Abstract

The physical mechanisms associated with precipitation in southeastern South America during spring are investigated using short-term integrations with the regional mesoscale Eta Model. An evaluation of the model’s performance using in situ measurements of precipitation as well as satellite estimates reveals that the model performed satisfactorily in the subtropics and extratropics. Deficiencies in tropical Brazil are partly related to the model’s convective adjustment scheme and possibly to surface parameterizations as well. The model forecasts reproduce all observed centers of precipitation south of about 20°S, although in some cases the magnitude is somewhat smaller. Of particular relevance for this study is the finding that spatial correlations between the model forecast and observed precipitation over Cuenca del Plata are almost as high as those obtained for the Mississippi River basin using forecasts of the National Centers for Environmental Prediction operational Eta Model. Cuenca del Plata is a basin in southeastern South America that is the water resource for a largely populated area and is well known for its agricultural production and other factors that sustain the region’s economies.

An important component of the circulation reproduced in the simulations is the low-level jet east of the Andes that feeds moisture from the Amazon basin to higher latitudes. It has a diurnal cycle with a nighttime maximum that favors increased moisture flux convergence in southeastern South America. This convergence, in turn, is associated with generalized nighttime ascent and precipitation. The results are consistent with previous observational studies that show a nighttime maximum of precipitation over the region. A second regime of precipitation is found toward the eastern coast, where maximum daytime precipitation appears to be associated with a convectively unstable atmosphere, with convection being triggered by a sea–land breeze enhanced by the topography of southern Brazil. These diurnal regimes of precipitation have a significant impact in the atmospheric water cycle in Cuenca del Plata.

The basin-averaged vertically integrated moisture flux convergence is about 4 mm day−1 and almost doubles the spring values for the Mississippi River basin. The large values may be related to the particular conditions of the period under analysis and the stronger low-level jet. The results reported here provide a preliminary description of the basin-averaged moisture flux convergence and its diurnal variability, but basin-averaged precipitation is still the component that needs to be improved. It is assumed that a blend of observations and high-resolution satellite estimates will be needed to complete the description of the atmospheric water cycle.

1. Introduction

Warm season precipitation variability over tropical South America plays an important role in atmospheric circulation due to the considerable amounts of associated diabatic heating. This heating affects the atmospheric circulation through local Hadley-type circulations (Zhou and Lau 1998) and through the excitation of waves that can propagate and influence other regions (e.g., Silva Dias et al. 1987). South American precipitation has been found to vary in a wide range of frequencies, but diurnal variations are, with the annual cycle, among the most prominent modes of variability (Kousky 1980; Silva Dias et al. 1987; Garreaud and Wallace 1997). Although the diurnal heating is a basic forcing throughout the Tropics, the diurnal cycle of precipitation exhibits marked regional variations, usually related to low-level, geographically tied circulations [e.g., land–sea breezes; Kousky (1980)].

Farther south into the subtropics, the maximum warm season precipitation over land tends to occur toward nighttime (Paegle et al. 1978; Hendon and Woodberry 1993). The region that includes this precipitation regime covers approximately northeastern Argentina, Paraguay, southern Brazil, and parts of Uruguay, and for simplicity will be referred as southeastern South America. The nighttime precipitation is in part due to mesoscale convective complexes (MCCs) that achieve their maximum size at night (Velasco and Fritsch 1987). Additionally, Virji (1981), using satellite data, found that an important mechanism providing moisture (and high θe) to higher latitudes is a northerly/northwesterly low-level jet (LLJ) east of the Andes, much in the way the Great Plains LLJ feeds moisture to the central United States (Rasmusson 1967; Bonner 1968). The relation between the LLJ and the moisture flux into northern Argentina is supported by the modeling study of Berri and Inzunza (1993).

Two other regions of precipitation are observed at the same subtropical latitudes. The first one, located over the eastern slopes of the Bolivian Altiplano, is also related to MCCs (Velasco and Fritsch 1987) and, as suggested by Garreaud (1999), depends on the strength and extent of the diurnally varying flow near the Altiplano. The second one is found over the Atlantic Ocean, where large precipitation is observed in a band known as the South Atlantic convergence zone (SACZ). According to Figueroa et al. (1995) precipitation in the SACZ is forced by the pulses of latent heat released over the Amazon basin and the forced circulation due to the Andes Mountains. Additionally, it has been found that sea surface temperatures affect the position and intensity of the SACZ precipitation (Lenters and Cook 1995). These different regions of precipitation are connected in various ways: for instance, Nogués-Paegle and Mo (1997) have shown that intraseasonal variations of precipitation in SACZ have associated changes in the southward moisture fluxes east of the Andes.

