This paper examines the origins of a secondary nocturnal maximum in cloudiness and precipitation in southwestern Amazonia, a diurnal feature observed previously by many investigators. Analysis is based on satellite, radar, sounding, and profiler observations of precipitating systems and cloudiness from the Tropical Rainfall Measuring Mission Large-Scale Biosphere–Atmosphere (TRMM-LBA) and the coincident Wet-Season Atmospheric Mesoscale Campaign (WETAMC) field programs during the early 1999 wet season. The general finding is that following the collapse of the nearly ubiquitous and locally generated afternoon (“noon balloon”) convection, organized deep convection contributes to a postmidnight maximum in raining area and high cloudiness, and to a lesser extent rainfall. Nocturnal convective systems have the effect of weakening and delaying the onset of the following afternoon's convection. Many of these nocturnal convective events are traced to large- scale squall lines, which propagate westward thousands of kilometers from their point of origin along the northeast coast of Brazil. In addition, a previously undescribed nocturnal stratiform drizzle phenomenon, generated above the melting layer independently from deep convection, contributes significantly to nocturnal cloud cover. Results from this study underscore the complex influence of propagating large-scale organized convection in locally modulating the diurnal variation in clouds and rain. The greatest significance of the nocturnal drizzle may be the potential effect on the diurnal radiation budget by the extensive midlevel nocturnal clouds rather than their marginal contribution to nocturnal rainfall.
The diurnal variation of clouds and precipitation over tropical landmasses is the net expression of a rich variety of phenomena spanning many time and space scales. Solar heating of the land surface increases sensible heat flux and thermodynamic instability, leading to vertical mixing of water vapor in the deepening boundary layer during the morning and favoring afternoon precipitating convection (Wallace 1975; Betts 1976). Dynamic processes, such as land/water circulations, density currents, and orographic lifting, are crucial to actually initiating moist convection by lifting parcels to their level of positive buoyancy (Wallace 1975). Preexisting convective systems may prime the atmosphere for further convection by lowering convective inhibition due to the net upward displacement resulting from the atmosphere's response to heating from nearby convective systems. Thus convective systems increase the chance that surface heating and local circulations will succeed at initiating more convection (Mapes 1993). Once formed, subsequent upscale growth and mesoscale organization strongly influences the timing of maximum rainfall and cloudiness (McAnelly and Cotton 1989).
Over the last decade, advances in satellite remote sensing and increased deployment of field instrumentation have led to a detailed view of diurnal changes in rainfall and cloudiness. Satellite studies and surface observations have provided guidance for model parameterization of convective processes (Lin et al. 2000; Yang and Slingo 2001; Betts and Jakob 2002). There has been a recent focus on convective cloud systems in the Amazon basin, in part because of field efforts to study the interaction between convection, land processes, topography, and large-scale circulations (Baidya Roy and Avissar 2002; Rickenbach et al. 2002; Laurent et al. 2002). The Tropical Rainfall Measuring Mission Large-Scale Biosphere–Atmosphere (TRMM-LBA) and the coincident Wet-Season Atmospheric Mesoscale Campaign (WETAMC) field programs during the 1999 wet season (Silva Dias et al. 2002) provided a detailed view of the structure and environment of convective systems in southwestern Amazonia.
Many studies of the diurnal cycle of precipitation and cloudiness over the Amazon indicated a secondary nocturnal maximum in addition to the primary afternoon maximum. A composite of two wet seasons of TRMM rain estimates from the precipitation radar (PR) over the Amazon (Lin et al. 2000) indicated that rainfall increased at 2300 LT to a broad nocturnal maximum, though weaker than the afternoon maximum following the explosive increase between 1100 LT and local noon. Lin et al. (2000) postulated that the weaker nocturnal maximum was associated with stratiform rain enhanced through instability driven by nocturnal radiative cooling of the extensive anvil cloud tops. Negri et al. (2002) found a similar nocturnal maximum in rainfall using a TRMM-adjusted geostationary infrared rainfall estimate applied to the Amazon basin. They hypothesized that the nocturnal peak was associated with an enhancement of rainfall along the wide Amazon River via a land/river circulation favoring nocturnal convection over the river. Nesbitt and Zipser (2003) used TRMM PR and passive microwave measurements to partition rainfall into contributions from isolated convection with little ice-scattering signature and mesoscale convective systems (MCSs) with extensive thick anvil (ice) cloud. In the composite of all tropical land regions, isolated convection with and without ice aloft had peak rainfall in the afternoon, while rainfall from MCSs was greatest around midnight, presumably from the upscale growth of afternoon convection.
