli et al. (2002), Petersen et al. (2002), and Halverson et al. (2002)]. The SVI-regime mean wind profile is associated with predominantly easterly flow throughout the depth of the troposphere, with the suggestion of a very slight local maximum in wind speed between 700 and 750 hPa. It seems probable that the air mass originated over the drier and hotter Nordeste region. Given the higher altitude of the Nordeste, it is also likely that further subsidence-driven drying and heating of the air mass would have occurred as it approached the SCAMZ, resulting in the observed lower humidity and warmer temperatures shown in Fig. 8. This warmer, drier profile is also similar to that observed during easterly wind regimes over the SWAMZ. The easterly wind, combined with the slightly drier troposphere encountered in SVI regimes, would support the propagation of coastal squall lines and associated mesoscale disturbances well inland (Sun and Orlanski 1981; Garstang et al. 1994; Cohen et al. 1995; Halverson et al. 2002; Rickenbach 2004). The SVI meridional wind component shifts from northerly to southerly near the 600-hPa level. The NVI regime exhibits northwesterly flow that increases in magnitude from the surface to approximately 900 hPa, and then decreases from near the surface to approximately 600 hPa. A shift to southeasterly flow occurs above the 600-hPa level.

Fig. 10.

Mean profiles of u (dash) and υ (solid) winds for the NVI (bold) and SVI (light) regimes. Speed is plotted on the abscissa, pressure on the ordinate.

Fig. 10.

Mean profiles of u (dash) and υ (solid) winds for the NVI (bold) and SVI (light) regimes. Speed is plotted on the abscissa, pressure on the ordinate.

Figures 11a,b show the difference between lower-tropospheric wind shear (as defined in section 2) in the two VI regimes. The SVI low-level shear distribution (Fig. 11a) exhibits a relatively broad easterly spread in direction, but a relatively narrow peak in magnitude of ∼4–8 m s−1. Conversely, shear in the NVI regime (Fig. 11a) exhibits a broad speed distribution with a similar magnitude 4–8 m s−1, but a more narrowly distributed directional shear that is west-southwesterly in its orientation. Mean hodographs of flow in both regimes (not shown) are only weakly curved at very low levels and approximately linear thereafter.

Deep tropospheric shear in both regimes (Fig. 11b) is dominated by relatively strong southeasterly flow aloft, leading to very similar southeasterly shear distributions in both instances. There is a tendency for the deep tropospheric shear magnitude to be larger in the NVI regime, however. The combination of reversed low and deep tropospheric shear vectors during NVI periods suggests that NVI convective events should be associated with east-northeastward-propagating convective systems that tilt westward with height, and perhaps extensive regions of trailing precipitation sheared rearward in a convective-system-relative sense (the relative increase in humidity present in the troposphere of NVI regimes would also support/sustain trailing regions of precipitation). Broadly speaking, SVI convective systems should propagate westward and, to the extent that the CAPE is larger and system propagation is in the direction of the deep tropospheric shear, maintain a more erect nature similar to that observed in easterly zonal wind regimes (e.g., Cifelli et al. 2002; Halverson et al. 2002).

4. Conclusions and implications

Intraseasonal variability in precipitation during the South American monsoon has been related to transitions in both the low-level zonal and meridional wind (Petersen et al. 2002; WF02, respectively). Previous work in the literature addressed oscillating low-level zonal wind regimes over the southern Amazon (Cifelli et al. 2002, 2004; Williams et al. 2002; Halverson et al. 2002; Petersen et al. 2002). These studies noted systematic changes in convective vertical structure, lightning frequency, and precipitation microphysical characteristics accompanying reversals in low-level zonal wind direction and/or zonal wind anomalies. The changes in zonal wind anomaly appear to be forced by the equatorward intrusion of extratropical synoptic systems (Rickenbach et al. 2002). However, WF02 found that precipitation variability over the Amazon Basin and subtropical South America was more tightly coupled to low-level meridional wind variability in the near-equatorial Amazon. WF02 developed the South American monsoon V index (VI) as a metric for detecting this variability.