The character of precipitation variability at diurnal scales and the associated moisture fluxes brings in the question of what their role is in regional moisture budgets. The subject is important because it affects Cuenca del Plata (see Fig. 1), the basin in subtropical South America encompassing the Uruguay, Paraná, and Paraguay Rivers. (Cuenca del Plata is of great economical importance, since it sustains a large fraction of the agricultural production in South America.) However, difficulties have been found when attempting to compute moisture flux convergence over such a region. Wang and Paegle (1996) found that primarily due to large uncertainties in wind analyses, and consequently in the moisture flux convergence, estimates of moisture budgets from different global analysis datasets have considerable disagreement. Moreover, the uncertainty is not reduced by using the more recent reanalysis products, as noted by Higgins et al. (1996) for North America and by Min and Schubert (1997) for regions around the world, including one east of the Andes. Part of the problem is related to the coarse resolutions of global analyses that may miss important aspects affecting the moisture budgets (Berbery et al. 1996; Berbery and Rasmusson 1999). Those studies showed that both the diurnal cycle and the smaller spatial scale features of the circulation can be resolved adequately using regional model analyses or forecasts, thus reducing sampling errors that significantly affect the estimates of the moisture flux convergence.

Fig. 1.

Domain and grid resolution of the model. The heavy contour in southeastern South America represents Cuenca del Plata, the basin formed by the Paraná, Paraguay, and Uruguay Rivers.

Fig. 1.

Domain and grid resolution of the model. The heavy contour in southeastern South America represents Cuenca del Plata, the basin formed by the Paraná, Paraguay, and Uruguay Rivers.

The objective of this study is to investigate the physical mechanisms facilitating precipitation over southeastern South America during springtime, with some emphasis on the Cuenca del Plata. To this end a series of short-term forecasts from a regional model are used to resolve the smaller-scale aspects of the circulation. The paper also discusses the relationship of precipitation to the moisture transports and the potential forcings (dynamical and thermodynamical) that may explain the character of the diurnal variations of precipitation over this region. The paper is organized as follows: Section 2 presents the regional model employed to generate a set of forecasts at higher frequency and spatial resolution than those of global reanalyses. Observational data used to evaluate the model are also presented in section 2, and the evaluation itself is discussed in section 3. Section 4 analyzes the daily mean moisture fluxes followed by a discussion of the diurnal variability and the potential forcings of precipitation in section 5. Section 6 addresses the implications the results have on studies of the atmospheric water cycle. Finally, section 7 presents the conclusions.

2. Methodology

a. The Eta Model

This study was based on the premise that, lacking regional analyses, short-term forecasts with a regional model will serve to add information in regions and times that otherwise would not be available. To this end we used the Eta Model, the regional mesoscale operational model used at the National Centers for Environmental Prediction (NCEP) (Mesinger et al. 1988; Rogers et al. 1996). Its vertical coordinate, called eta, is a generalization of the sigma coordinate, but its surfaces are quasi-horizontal. Orography can then be defined as “steps” that are particularly well suited to reproduce sharp slopes like those of the Andes Mountains. Thus, it is not unexpected that an adapted version of NCEP’s Eta Model is currently being used operationally at the Brazilian Center for Weather Forecasts and Climate Studies.

The 1997 version of the Eta Model developed at NCEP was adapted to a domain (Fig. 1) that covers all South America and parts of the adjoining oceans with a horizontal resolution of 80 km. It has 38 unevenly distributed vertical levels (see Table 1) so that about 17 of them are found below 700 hPa over areas with low topography. This resolution is good at lower levels but may be too coarse over the high Bolivian Plateau, where vertical levels are about 35 hPa apart and may not appropriately resolve the interaction between surface and atmospheric processes in that area.

Table 1.

Eta Model vertical levels.

Eta Model vertical levels.
Eta Model vertical levels.