A further complication to this picture was that the phase of the diurnal cycle of clouds and rainfall varied across Amazonia. Satellite retrievals of the diurnal cycle of precipitation and cloudiness over the Amazon basin revealed an alternating banded structure in the diurnal phase, from the northeast coast to western Amazonia. Negri et al. (1994) used passive microwave measurements to construct day versus evening precipitation composites for three wet seasons. They found that the prevalence of afternoon rainfall was not uniform over the basin, but instead was focused in parallel bands along the entire northeast coast of Brazil, central Amazonia, and western Amazonia. Wet-season composites of the diurnal phase from infrared cloud-top temperature data showed a similar pattern (Garreaud and Wallace 1997; Yang and Slingo 2001). Garreaud and Wallace (1997) posited that the banded structure in diurnal phase might be influenced by gravity wave propagation from regions of strong diurnal modulation on both sides of the basin (orographic convection in the Andes to the west, and afternoon sea-breeze convection along the northeast coast to the east).
Other investigations further suggested that the phase propagation of the diurnal rainfall maximum in Amazonia was tied to the effects of long-lived, propagating convective systems. Greco et al. (1990) found that in central Amazonia 1200 km from the northeast coast, convective systems that formed locally produced a rainfall maximum around 1800 LT, while rainfall associated with propagating squall-line systems that formed along the northeast coast produced a maximum between 1000 and 1400 LT. They hypothesized that the late morning maximum corresponded to the time of arrival of the coastal squall lines to central Amazonia. Cutrim et al. (2000) analyzed a different year of rain gauge data to show that a station in central Amazonia had a maximum near local noon, much earlier than the 1600 LT maximum at a station on the northeast coast, possibly tied to the timing of coastal squall-line passage as described in Greco et al. (1990). A satellite climatology of these large-scale westward propagating coastal systems presented by Cohen et al. (1995) showed that though they were more frequent (or more easily identified) during the dry season, they occurred all year round and could be tracked at times more than 2000 km inland (westward). The results of Garreaud and Wallace (1997) suggested that as large-scale coastal squall lines propagate westward to the basin interior, the diurnal maximum in cloudiness follows them in space and time, both during the wet and dry seasons. Similar diurnal rainfall phase propagation occurred across West Africa for related reasons (Shinoda et al. 1999; Fortune 1980).
Recent results from the TRMM-LBA campaign in southwestern Amazonia provided additional insight. Diurnal composites of radar-derived rain area and conditional rainfall rate were produced for periods of northwesterly low-level flow in Rondônia State, Brazil [when the South Atlantic convergence zone (SACZ) penetrated the deep Tropics] and were compared with the interim period (Rickenbach et al. 2002). In addition to the afternoon maxima in rainfall and rain area, they found for both regimes another large maximum in rain area after local midnight following a minimum near 2300 LT. The presence of this minimum suggested nocturnal resurgence of light rainfall, well after the decay of the afternoon systems, though the reasons for this were not clear. Machado et al. (2002) also noted a similar secondary nocturnal peak in radar rain area and total cloud cover during the TRMM-LBA campaign.
These studies highlighted several unresolved questions concerning the nocturnal maximum in clouds and precipitation over tropical South America. What was the origin of the secondary nocturnal maximum in rainfall and cloud area? How important were stratiform precipitation processes in the nocturnal maximum of clouds and rainfall? Did propagating coastal squall lines influence nocturnal rainfall in southwestern Amazonia? The goal of this paper is to investigate nocturnal precipitation in Amazonia with those specific questions in mind. This study takes the approach of composite and case study analysis of TRMM-LBA surface observations and geostationary satellite data for nearly 2 months in southwestern Amazonia during the early 1999 wet season. Results from this study highlight the complex influence of propagating large-scale organized convection in locally modulating the diurnal variation in clouds and rain. These findings may help to guide the interpretation of multiyear diurnal rainfall and cloudiness composites from satellite platforms such as TRMM and may lead to improvements in the representation of precipitation processes in model parameterizations of convection.