Motivated by the question of convective vertical structure variability as a function of the VI, this study composited TRMM PR and LIS datasets to describe ensemble convective vertical structure profiles and lightning frequency in periods of north and south VI. Our results revealed substantial changes in convective vertical structure and microphysical processes as a function of VI. Initial comparisons of precipitation and lightning flash density indicated the strongest response to VI changes over the south-central Amazon; hence the southern Amazon became the focus of this study. The TRMM observations were subsequently placed in the context of background thermodynamic and wind environments using sounding data collected from three different locations within the south-central Amazon. Continuous ground-based observations of lightning from the Brazilian Lightning Detection Network (BLDN), located in the southwestern Amazon, were also used as a quick and direct proxy for testing convective system vertical structure variability as a function of transitions in VI.

Analysis of the observations revealed the following:

  1. Lightning flash rates respond strongly to changes in the monsoon VI in both time series and composited mean analyses. Southerly VI regimes are associated with more than twice the lightning flash density of the northerly VI regime, but average precipitation amounts are 20%–25% lower during periods of SVI. Consistent with the marked differences in lightning activity, the TRMM PR data show that convection in the SVI regime is 1.5–2 times more likely to produce reflectivities in excess of 30 dBZ above the freezing level and that SVI convection is associated with significantly larger precipitation ice water path. These consistent changes reinforce hypotheses that mixed phase processes play a greater role in precipitation production during periods of SVI. The deeper penetration of convection and lower area-mean rainfall indicate the occurrence of more intense but spatially confined precipitation during the SVI. Alternatively, in NVI, the lightning flash density, TRMM PR reflectivity, and ice water path above the freezing level are all consistently lower. The 20%–25% higher area-averaged rainfall is mainly contributed by a higher nocturnal stratiform rainfall. In general, NVI rainfall is less intense but is horizontally more widespread and/or more frequent.

  2. The diurnal cycle of lightning flash rate was strongly modulated by the phase of the VI. For southerly VI phases, the diurnal cycle was significantly more amplified in the late afternoon/early evening hours, exhibiting an archetypical continental shape. For the northerly phase of the VI, the diurnal cycle of lightning still exhibited an afternoon peak, but was much weaker in amplitude (diurnal peak to trough) by a factor of 4 to 5 relative to the southerly phase. The diurnal cycles of rainfall in each VI regime also exhibited strong diurnal modulation, but in contrast to the diurnal cycle of lightning, relative differences in amplitude were considerably smaller and driven primarily by a preference for more stratiform precipitation during northerly VI regimes. Increases in stratiform precipitation during periods of northerly VI were also reflected in diurnal cycles of mid- to upper level cloudiness. The much larger difference between the lightning diurnal cycles was better correlated to a similar diurnal response in precipitation ice water path. The results point to a more robust contribution of ice processes to rainfall during periods of southerly VI in the late afternoon hours. When considered in light of the small relative differences in rainfall, the diurnal cycle results also suggest that significant differences in latent heating and mass flux profiles must exist between the two regimes.

  3. Changes in the large-scale environment at 0000 UTC appeared consistent with the observed changes in convective structure and lightning frequency. During periods of southerly VI, the lower (middle–upper) troposphere was warmer (cooler) than that of the northerly. Differences in moisture content appeared to be most prevalent in the lower–middle troposphere where periods of southerly VI were on average approximately 0.5 g m−3 drier than northerly periods. The temperature and moisture differences resulted in larger virtual temperature lapse rates, a tendency for larger integrated thermal buoyancy, and a steeper slope in the low- to mid-level thermal buoyancy profile during periods of southerly VI. The cooler and slightly more moist conditions near the surface present during northerly VI periods should also result in relatively lower cloud-base heights at the peak of the diurnal cycle [O(200–300 m)].