The model’s dynamics and physics are discussed in several articles: Rogers et al. (1996) give a general overview, the model boundary layer formulation is discussed in Janjić (1990, 1994), the convective scheme is discussed in Betts and Miller (1986) with modifications by Janjić (1994), and the explicit cloud microphysics are described in Zhao et al. (1997). The model land surface physics for surface energy–water fluxes and soil moisture–temperature are discussed in Chen et al. (1996, sections 3.1.1. and 3.1.2), Chen et al. (1997), and Betts et al. (1997). The solar and longwave radiation schemes (Lacis and Hansen 1974; Fels and Schwartzkopf 1975) interact with the model-predicted water vapor and cloud water/ice.

The topography used by the model (with heights up to 5100 m; Fig. 2) reproduces the massive block of the Andes Mountains and the sharp slopes that in some regions become practically vertical walls. The wide region with heights of about 4000 m between Chile, Peru, and Bolivia is known as the Altiplano. An extensive region of lower-level mountains over southeastern Brazil, the Brazilian Plateau, is represented in the model with heights of about 500–700 m. Between the Andes Mountains and the Brazilian Plateau, terrain heights of no more than 200 m connect the Amazon basin with Argentina and Uruguay. Finally, a smaller-scale mountain plateau also of 500–700 m over northern Brazil/southern Venezuela, known as the Guiana Highlands, also has an area of lower altitudes to the west that connects the Caribbean Sea with the Amazon basin.

Fig. 2.

Topography used in the Eta Model. Contour intervals are indicated at the bottom (note that they are not uniform).

Fig. 2.

Topography used in the Eta Model. Contour intervals are indicated at the bottom (note that they are not uniform).

b. Initial state and boundary conditions

The basic dataset for this study consists of a series of 36-h forecasts at 3-h intervals starting at 0000 UTC for each day in November 1997. NCEP–NCAR (National Center for Atmospheric Research) global reanalyses (Kalnay et al. 1996) were used to initialize the model and provide the boundary conditions (one-way interaction) every 6 h. Occasionally, interpolation from the coarser grid may result in a large value at one or a few more grid points, and in some cases the attenuation mechanisms in the model may not be enough to damp the initial large value, so that the model run could fail. This occurred in six instances, but the monthly averages are still stable.

Boundary conditions are linearly interpolated in time to fit the time step of the Eta Model. The outermost row of the Eta Model grid is interpolated from the reanalyses using a bilinear interpolation, while the second outermost row is an average of the outermost and the third row (Black 1994; Mesinger 1997). Inflow points at the perimeter of the Eta Model domain have all prognostic variables prescribed from the global reanalyses, while outflow points have the velocity component tangential to the boundary linearly extrapolated from the interior (Black 1994). This is done to avoid overspecification of the lateral boundary conditions and, thus, to obtain a well-posed problem; in this case, imprecise boundary information will propagate slowly and will not contaminate the forecast in the interior of the domain (Pielke 1984; Staniforth 1997). Possible incompatibilities between the Eta Model physics and the boundary conditions provided by the reanalyses could also be expected. McGregor (1997) reviews several applications that combine global and local models that have differing treatments of physical processes and numerical methods as well as resolution. Resulting incompatibilities near model interfaces may provoke spurious response near the perimeter of the highly resolved domain that could degrade the interior simulation. Nevertheless, this was not apparent in our integrations, probably due to the large domain used in our experiments, and the relatively short integration period. [See Warner et al. (1997) for a discussion on domain sizes using the Eta Model.]

Monthly averages were computed for each forecast time to produce a “mean” diurnal cycle, and the 12–36-h forecasts were averaged together to construct monthly fields. The choice of November 1997 was based on the fact that the warm season precipitation processes in subtropical South America are already established (see, e.g., Horel et al. 1989; Zhou and Lau 1998; Grimm et al. 1998). Also, November is the month with largest MCC activity south of 20°S (Velasco and Fritsch 1987), and, further, 1997 was an El Niño year, which typically increases MCC activity (Velasco and Fritsch 1987). Other observational studies (Ropelewski and Halpert 1987;Grimm et al. 1998) also identify the region of southeastern South America as influenced by El Niño. During these events, large positive precipitation anomalies develop and result in widespread and devastating floods covering areas of Cuenca del Plata (e.g., 1983 and 1997).

c. Observational data

To evaluate the model’s performance and to complement the Eta Model forecasts, independent measurements of precipitation were obtained in the form of in situ observations and satellite estimates. Observed precipitation measurements at more than 400 rain gauge stations distributed in Argentina, Brazil, Chile, Paraguay, and Uruguay were used for model evaluation. Although the distribution of stations is not even (see Fig. 3a) it provides an opportunity to estimate the model’s performance quantitatively.