The TRMM-LBA/WETAMC field campaign was conducted in Rondônia in southwestern Amazonia during January and February 1999. The main objectives of these joint experiments were to examine biosphere–atmosphere coupling and to provide validation for TRMM algorithms in a tropical continental environment (for details see Silva Dias et al. 2002). In this study, volumes of radar reflectivity data, taken approximately every 10 min from the National Aeronautics and Space Administration (NASA)C-band scanning radar [known as the Tropical Ocean Global Atmosphere (TOGA) radar], were used to examine the three-dimensional structure of convection between 11 January and 28 February. Radar reflectivity data were quality controlled to remove anomalous echo, and calibration biases were corrected by comparison with the TRMM precipitation radar (Rickenbach et al. 2002; Anagnostou and Morales 2002). The data were interpolated to a 1 km × 1 km × 1 km grid for analysis. Spatial gradients and local maxima provided the basis to partition the reflectivity field into regions of active convective cells and weaker, horizontally uniform stratiform rain (see Rickenbach et al. 2002 for details). Analysis was restricted to within a range of 130 km from the radar since beam broadening and gaps between radar scans degraded the usefulness of the data beyond that range. Reflectivity data at a height of 2 km above ground level (AGL) were converted to rainfall rate using an empirical relationship based on surface disdrometer data (Tokay et al. 2002). Only relative differences rather than absolute values of rainfall rate are emphasized in this analysis. Rain area was based on the areal coverage of reflectivity greater than 10 dBZ (about 0.5 mm h−1, or roughly speaking, drizzle).
Infrared brightness temperature data for the Amazon basin (every 30 min, nominally 5-km spacing) from the Geostationary Operational Environmental Satellite-8 (GOES-8) revealed the height, organization, and motion of cloud tops associated with convective systems and other cloud features. These data were obtained from the NASA Goddard Space Flight Center's Distributed Active Archive Center (GSFC DAAC) and navigated to earth coordinates in order to produce imagery and regional composites over different areas of interest. Composites were made within the radar coverage area and within a 10° diameter circle covering western and central Amazonia inclusive of the radar area. Sounding data from the Abracos site (collocated with the radar) and the Rancho Grande site (at Ariquemes, 70 km northwest of the radar) provided thermodynamic and wind profiles for two case studies presented in this work [see Halverson et al. (2002) for details on the radiosonde network in the TRMM-LBA campaign]. Figure 1 shows the location of the observational network and analysis regions used in this study.
We begin by presenting general features of the diurnal cycle of rainfall and cloudiness, from 11 January 1999 to 28 February 1999 (49 days). Diurnal composites of conditional rainfall rate (rainfall within the raining area), rain area, and vertical reflectivity structure (Fig. 2) were constructed from the radar reflectivity volumes and compared to composites of cloudiness from geostationary satellite. Rain area, rainfall rate, and mean echo-top height were separated into convective and stratiform components. Included in the radar parameters was the mean height of the 30-dBZ surface, which is an indicator of the vertical intensity of convective cells (e.g., DeMott and Rutledge 1998). A greater height of the 30-dBZ surface suggests larger concentrations and/or larger particle sizes of condensate in the mixed-phase region of the cell, and thus stronger convective updrafts. Infrared cloudiness area (number of pixels) was calculated for various brightness temperature thresholds. The colder thresholds represented convective cloud tops and cold anvil cloud, while the warmest thresholds were related to the total cloudiness including midlevel clouds.
a. Radar and infrared diurnal composites for the TRMM-LBA campaign (49 days)
The radar observations showed that, following a midmorning minimum, a sharp increase in total rain area (Fig. 2a) between 1000 and 1400 LT was coincident with a similarly sharp increase in convective rainfall (Fig. 2b), echo-top height, and 30-dBZ surface height (Fig. 2c) over the same time period. This was the signature of rapid, widespread formation of precipitating convective cells near local noon (hereafter “noon balloon,” a term coined by E. Williams). Convective rain area and rainfall decreased after 1600 LT, while stratiform rain area remained constant and mean echo-top height continued to increase. These observations suggested that by 1600 LT, active convection weakened, stratiform anvil clouds expanded in area, and rainfall decreased, all pointing to the decay stage of the noon- balloon convection. By about 2200 LT, the composites suggested that the afternoon convective activity had come to an end, having collapsed in vertical intensity and diminished in area and rainfall. Animation of radar imagery for each day showed a consistency with this general scenario for most days.
After 2200 LT, the radar composites indicated a resurgence of activity. Stratiform rain area increased sharply again between 2300 and 0300 LT, corresponding in time with increased mean echo-top height and a slight rise in convective intensity (mean 30-dBZ surface height). The 0300 LT stratiform rain area maximum was equal in amplitude to the afternoon maximum at 1500 LT. However, the convective component of rain area showed only a weak and gradual increase, in contrast to the afternoon situation, with a 0300 LT maximum. Rainfall rate showed a weak downward trend after 2200 LT into the early morning hours, with some fluctuations in convective rain during that time, particularly near 0200 LT. By 0600 LT, this nocturnal resurgence had diminished, ending in a general cessation of all activity by 0900 LT. The scenario suggested by these observations was of weak convective regeneration and a large expansion of stratiform rain area after midnight. Perusal of radar imagery revealed that nocturnal convection was indeed common, but in contrast to the afternoon convection there was no single or consistent scenario to explain what was happening at night. This will be explored further in section 3b.