  4. Low-level tropospheric shear changed markedly between the two regimes. The SVI low-level shear distribution (Fig. 11a) was broad in direction, exhibiting an easterly mean but a rather narrow and modest peak in magnitude of ∼4–8 m s−1. Conversely, shear in the NVI regime exhibited a broader speed distribution with a mean similar to or slightly larger than the SVI regime, but a narrower distribution of directions centered on a west-southwesterly orientation. Both VI regimes exhibited a very similar southeasterly deep tropospheric shear direction but with the NVI regime exhibiting a tendency for slightly larger values than the SVI. We hypothesize that the reversal in direction between the low and deep tropospheric shear vectors of the NVI regime, coupled with a moister troposphere, would lead to more tilted convection and more widespread precipitation events.

In conclusion, the data suggest that vertical structure properties of convection over the southern Amazon respond strongly to modulation of the large-scale intraseasonal monsoon circulation as measured by the magnitude of the V index. The monsoonal variability in atmospheric thermodynamic and dynamic structure provides a background forcing consistent with the observed transitions in convective precipitation microphysics, which, in turn, are manifested in marked differences in lightning activity and precipitation ice water path. Though the observed changes in dynamic and thermodynamic background forcing discussed herein provide the most straightforward explanation for the observed changes in monsoon convective structure over the southern Amazon, variability in background concentrations of biomass burning aerosols may also exert a microphysical influence (e.g., Rosenfeld and Lensky 1998; Williams et al. 2002; Andrea et al. 2004). Relative to climate study applications the degree to which either or both types of forcing alter convective precipitation structure still needs to be resolved by future research.

Following the marked changes in monsoon-regime convective structure and associated rainfall properties, one important implication is that convective updraft profiles, and hence mass flux and diabatic heating, are also likely to change in concert with the VI. In turn, VI-regime changes in the heating profile will feed back to the large-scale circulation (e.g., Hartmann et al. 1984; Rind and Rossow 1984; Schumacher et al. 2004). Future research should address/define the changes in mass flux profile using tools such as cloud-resolving models.

Last but not least, the apparent response of lightning activity to intraseasonal modulation of the monsoon is pronounced and strongly reflects underlying changes in convective structure. Observations of this type have been noted numerous times in the literature for other monsoon-affected regions of the globe (e.g., northern Australia, India, the southwestern United States), pointing to the utility of lightning activity for monitoring intraseasonal variability in local climate.

Acknowledgments

This research was supported by grants from the NASA EOS Program (NNM05AA22A), the NASA Precipitation Measurement Mission (NNM05AA22A), and the NASA Global Water and Energy Cycle program. The authors thank Dr. Hui Wang and Ms. Allison Walker for their contributions of Figs. 1 and 5. We also thank two anonymous reviewers for their insights and helpful comments.