Fig. 3.

(a) Distribution of rain gauges available for model evaluation; (b) Nov 1997 observed precipitation in mm; (c) same as (b) but estimated from satellites (Xie and Arkin 1997); (d) Eta Model forecast precipitation, computed as the sum of the 12–36-h forecast for all days in Nov 1997. Contour intervals are denoted at the bottom.

Fig. 3.

(a) Distribution of rain gauges available for model evaluation; (b) Nov 1997 observed precipitation in mm; (c) same as (b) but estimated from satellites (Xie and Arkin 1997); (d) Eta Model forecast precipitation, computed as the sum of the 12–36-h forecast for all days in Nov 1997. Contour intervals are denoted at the bottom.

Satellite estimates of precipitation (Xie and Arkin 1997) were employed to assess the spatial distribution of precipitation in regions where observations were not available. Estimates on a 2.5° × 2.5° latitude–longitude grid are the result of merging several satellite products and observations, including the infrared-based Geostationary Operational Environmental Satellite precipitation index, outgoing longwave radiation–based precipitation index, and microwave measurements. For our evaluation we used the dataset version that does not include the NCEP–NCAR reanalysis precipitation forecasts (see Xie and Arkin’s paper for more details).

Estimates of precipitation at 0.5° × 0.5° latitude–longitude resolution from National Aeronautics and Space Administration (NASA) polar-orbiting satellites using Special Sensor Microwave/Imager instruments [GPROF4.0 dataset; see Negri et al. (1994)] were also used to verify results but are not shown. Finally, a 0.5° × 0.5° latitude–longitude resolution November 1985–91 climatology of convective cloudiness [clouds with tops colder than 235 K; Garreaud and Wallace (1997)] obtained from geostationary satellites at 3-h intervals was employed to complement the analysis of diurnal precipitation variability.

3. Model evaluation

a. Precipitation fields

This version of the Eta Model had not been tested before over South America. Thus, its performance was first assessed by using independent precipitation observations. To this end, and as mentioned in section 2c, more than 400 rain gauges (Fig. 3a) over five countries were employed. The stations are unevenly distributed, so that several regions with missing information are noticed in the monthly precipitation field (Fig. 3b). Distinguishable areas of large precipitation are observed over southeastern South America, over the northern slopes of the Brazilian Plateau, and over the Amazon basin in northwestern Brazil near the border with Colombia. The satellite estimates of precipitation (Fig. 3c) tend to agree with the in situ observations (Fig. 3b) where available, although there are some discrepancies with respect to magnitude. Apart from the regions identified from in situ observations, other regions of large precipitation include Colombia, Central America, and the nearby Pacific Ocean intertropical convergence zone (ITCZ); the Atlantic sector of the ITCZ; and, to the south, the SACZ.

The Eta Model 12–36-h forecast precipitation accumulated for November (Fig. 3d) has large values over northeastern Brazil that are not supported by observations or satellite estimates. Also, there is a deficit of precipitation over the nearby Atlantic sector of the ITCZ and over northwestern Brazil. These regions will be discussed later in detail. Nevertheless, the Eta Model forecasts reproduce the other observed maxima, including the region over Colombia, Central America, and the nearby Pacific sector of the ITCZ; the widespread precipitation over southeastern South America; and, finally, the SACZ. In addition, other smaller scale maxima are noticed in the forecast precipitation that are not well captured in the coarse resolution of the satellite estimates. The first region is near the Altiplano in northern Bolivia and southeastern Peru, where warm season precipitation has been documented (Garreaud and Wallace 1997; Garreaud 1999). The second region is in southern Chile, where average November precipitation over Punta Arenas, Chile, is about 350 mm month−1 (P. Aceituno 1998, personal communication). The forecast precipitation in southern Chile appears to be slightly shifted eastward toward the higher orography; a similar shift was documented by Dunn and Horel (1994) over the Rockies, where the Eta Model had a tendency to put precipitation over the top of the mountains instead of the slopes.

b. Northern region

The Eta Model forecast precipitation in the tropical band over Brazil and the Atlantic ITCZ differs from that estimated from satellites and observations. The forecasts exhibit large precipitation over northeastern Brazil and deficits to the west and over the Atlantic ITCZ. The large precipitation is not merely a problem of spinup, since it persists at all times.