Diurnal composites of infrared pixel count (areal coverage) at various brightness temperature thresholds added insight into the nocturnal activity and provided an independent perspective from the radar. The afternoon noon-balloon convection left a clear cloud-top area signature in each IR threshold (Fig. 3a), with a maximum in cloud cover between 1700 and 1800 LT. This was consistent with the time of maximum mean radar echo- top height, expected since radar echo-top patterns typically corresponded well to IR cloud-top structure within convective systems (Zipser 1988; Rickenbach 1999). The cloud-top area maximum occurred 2–3 h after the maximum in low-level rain area and rainfall, as found by the TRMM-LBA observations of Negri et al. (2002) and Machado et al. (2002), and quite consistent with the general evolution of developing, heavily raining convective cells to extensive stratiform anvil cold cloud shields (Zipser 1988; McAnelly and Cotton 1989; Yuter and Houze 1998; Rickenbach 1999).
Afternoon cloudiness diminished during the early evening hours; however, the cloud coverage at 250 K (low and midlevel cloud) began to increase again by 2200 LT, near the time of the stratiform rain area increase seen by the radar. The dominant nocturnal cloud maximum was in the low and midlevel clouds (250 K) at 0100 LT, which occurred 1 h before the peak in radar echo-top height and 2 h before the rain area maximum. Only later, at 0300 LT, a weak maximum was observed at the colder cloud area threshold (210 K; Fig. 3a), which matched the mean radar echo-top area maximum and was suggestive of some deep convective activity.
In summary, the IR and radar composites were quite consistent with one another and suggested a rapid expansion of midlevel stratiform cloud, with a relatively weak convective resurgence after midnight. These observations led to the conclusion that the nocturnal cloud systems were decoupled from the afternoon convection and that the nighttime systems evolved differently than the afternoon convective activity, with a weak convective maximum an hour or two after the expansion of midlevel stratiform cloud. This evolution of the nocturnal systems appeared to be in contrast with our understanding of the development of mesoscale convective systems, where a peak in convective rain followed by the expansion of stratiform cloud would be expected (Smull and Houze 1985; Rasmussen and Rutledge 1993). This anomaly was likely explained by the juxtaposition of two types of nocturnal precipitating systems, as will be discussed in section 3b.
One possible source of bias in these composites was that the relatively small areal coverage of the radar introduced a sampling bias from partially sampled or propagating convective systems. For example, the chance timing of convective systems entering the radar's view at arbitrary life cycle stages might account for the nocturnal observations. One way to address this was to reproduce the IR composites for a much larger area, to ensure that the life cycle of one or several mesoscale convective systems was included, yet not so large as to include regions of potentially distinct diurnal rainfall characteristics (such as the Andes Mountains or the northeast coastal region). A 10° diameter circular region centered at 10°S, 60°W, which included the radar area (but was 50 times larger in area), was used for this purpose, as shown in Fig. 1. Perusal of IR imagery indicated that multiple mesoscale cloud shields are commonly observed simultaneously in this large area, so the problem of cloud systems propagating in or out of the composite region was greatly diminished. The resulting diurnal composite of IR cloud-top area (Fig. 3b) was nearly identical to that of the IR composite within the radar area. The 0100 LT maximum in midlevel cloud area (250 K) and 0300 LT maxima at the colder thresholds were all present (though slightly weaker), as were the afternoon maxima. This consistency between the two IR diurnal composites gave confidence that system propagation or partially sampled life cycle did not contaminate the radar (and IR within the radar area) diurnal composites.
b. Nocturnal precipitation events
The next step toward understanding the nocturnal rainfall and cloudiness maxima was to look for consistent patterns in the daily time series of radar and IR parameters and animations of radar and IR imagery. Examination of the midlevel IR cloud threshold (250 K) time series showed that 38 of the 49 experiment days had a clear and distinct postmidnight maximum in cloudiness. Of these 38 days, 13 had significant gaps (at least 10 missing hours in a 24-h period) in either the IR or radar data. These days were removed from subsequent analysis, leaving 25 days in which the nocturnal cloudiness maxima were well sampled. Careful perusal of radar and IR image animations and daily time series suggested two general scenarios qualitatively describing the nocturnal systems. On many of these nights, one or more squall line systems propagated into the radar view. These systems typically consisted of either a large squall-line system, or a series of smaller groups of active convection (hereafter referred to as “nocturnal convection”), with associated stratiform rain. On other nights, large regions of light stratiform precipitation and midlevel cloudiness formed within and well beyond the radar area, with no clearly associated deep convection (termed “nocturnal drizzle” in this paper). During the TRMM-LBA campaign, many scientists in the field observed extensive nocturnal drizzle and midlevel overcast that at times persisted after sunrise. Such events were quite distinct from nocturnal convective systems, and of sufficient frequency and uniqueness to warrant further investigation.