REFERENCES

REFERENCES
Andrea
,
M. O.
,
D.
Rosenfeld
,
P.
Artaxo
,
A. A. P.
Costa
,
G. P.
Frank
,
K.
Longo
, and
M. A. F.
Silva-Dias
,
2004
:
Smoking rain clouds over the Amazon.
Science
,
303
,
1337
1342
.
Betts
,
A. K.
,
J. D.
Fuentes
,
M.
Garstang
, and
J. H.
Hall
,
2002
:
Surface diurnal boundary layer structure over Rondônia during the rainy season.
J. Geophys. Res.
,
107
.
doi:10.1029/2001JD000356
.
Black
,
R. A.
,
1990
:
Radar reflectivity–ice water content relationships for use above the melting level in hurricanes.
J. Appl. Meteor.
,
29
,
955
961
.
Blakeslee
,
R. J.
,
J. C.
Bailey
,
O.
Pinto
,
A.
Athayde
,
N.
Renno
, and
C. D.
Weidman
,
2003
:
The Rondonia lightning detection network: Network description, science objectives, data processing/archival methodology and results. Proc. 12th Int. Conf. on Atmospheric Electricity, Versailles, France, International Commission on Atmospheric Electricity, 77–80
.
Christian
,
H. J.
, and
Coauthors
,
2003
:
Global frequency and distribution of lightning as observed from space by the Optical Transient Detector.
J. Geophys. Res.
,
108
.
doi:10.1029/2002JD002347
.
Cifelli
,
R. C.
,
W. A.
Petersen
,
L. D.
Carey
, and
S. A.
Rutledge
,
2002
:
Radar observations of the kinematic, microphysical, and precipitation characteristics of two MCSs in TRMM-LBA.
J. Geophys. Res.
,
107
.
8077, doi:10.1029/2000JD000264
.
Cifelli
,
R. C.
,
L. D.
Carey
,
W. A.
Petersen
, and
S. A.
Rutledge
,
2004
:
An ensemble study of wet season convection in southwest Amazonia: Kinematics and implications for diabatic heating.
J. Climate
,
17
,
4692
4707
.
Cohen
,
J. C. P.
,
M. A. F.
Silva Dias
, and
C. A.
Nobre
,
1995
:
Environmental conditions associated with Amazonian squall lines: A case study.
Mon. Wea. Rev.
,
123
,
3163
3174
.
Garreaud
,
R. D.
, and
J. M.
Wallace
,
1998
:
Summertime incursions of midlatitude air into subtropical and tropical South America.
Mon. Wea. Rev.
,
126
,
2713
2733
.
Garstang
,
M.
,
H.
Massie
Jr.
,
J.
Halverson
,
S.
Greco
, and
J.
Scala
,
1994
:
Amazon coastal squall lines. Part I: Structure and kinematics.
Mon. Wea. Rev.
,
122
,
608
622
.
Halverson
,
J. B.
,
T.
Rickenbach
,
B.
Roy
,
H.
Pierce
, and
E.
Williams
,
2002
:
Environmental characteristics of convective systems during TRMM-LBA.
Mon. Wea. Rev.
,
130
,
1493
1509
.
Hartmann
,
D. L.
,
H. H.
Hendon
, and
R. A.
Houze
Jr.
,
1984
:
Some implications of the mesoscale circulations in tropical cloud clusters for large-scale dynamics and climate.
J. Atmos. Sci.
,
41
,
113
121
.
Houze
Jr,
R. A.
,
1993
:
Cloud Dynamics.
Academic Press, 573 pp
.
Iguchi
,
T.
,
T.
Kozu
,
R.
Meneghini
,
J.
Awaka
, and
K.
Okamoto
,
2000
:
Rain-profiling algorithm for the TRMM precipitation radar.
J. Appl. Meteor.
,
39
,
2038
2052
.
Jones
,
C.
, and
L. M. V.
Carvalho
,
2002
:
Active and break phases in the South American monsoon system.
J. Climate
,
15
,
905
914
.
Lau
,
K-H.
, and
N-C.
Lau
,
1990
:
Observed structure and propagation characteristics of tropical summertime synoptic scale disturbances.
Mon. Wea. Rev.
,
118
,
1888
1913
.
Li
,
W.
, and
R.
Fu
,
2006
:
Influence of cold air intrusions on the wet season onset over Amazonia.