Failure in the Tropics is a common deficiency in many models (see, e.g., comments in the appendix of Zhou and Lau 1998) and the reasons are not clear yet, as there is a broad range of possibilities. Here, the effect of the parameterization of convection will be discussed, but it has been suggested (P. L. Silva Dias 1999, personal communication) that the deficiency in many models may be due to poor representation of surface processes over the tropical forests of central Brazil that could lead to a too low sensible heating and low precipitation. An additional possibility is related to the initial distribution of soil moisture: Given that the NCEP–NCAR reanalyses also have a positive bias of precipitation over northeastern Brazil and a dry bias to the west, soil moisture will have similar biases affecting our integrations through the lower boundary initial state. An additional factor in the case of the Atlantic ITCZ is its proximity to the northern and eastern boundaries of the model’s domain, which may negatively impact the forecasts as discussed in section 2. It is also likely that the mechanisms producing/inhibiting precipitation over northwestern and northeastern Brazil and the Atlantic ITCZ are linked, in a manner similar to that discussed by Gandu and Silva Dias (1998). Thus, the excess of precipitation (and therefore the release of latent heat and ascent) over northeastern Brazil may produce subsidence in neighboring areas, where precipitation could be inhibited.

To evaluate the extent of the overestimation of precipitation over northeastern Brazil in the Eta Model, a point-to-point comparison was performed in the region north of 14°S and east of 51°W. The results are summarized in the form of a scatter diagram (Fig. 4a) where each point represents the monthly precipitation at a station location. According to Fig. 4a, the model produces precipitation up to 500 mm month−1 when in fact there was no observed precipitation or it did not exceed 100 mm month−1. However, the model precipitation is not completely random, as their spatial correlation is 0.41.

Fig. 4.

Nov 1997 Eta Model precipitation vs observed precipitation at each rain gauge in (a) northeastern Brazil, (b) southeastern South America, and (c) Cuenca del Plata. Units are mm month−1.

Fig. 4.

Nov 1997 Eta Model precipitation vs observed precipitation at each rain gauge in (a) northeastern Brazil, (b) southeastern South America, and (c) Cuenca del Plata. Units are mm month−1.

This bias seems similar to that noted in the southeastern United States, where the model forecast precipitation has a bias that about doubles the observed precipitation (Berbery and Rasmusson 1999). The problem has been addressed by Manikin et al. (1998) who discuss the possible causes for excessive precipitation near coastal regions of the Gulf of Mexico. Several experiments led them to conclude that at least two factors in the convective scheme may be responsible for the bias. First, the Betts–Miller–Janjić scheme uses “reference profiles” to which the ambient temperature and moisture are nudged when deep convection is initiated. These profiles are different over land and sea, so that moist air masses being transported from the sea into land areas may be too moist for the assumed drier atmosphere over land, hence being unstable and triggering convection and heavy precipitation. Second, the model has no explicit check of capping inversions that may inhibit convection. Both factors discussed in Manikin et al. (1998) are likely to be valid over northeastern Brazil, where the trade winds produce a sustained inflow from the Atlantic Ocean (with wind speeds of about 10 m s−1 in the lowest 50-hPa layer). Additionally, the characteristics of the troposphere over land surfaces in northeastern Brazil may differ from those of the troposphere over land in North America (see section 4), so that inadequate reference profiles over land could also be an issue.