Table 1 presents a listing of the two general categories of nocturnal systems (nocturnal convection and nocturnal drizzle) observed in the early morning hours for the 25 days. Maximum cloudiness occurred between 0000 and 0700 LT, with no specific time favored by either category of nocturnal systems. The 12-h (2200–1000 LT) mean rainfall rate tended to be about 5 to 10 times less for nocturnal drizzle events. Occurrence frequency was nearly the same (14 nocturnal convection days versus 11 nocturnal drizzle days), with no strong tendency for either category to be clustered in time. The exception was the grouping of nocturnal drizzle events between 29 January and 4 February, which corresponded to a weak northwesterly low-level wind regime (Rickenbach et al. 2002; Halverson et al. 2002). Diurnal composites and statistics of nocturnal convection and nocturnal drizzle will be presented in section 3c. However, the composite analyses obscured important details of the structure and evolution of these systems that gave insight to their occurrence at night. Thus we turn first to two illustrative case studies.
1) Nocturnal convection: The 18 February squall line system
Afternoon noon-balloon convection was widespread in the radar area on 17 February and had decayed to patchy stratiform remnants by 2200 LT (0200 UTC 18 February). Radar and IR imagery indicated that by that time, a large, preexisting squall line began to enter the eastern edge of the radar area. Animation of IR imagery covering Amazonia strongly suggested that this squall line was a part of a large-scale, westward propagating line of cloudiness that formed 1.5 days earlier (the afternoon of 16 February) on the northeast coast of Brazil, of the type described by Cohen et al. (1995). By 1415 UTC 17 February (12 h earlier, late morning local time) the squall-line system had survived its first night and was clearly evident on IR imagery (Fig. 4a) moving westward over central Amazonia, oriented northwest to southeast and at least 1000 km in length. The line was identified by a sharp leading edge of high brightness temperature gradient (the convective line) and cold cloud trailing with respect to its westward motion (stratiform anvil region), typical of IR signatures of large squall-line systems (Smull and Houze 1985; Scofield 1987; Zipser 1988; Heymsfield and Fulton 1994; Rickenbach 1999). Twelve hours later, IR imagery at 0215 UTC 18 February (Fig. 4b) showed the squall line still well defined on its second night, 3000 km west of where it had formed, with new cold cloud elements on the southern end entering the TOGA radar view. The leading edge of the cloud line traveled approximately 550 km during this 12-h period (a propagation speed of roughly 13 m s−1), consistent with the speed of coastally generated lines reported in Cohen et al. (1995). This event occurred in a regime of low and midlevel easterly flow in Rondônia (Rickenbach et al. 2002). An easterly low-level jet has been found to be an important feature of westward propagating coastal squall lines (Silva Dias and Ferreira 1992; Warner et al. 2003). Betts et al. (2002b) implicated this particular nocturnal squall-line system in raising surface concentrations of ozone at night in the TRMM-LBA region, by vertical transport in convective downdrafts. That study also suggested, as is done herein, that this system originated along the northeast coast of Brazil.
The daily time series of radar rain area, radar rainfall rate, radar echo-top height, and IR pixel count (Fig. 5) captured the evolution of the southern portion of the squall line beginning at 0000 LT. As the squall line entered the radar view, stratiform rain area (Fig. 5a) and IR cloud-top area (250 and 230 K; Fig. 5d) increased. Radiosonde data from 2300 LT (0300 UTC; Fig. 6a) indicated an atmosphere primed for organized deep convection, with ample CAPE distributed through the column and an easterly jet centered at 4-km height. The convection on the leading edge weakened between 0000 and 0300 LT, as shown by the downward trend in the mean 30-dBZ surface height, convective rain rate, and convective rain area. This can be seen in the local IR imagery, radar reflectivity maps (2-km height), and vertical cross sections of radar reflectivity normal to the squall line (Fig. 7). The vertical cross section at 0015 LT (0415 UTC) showed the northern portion of the line with deep convection reaching 16-km echo-top height, and both a trailing and forward-sheared anvil region (consistent with northeasterly winds in the upper troposphere; Fig. 6a). Two hours later (Fig. 8) at 0215 LT (0615 UTC) the leading-edge convection had collapsed in that portion of the line, leaving an extensive stratiform rain region (11-km-maximum echo top) and large cold anvil cloud (210 K). During this period the southern portion of the line remained active. A subsequent resurgence in convective intensity and rainfall between 0300 LT (0700 UTC) and 0600 LT (1000 UTC) seen in Fig. 5 was traced to the regeneration of the northern portion of the squall line (Fig. 9). This portion of the line decayed to extensive and intense stratiform rain with a strong radar brightband signature (not shown) likely due to aggregation of melting snow near the 0°C level, suggestive of the growth of convective debris settling through mesoscale ascent (Rutledge and Houze 1986). By midmorning, all activity had ceased, and the subsequent noon-balloon convection was suppressed (Fig. 5), presumably from the stabilizing effect of the low-level cooling and drying induced by the system passage (Betts et al. 2002a; postline sounding shown in Fig. 6b).