J. Climate
,
19
,
257
275
.
Liebmann
,
B.
,
G. N.
Kiladis
,
J. A.
Marengo
,
T.
Ambrizzi
, and
J. D.
Glick
,
1999
:
Submonthly convective variability over South American and the South Atlantic convergence zone.
J. Climate
,
12
,
1877
1891
.
Lin
,
X.
,
D. A.
Randall
, and
L. D.
Fowler
,
2000
:
Diurnal variability of the hydrologic cycle and radiative fluxes: Comparisons between observations and a GCM.
J. Climate.
,
13
,
4159
4179
.
Lu
,
E.
, and
J. C. L.
Chan
,
1999
:
A unified monsoon index for south China.
J. Climate
,
12
,
2375
2385
.
Lucas
,
C.
,
E. J.
Zipser
, and
M. A.
Lemone
,
1994
:
Vertical velocity in oceanic convection off tropical Australia.
J. Atmos. Sci.
,
51
,
3183
3193
.
MacGorman
,
D. R.
, and
W. D.
Rust
,
1999
:
The Electrical Nature of Storms.
Oxford University Press, 422 pp
.
Madden
,
R. A.
, and
P. R.
Julian
,
1971
:
Description of a 40-50 day oscillation in the zonal wind in the tropical Pacific.
J. Atmos. Sci.
,
28
,
702
708
.
Mapes
,
B.
, and
R. A.
Houze
Jr.
,
1993
:
An integrated view of the 1987 Australian monsoon and its mesoscale convective systems. II: Vertical structure.
Quart. J. Roy. Meteor. Soc.
,
119
,
733
754
.
McCaul
Jr.,
E. W.
, and
C.
Cohen
,
2002
:
The impact on simulated storm structure and intensity of variations in the mixed layer and moist layer depths.
Mon. Wea. Rev.
,
130
,
1722
1748
.
Negri
,
A. J.
,
T. L.
Bell
, and
L.
Xu
,
2002
:
Sampling of the diurnal cycle of precipitation using TRMM.
J. Atmos. Oceanic Technol.
,
19
,
1333
1344
.
Nesbitt
,
S. W.
, and
E. J.
Zipser
,
2003
:
The diurnal cycle of rainfall and convective intensity according to three years of TRMM measurements.
J. Climate
,
16
,
1456
1475
.
Nogues-Paegle
,
J.
, and
K. C.
Mo
,
1997
:
Alternating wet and dry conditions over South America during summer.
Mon. Wea. Rev.
,
125
,
279
291
.
Petersen
,
W. A.
, and
S. A.
Rutledge
,
2001
:
Regional variability in tropical convection: Observations from TRMM.
J. Climate
,
14
,
3566
3586
.
Petersen
,
W. A.
,
S. W.
Nesbitt
,
R. J.
Blakeslee
,
R.
Cifelli
,
P.
Hein
, and
S. A.
Rutledge
,
2002
:
TRMM observations of intraseasonal variability in convective regimes over the Amazon.
J. Climate
,
15
,
1278
1294
.
Petersen
,
W. A.
,
R.
Cifelli
,
D. J.
Boccippio
,
S. A.
Rutledge
, and
C.
Fairall
,
2003
:
Convection and easterly wave structure observed in the eastern Pacific warm pool during EPIC-2001.
J. Atmos. Sci.
,
60
,
1754
1773
.
Pickering
,
K. E.
,
A. M.
Thompson
,
W-K.
Tao
,
R. B.
Rood
,
D. P.
McNamara
, and
A. M.
Molod
,
1995
:
Vertical transport by convective clouds: Comparisons of three modeling approaches.
Geophys. Res. Lett.
,
22
,
1089
1092
.
Price
,
C.
,
2000
:
Evidence for a link between global lightning activity and upper tropospheric water vapor.
Nature
,
406
,
290
293
.
Rickenbach
,
T. M.
,
2004
:
Nocturnal cloud systems and the diurnal variation of clouds and rainfall in southwestern Amazonia.
Mon. Wea. Rev.
,
132
,
1201
1219
.
Rickenbach
,
T. M.
,
R. N.
Ferreira
,
J.
Halverson
, and
M. A. F.
Silva Dias
,
2002
:
Modulation of convection in the southwestern Amazon basin by extratropical stationary fronts.