A case study test was performed along the lines suggested by Manikin et al., fixing the reference profile to one unique profile over land and sea. The results (not shown, but to be reported elsewhere) indicate that for the given day there was an overall decrease of precipitation in most of northeastern Brazil and an increase inland. This would suggest that, at least in part, the difficulties in simulating the precipitation over the tropical Brazil could be alleviated by using the corrected convection parameterization scheme.

c. Southern region

The region of interest in this study is southeastern South America, so again a point-to-point comparison of observed and forecast precipitation was done for a region between 40° and 26°S and east of 63°W. The results, summarized in Fig. 4b, show that while the model precipitation tended to be less than the observed, they have a remarkable correlation (0.72). This is of particular interest because part of Cuenca del Plata is found in this region. A similar analysis was performed with all stations within this basin (Fig. 4c) showing again that the model forecast precipitation was less than the observations but still with a correlation of 0.54. [For comparison purposes only, a similar correlation for the Mississippi River basin during summer yielded a value of 0.56 (Berbery and Rasmusson 1999).] The reduced magnitude of the forecast precipitation may be related to the excessive heat source over the eastern Amazon described in the previous section, which according to Gandu and Silva Dias (1998) may favor increased subsidence (or at least oppose the ascent) to the southwest.

Similar spatial correlations between observed precipitation and the 0–24-h forecast precipitation are markedly lower, by as much as 0.2–0.3. The reason is that the model needs some time to develop the circulations at higher resolution. The fact that the correlations increase with time is an encouraging factor supporting the approach followed in this study, by which we let a regional model fill in regions with no information.

4. Character of moisture related fields

Despite the caveats that arise from the previous discussion, the model forecasts can describe moisture-related fields in an adequate manner in the subtropics and extratropics. South America’s large landmass over the tropical region, with large evapotranspiration from the Amazon basin, results in moisture content over land comparable to or larger than over the adjoining oceans. This is noticeable in Fig. 5a, where the precipitable water exceeds 45 mm over the Amazon basin, in close agreement with the November 1972–1975 observational climatology prepared by Marques et al. (1979b). At 20°S over land, precipitable water is double that over the Pacific Ocean, and is somewhat larger than that of the Atlantic Ocean. Clearly, the subsiding effect of the subtropical anticyclones dries the atmosphere over the oceans, most markedly over the Pacific. The model-produced vertical distribution of specific humidity at three locations (Belém, near the mouth of the Amazon River; Manaus, in the heart of the Amazon River basin;and Brasilia) is closely similar to the observed counterparts described in Marques et al. (1979b) (not shown). Also, as illustrated in Fig. 5b, the layer of moist air is deeper over the continent, which may have relevant consequences for static stability as well as the sources of moisture.

Fig. 5.

(a) Precipitable water in mm and (b) cross section of specific humidity (in g kg−1) at 20°S.

Fig. 5.

(a) Precipitable water in mm and (b) cross section of specific humidity (in g kg−1) at 20°S.

The moisture flux at 950 hPa, depicted in Fig. 6a, illustrates the most common and persistent aspects of the low-level circulation during the warm season. Large moisture fluxes associated with the trade winds are noticed over the tropical Atlantic Ocean and into the Amazon basin. A region of southward low-level moisture flux into the Amazon basin occurs along the low topography between the Guiana Highlands and the northern Andes Mountains. Large southward low-level moisture fluxes also occur east of the Andes Mountains and on the southeastern coast of Brazil. The latter is related to the western boundary of the Atlantic anticyclone and the topography of southeastern Brazil, and transports moisture along the coast toward the SACZ. A narrow region with southward flow west of the Andes Mountains does not agree with the expected northward flow associated with the eastern boundary of the subtropical Pacific anticyclone, and could be related to a problem of the Eta Model handling the steep orography.

Fig. 6.

(a) Mean moisture flux at 950 hPa in m s−1 g kg−1; (b) cross section of meridional wind at 20°S. Units are m s−1; (c) cross section of meridional moisture flux at 20°S. Units are m s−1 g kg−1.

Fig. 6.

(a) Mean moisture flux at 950 hPa in m s−1 g kg−1; (b) cross section of meridional wind at 20°S. Units are m s−1; (c) cross section of meridional moisture flux at 20°S. Units are m s−1 g kg−1.