2) Nocturnal drizzle: The 19 February event
This event occurred the night following the 18 February squall line discussed in the previous section. By the evening of 18 February, GOES-8 observed a general lack of convective activity in the entire western half of Amazonia, including the TOGA radar area, as shown in 2145 LT (0145 UTC 19 February) IR image (Fig. 10a). An hour later, IR imagery (Fig. 11a) revealed an expanding region of midlevel cloudiness (250–240-K range) over the radar domain. This cloudiness formed in situ and was not associated with nearby convection. At the same time the radar indicated the initial formation of weak echo above the 0°C level (Fig. 11b), between 5 and 8 km, within the growing midlevel cloud. In the center of the cloud mass the radar observations suggested light rainfall at or near the surface (Fig. 11b). A sounding released from the Rancho Grande site at that time (Fig. 12) passed through the western edge of this cloud feature. The sounding profiles indicated two thin saturated layers in the middle troposphere, one at 580 mb (the 0°C level, 4.5 km above the surface) and the other at about 440 mb (−15°C or 7.5 km above the surface), and a relatively low CAPE value of 858 J kg−1. This implied two stratus cloud layers with the lower cloud base at 4.5 km (the 0°C level) and upper cloud top at 7.5 km (consistent with the IR brightness temperature image in Fig. 11a). Strong southeasterly flow in the upper troposphere transitioned to weak northeasterly winds at and below cloud top, raising the possibility that shear-induced instability had a role in generating the midlevel cloud. However, such a mechanism does not explain the nocturnal timing of the cloud formation.
Cloud cover and light precipitation area continued to increase for the next couple of hours. Time series of radar and IR parameters (Figs. 5 and 13) in the radar domain showed that the midlevel cloud cover maximum (250 K) at 0000 LT (0400 UTC) preceded the rain area maximum by 1 h. This was consistent with the observed formation of precipitation (and clouds) in the middle troposphere and subsequent descent by gravitational settling. Figure 14a shows the cloud cover near the time of maximum areal extent. Note the expanded area of low-level stratiform rain and the descent of midlevel radar echo to the surface (Fig. 14b), particularly in the northern portion of the domain. Observations from the National Oceanic and Atmospheric Administration (NOAA) 915-MHz profiler (location shown in Fig. 1) captured a similar pattern of evolution that night, with echo formation between 5 and 7 km AGL, descent and brightband formation, and light surface rainfall. The profiler data imagery for this time (which may be seen online at http://daac.gsfc.nasa.gov/CAMPAIGN_DOCS/TRMM_FE/lba/prof_prelim.shtml) did not indicate evidence of convective upward motion, with 1–2 m s−1 fall speeds above the bright band suggestive of falling snow. However, the formation of midlevel cloudiness, height of first radar echo, and presence of the radar bright band strongly suggested that weak upward air motion occurred above the melting level.
Two hours later [0145 LT (0545 UTC)], it was clear from IR imagery (Fig. 15a) that much of the western Amazon basin was covered by midlevel clouds, which had evolved similarly to the cloudiness over the radar domain. By then, the cloud deck near the radar began to gradually dissipate. Only at this later stage, a well- defined radar bright band formed between 3 and 4 km above the surface, suggesting melting of snow or snow aggregates. Streamers from the more intense brightband features coincided with higher values of near-surface radar reflectivity (Fig. 15b), which suggested that the melting of snow aggregates enhanced rainfall at the ground. Surface rainfall data were unavailable for this event. However, Joss disdrometer data (at the profiler location; cf. Fig. 1) for the very similar 2 February nocturnal drizzle event indicated 0.5-mm rain accumulation over 2 h, with a peak rainfall rate of 0.6 mm h−1. On 19 February, light rain continued to fall under a dissipating bright band 2 h later [0345 LT (0745 UTC); Fig. 16b], associated with the remaining cloud (250– 260-K cloud-top temperatures), as the stratiform cloud dissipated further (Fig. 16a). The breakup of the nocturnal midlevel cloud was underway over western Amazonia by the approach of sunrise [0545 LT (0945 UTC); Fig. 10b].