J. Geophys. Res.
,
107
.
8040, doi:10.1029/2000JD000263
.
Riehl
,
H.
,
1954
:
Tropical Meteorology.
McGraw Hill, 392 pp
.
Rind
,
D.
, and
W. B.
Rossow
,
1984
:
The effects of physical processes on the Hadley Circulation.
J. Atmos. Sci.
,
41
,
479
507
.
Rosenfeld
,
D.
, and
M.
Lensky
,
1998
:
Satellite-based insights into precipitation formation processes in continental and maritime convective clouds.
Bull. Amer. Meteor. Soc.
,
79
,
2457
2476
.
Rutledge
,
S. A.
,
E. R.
Williams
, and
T. D.
Keenan
,
1992
:
The Down Under Doppler and Electricity Experiment (DUNDEE): Overview and preliminary results.
Bull. Amer. Meteor. Soc.
,
73
,
3
16
.
Scala
,
J. R.
, and
Coauthors
,
1990
:
Cloud draft structure and trace gas transport.
J. Geophys. Res.
,
95
,
17015
17030
.
Schumacher
,
C.
,
R. A.
Houze
Jr.
, and
I.
Kraucunas
,
2004
:
The tropical dynamical response to latent heating estimates derived from the TRMM Precipitation Radar.
J. Atmos. Sci.
,
61
,
1341
1358
.
Silva Dias
,
M.
, and
Coauthors
,
2002
:
Clouds and rain processes in a biosphere atmosphere interaction context in the Amazon region.
J. Geophys. Res.
,
107
.
8072, doi:10.1029/2001JD000335
.
Sun
,
D-Z.
, and
R. S.
Lindzen
,
1993
:
Distribution of tropical tropospheric water vapor.
J. Atmos. Sci.
,
50
,
1643
1660
.
Sun
,
W-Y.
, and
I.
Orlanski
,
1981
:
Large mesoscale convection and sea breeze circulation: Part I: Linear stability analysis.
J. Atmos. Sci.
,
38
,
1675
1693
.
Tao
,
W-K.
,
J.
Simpson
,
S.
Lang
,
M.
McCumber
,
R.
Adler
, and
R.
Penc
,
1990
:
An algorithm to estimate the heating budget from vertical hydrometeor profiles.
J. Appl. Meteor.
,
29
,
1232
1244
.
Wang
,
H.
, and
R.
Fu
,
2002
:
Cross-equatorial flow and seasonal cycle of precipitation over South America.
J. Climate
,
15
,
1591
1608
.
Webster
,
P.
,
V.
Magaña
,
T. N.
Palmer
,
J.
Shukla
,
R. A.
Tomas
,
M.
Yanai
, and
T.
Yasunari
,
1998
:
Monsoons: Processes, predictability and the prospects for prediction.
J. Geophys. Res.
,
103
,
14451
14510
.
Williams
,
E. R.
, and
S.
Stanfill
,
2002
:
The physical origin of the land–ocean contrast in lightning activity.
Compt. Rend. Phys.
,
3
,
1277
1292
.
Williams
,
E. R.
, and
G.
Satori
,
2004
:
Lightning, thermodynamic and hydrological comparison of the two tropical continental chimneys.
J. Atmos. Solar-Terr. Phys.
,
66
,
1213
1231
.
Williams
,
E. R.
,
S. A.
Rutledge
,
S. C.
Geotis
,
N.
Renno
,
E.
Rasmussen
, and
T.
Rickenbach
,
1992
:
A radar and electrical study of tropical hot towers.
J. Atmos. Sci.
,
49
,
1386
1395
.
Williams
,
E. R.
, and
Coauthors
,
2002
:
Contrasting convective regimes over the Amazon: Implications for cloud electrification.
J. Geophys. Res.
,
107
.
8082, doi:10.1029/2001JD000380
.
Yuter
,
S. E.
, and
R. A.
Houze
,
1995
:
Three-dimensional kinematic and microphysical evolution of Florida cumulonimbus. Part II: Frequency distributions of vertical velocity, reflectivity, and differential reflectivity.
Mon. Wea. Rev.
,
123
,
1941
1963
.

Footnotes

Corresponding author address: Dr. Walter A. Petersen, ESSC/NSSTC, University of Alabama in Huntsville, Huntsville, AL 35899. Email: walt.petersen@msfc.nasa.gov