The large southward flux east of the central Andes Mountains that transports moisture from the Tropics into the subtropics and even into middle latitudes is related to the LLJ discussed in the introduction. The cross section of meridional wind at 20°S, depicted in Fig. 6b, has a southward maximum of about 15 m s−1 in the 750–900-hPa layer on the lee side of the Andes. It is of interest to compare these results with the Great Plains LLJ estimated from Eta Model analyses during May 1995–97 (May in the Northern Hemisphere may be considered equivalent to November in the Southern Hemisphere). Both the average and individual months reveal that the Great Plains LLJ, depicted in Fig. 7, tends to be somewhat wider and weaker: the average of May 1995–97 has a core of about 6 m s−1 while the largest individual monthly value, May 1996, was 12 m s−1. Thus, it is possible that the South American LLJ could be more intense than the one in the Great Plains. On the other hand, this is just a one-month estimate, and the larger values could be the result of 1997 being an intense El Niño year. (A more intense LLJ during El Niño years would be consistent with the increased precipitation in southeastern South America.) Finally, another difference between the two LLJs refers to their vertical extent in comparison to the mountains. The Great Plains LLJ exceeds the height of the Rockies (Fig. 7) while the one in South America tends to be bound by the Andes Mountains, suggesting a stronger mechanical influence of the orography.

Fig. 7.

Cross section of the meridional wind (heavy lines) and specific humidity (dashed thin lines) at 30°N computed from 3 yr of Eta Model analyses. Wind units are m s−1 and humidity units are g kg−1. See Berbery and Rasmusson (1999) for a description of this dataset.

Fig. 7.

Cross section of the meridional wind (heavy lines) and specific humidity (dashed thin lines) at 30°N computed from 3 yr of Eta Model analyses. Wind units are m s−1 and humidity units are g kg−1. See Berbery and Rasmusson (1999) for a description of this dataset.

The cross section of meridional moisture flux (Fig. 6c) shows that southward fluxes are also concentrated in a narrow longitudinal band below 700 hPa with a mean value at the core of about 180 g kg−1 m s−1 at 850–900 hPa. Again, the magnitude is larger than that estimated for the Great Plains LLJ (not shown), due to the stronger winds and somewhat larger specific humidity (Fig. 7). Still, the sources are different: Nogués-Paegle and Mo (1997) point out that while the moisture source of the Great Plains LLJ is a water mass (the Gulf of Mexico), the South American LLJ has a continental moisture source (the Amazon basin).

It is of interest to analyze some elements in the atmospheric water cycle: Fig. 8a shows that the vertically integrated moisture flux has a similarity with the moisture flux at lower levels (Fig. 6a), but now it seems to be widespread from the tropical Atlantic Ocean, which in turn is larger than the contribution from the Caribbean Sea. The model produces an easterly moisture flux near the mouth of the Amazon River of about 300 kg (m s)−1, with the November climatology of Marques et al. (1979a) showing values of 283 kg (m s)−1 at Belém (1°S, ∼48°W). The meridional component of the vertically integrated moisture flux in the forecasts is larger than in Marques et al.’s observations, but still markedly smaller than the zonal component. These results suggest that the model correctly represents the moisture transports in the Tropics, and its failure to produce the correct precipitation over tropical Brazil is mostly due to the parameterization schemes.

Fig. 8.

(a) Vertically integrated moisture flux. Units are kg (ms)−1 and the scale is presented at the bottom; (b) convergence of the vertically integrated moisture flux in mm month−1; contour interval is 50 mm month−1, and values larger than 100 mm month−1 are shaded.

Fig. 8.

(a) Vertically integrated moisture flux. Units are kg (ms)−1 and the scale is presented at the bottom; (b) convergence of the vertically integrated moisture flux in mm month−1; contour interval is 50 mm month−1, and values larger than 100 mm month−1 are shaded.

In the subtropics, the southward moisture flux related to the LLJ enters Cuenca del Plata and branches out transporting moisture to the southeast over southeastern South America, and to the south along the mountains at 30°S. Despite the realistic values of moisture flux over northeastern Brazil, the vertically integrated moisture flux convergence (Fig. 8b) may be unreliable due to the deficiencies in precipitation discussed earlier. In the other regions, moisture flux convergence is consistent with precipitation, as positive values are collocated with the areas of precipitation. This is true on the eastern side of the LLJ, over the exit region of the LLJ toward the northeastern tip of Argentina, over southern Chile, and over the SACZ.