This nocturnal drizzle event was typical of those listed in Table 1, with the main variation being the timing of cloud and rain area maxima. The 19 February event observed by the radar was clearly part of a nocturnal cloudiness pattern that affected a very large region, essentially the western half of Amazonia. Three other events clearly showed a similarly broad coverage (see Table 1), and the other days also suggested widespread occurrence of midlevel cloud at night, though more localized than the four events highlighted in Table 1.
The large-scale extent of this event (Figs. 10 and 15) argued for a midlevel cloud generation mechanism acting on a similarly large scale. Topographic effects (i.e., upslope, orographic lifting) may be discounted by observing that the spatial pattern of cloud formation did not correspond to the location of increasing terrain slope in western Amazonia (cf. Fig. 1). Instead, the nocturnal timing of cloud generation pointed to a radiative mechanism.
It seems likely that the nocturnal squall-line system the previous morning [discussed in section 3b(1)] played some role in the formation of this nocturnal drizzle event, either directly or as a result of the suppression of the prior afternoon's convection. In the first case, 12 h had passed between the breakup of the stratiform region of the prior nocturnal system and the onset of nocturnal midlevel cloud formation, with clear conditions in the interim. Soundings during that period (not shown) indicated that the midtroposphere dried and warmed. Thus it was unlikely that subsequent nocturnal midlevel cloudiness had a direct link to the earlier convective system. In the second case, the lack of afternoon convection led to clear, moist conditions in the lower troposphere, and increasing relative humidity after sunset. The first clouds appeared to be shallow and isolated (western portion of Fig. 11) but rapidly filled in horizontally at and above the 0°C level where the precipitation process became active. The speculation was that the initial low-level clouds formed as radiational cooling of the surface brought low-level air to saturation. The reason for their rapid horizontal expansion and precipitation formation just above the 0°C level was not clear. Perhaps the weak mesoscale ascent that likely caused this expansion was driven by radiative destabilization of the initial shallow cloud tops, some of which extended to the 0°C level.
c. Nocturnal precipitation: Composite analysis
Composite diurnal time series for the days with nocturnal convection and nocturnal drizzle are presented in Figs. 17 and 18. The nocturnal convection composites (Fig. 17) showed general features similar to those discussed in the 18 February squall-line case study. These include a rapid buildup to a convective rainfall, echo top, and 30-dBZ height maximum at 0200 LT, a coincident peak in cloudiness at all IR thresholds, and an expansion of stratiform rain area to about 0600 LT. The transition from a majority of convective rain to a majority of stratiform rain occurred by 0500 LT, a time of decreasing vertical intensity, indicating that deep convection had largely diminished by then. This composite evolution typified large nocturnal squall-line passage through the radar area. Such squall-line systems occurred on 7 of the 11 nocturnal convective events (Table 1). Animation of infrared imagery indicated that five of these seven large-scale squall-line systems had their origin 1.5 days earlier on the northeastern coast of Brazil [section 3b(1)]. Together, these five coastal squall lines alone accounted for 50% of the total rainfall from the nocturnal convection category. The remaining two were traced to large propagating cloud systems that formed more than a day earlier, one in the Brazilian Highlands of southeastern Amazonia and the other from extreme western Amazonia (Fig. 1). The other 4 (of 11) systems were associated with cloud systems that formed locally in southwestern Amazonia.
The nocturnal convective systems influenced the timing and intensity of the noon-balloon convection the following afternoon. The late morning increase in convective rainfall and rain area (Fig. 17) was much less rapid and was delayed by 2 h (to local noon) compared to the 49-day composite (Fig. 2). Increased blockage of solar insolation in the hours after sunrise, as suggested by twice as much midlevel (250 K) cloud area at the 1000 LT minimum compared to the 49-day composite, likely contributed to the delay of the noon balloon, in addition to the low-level cooling and drying by the previous night's convective overturning [section 3b(1)]. Furthermore, noon-balloon echo-top height and 30-dBZ surface height for the 49-day mean remained constant or increased between 1200 and 1800 LT, while for the afternoon composite following nocturnal convection, the convective height and intensity decreased after about 1400 LT (Fig. 17c). This suggested that the onset of deep afternoon convection was delayed, with somewhat fewer, shallower, and weaker convective cells on the day following the nocturnal convective systems.
In contrast to the large rainrates associated with nocturnal convection, the main signature of nocturnal drizzle events (Fig. 18) was a peak in the stratiform rain area near 0200 LT (with a resurgence after 0600 LT). However, a weak maximum in the 30-dBZ surface height by 0400 LT, though 2 km lower than that for nocturnal convection, hinted that weak, embedded convective cells might at times have played a role in the maintenance of nocturnal drizzle, although a strengthening bright band could also explain this observation.
The weak rainfall intensity associated with nocturnal drizzle events suggested that these systems played a minor role in the total nocturnal rainfall. Nevertheless, the expansive midlevel cloudiness associated with these systems was likely important, for example, to the total radiation budget. The nocturnal 250-K cloud area maximum was nearly as large (and 2 h earlier) as the nocturnal convection cloudiness. Nocturnal drizzle systems contributed only 3% of the 48-day rainfall total (Table 2), while 25% of the total was associated with nocturnal convective events. The majority of rainfall (72%) was associated with the afternoon noon-balloon convection. Moreover, in contrast to the nighttime convection, nocturnal drizzle events apparently had little influence on the next afternoon's noon balloon, as seen by the sharp increase in convective rainrate and rain area in Fig. 18.
This paper examined the origins of a secondary nocturnal maximum in cloudiness and precipitation in southwestern Amazonia during the TRMM-LBA field campaign, which had been observed previously by many investigators. The general finding was that following the collapse of the nearly ubiquitous and locally generated afternoon (“noon balloon”) convection by 2200 LT, organized deep convection contributed to a postmidnight maximum in raining area and high cloudiness and to a lesser extent rainfall, following a nocturnal resurgence in vertical intensity of convection and subsequent expansion of stratiform rain area. For 7 of 11 cases, the timing of the nocturnal convective resurgence was tied to the propagation of large-scale bands of cloudiness associated with squall lines generated more than 2000 km away. Of these seven cases, five could be traced to large-scale squall-line systems that had formed 1.5 days earlier along the northeastern coast of Brazil (e.g., Cohen et al. 1995; Greco et al. 1990). These five coastal squall lines alone accounted for half of the rainfall from all nocturnal systems sampled by the radar. Each of these squall lines occurred in a regime of easterly flow in the lower troposphere (Rickenbach et al. 2002), consistent with the importance of an easterly jet in the formation and maintenance of the coastal squall lines (Silva Dias and Ferreira 1992). This scenario was consistent with the findings of westward propagation in the phase of diurnal rainfall maxima across the Amazon basin (Garreaud and Wallace 1997). Nocturnal convective systems were found to delay the onset of and weaken the noon-balloon convection the following afternoon. Observations suggested that these nocturnal systems delayed the initiation of the next day's local convection by cooling and drying from widespread precipitation and by blocking of solar insolation by extensive morning cloud cover. Moreover, there was a tendency for decreased 12-h-mean rainfall on the afternoon following nocturnal convective events. These results suggest the importance of remote, large-scale, propagating convective systems in modulating the timing and intensity of the diurnal variation of clouds and rainfall in western Amazonia.
In addition, a previously undescribed nocturnal drizzle phenomenon, independent from nocturnal convective systems, was shown to contribute significantly to nocturnal midlevel cloudiness, and slightly to precipitation. On nearly one of every three nights in the 49- day sample, widespread cloudiness and precipitation was generated in situ between 5 and 8 km in height, suggesting that the origin of this phenomenon lay above the 0°C level and was not rooted in the boundary layer. Nocturnal midlevel cloudiness was often observed to form over large portions of western Amazonia in regions undisturbed during the previous afternoon by deep convective systems. Radar observations showed that the fallout of ice particles produced a radar brightband signature, suggesting the presence of snow aggregates and implying mesoscale or large-scale ascent. The generating mechanisms of nocturnal drizzle events were not known, but observations suggested weak upward vertical motion above the melting level, possibly driven by nocturnal destabilization of shallow cloud tops. The greatest significance of nocturnal drizzle events may be the potential effect on the diurnal radiation budget by the extensive midlevel nocturnal clouds rather than their marginal contribution to nocturnal rainfall. Further research should be directed at the full life cycles of westward propagating coastal squall lines and on detailed case studies of nocturnal drizzle events.
This project was funded by NASA Tropical Rainfall Measuring Mission Grant NAG5- 9816-6. Helpful discussions with Drs. Robert Cifelli, Vanderlei Martins, Steve Nesbitt, Ali Tokay, and Earle Williams are gratefully acknowledged. The comments of the reviewers were particularly constructive and improved the paper. Mr. Andrew Negri provided the South America topography data. Dr. Biswadev Roy produced the radiosonde data plots.
Corresponding author address: Dr. Thomas M. Rickenbach, Code 912, NASA Goddard Space Flight Center, Greenbelt, MD 20771. Email: email@example.com