5. Diurnal variations

a. Water vapor budget terms

It was indicated in the introduction that there is a significant diurnal cycle in precipitation, with characteristics that change from region to region. This is verified in the Eta Model forecast, which has two large precipitation maxima during nighttime (Fig. 9a), one over northeastern Argentina and the other over central Brazil. During daytime (Fig. 9b) precipitation tends to be found near the coastline with an overall decrease over southeastern South America. Intensified precipitation is noticed over the plains east of the Bolivian Altiplano. The night–day differences, depicted in Fig. 9c, show a conspicuous nighttime maximum centered in Cuenca del Plata. Daytime precipitation prevails to the south of Cuenca del Plata and along the LLJ. The first region is known to have a maximum in the early morning hours (see, e.g., Paegle et al. 1978), but the cutoff times selected here for defining daytime and nighttime may be misleading in this case.

Fig. 9.

Eta Model forecast precipitation at (a) nighttime (0600–1200 UTC) and (b) daytime (1200–1800 UTC); (c) their difference. Contour interval is 0.2 mm h−1; values larger than 0.4 mm h−1 are shaded in (a) and (b), while values larger (smaller) than 0.2 (−0.2) are shaded dark (light) in (c).

Fig. 9.

Eta Model forecast precipitation at (a) nighttime (0600–1200 UTC) and (b) daytime (1200–1800 UTC); (c) their difference. Contour interval is 0.2 mm h−1; values larger than 0.4 mm h−1 are shaded in (a) and (b), while values larger (smaller) than 0.2 (−0.2) are shaded dark (light) in (c).

The diurnal cycle of forecast precipitation over Brazil (Fig. 9) is not consistent with the observational study of Garreaud and Wallace (1997) or the simulation results of Silva Dias et al. (1987). An incorrectly represented diurnal cycle of precipitation in the Tropics might produce erroneously timed subsidence to the south, affecting the diurnal cycle in the subtropics. However, as it will be shown in the next section, the realistic diurnal cycle of forecast precipitation over southeastern South America suggests that other mechanisms may have a primary role over this region, with the subsidence effect due to the heat source over Brazil possibly modulating the intensity of precipitation.

Weak nighttime precipitation and larger daytime values east of the Altiplano (Fig. 9) are supported by the study of Garreaud (1999), who has shown a daytime maximum east of the Altiplano opposing the phase of the diurnal cycle of precipitation over the Altiplano itself. Hints of diurnal variations of precipitation are noticed over the SACZ, where high-resolution satellite estimates of precipitation (GPROF4.0; not shown) also suggest a diurnal cycle. Still, more needs to be done to clarify the accuracy of these results. Finally, precipitation over southern Chile depicts no diurnal variability, consistent with the orographic nature of the forcing (Lenters and Cook 1995).

Figure 10 presents the corresponding fields of vertically integrated moisture flux convergence. As with the overall monthly mean, there is a consistency with the nighttime–daytime forecast precipitation; according to Fig. 10a, nighttime convergence is widespread and larger than the daytime values (Fig. 10b). These changes are consistent with the nighttime increase of the intensity of the LLJ, as depicted in Fig. 10c, which shows that the nighttime–daytime moisture flux difference is as large as 100 g kg−1 m s−1. The corresponding difference in vertically integrated moisture flux convergence (Fig. 10d) reveals large convergence that maximizes in the exit region of the LLJ (northern Argentina–Paraguay), surrounded by three areas of weaker nighttime convergence: to the south (the Cuenca del Plata delta), to the west associated with the southward branch of the LLJ, and last over the coast of southern Brazil. These results suggest that the regions have different forcings, a subject that will be addressed next.

Fig. 10.

(a), (b) and (d) Same as Fig. 9 except for the vertically integrated moisture flux convergence. (c) Cross section at 20°S of nighttime minus daytime meridional moisture flux (units are m s−1 g kg−1).

Fig. 10.

(a), (b) and (d) Same as Fig. 9 except for the vertically integrated moisture flux convergence. (c) Cross section at 20°S of nighttime minus daytime meridional moisture flux (units are m s−1 g kg−1).

b. Potential forcings

In an attempt to understand the processes that contribute to the development of precipitation, the convective available potential energy (CAPE) was computed; it can be viewed as the thermodynamical forcing facilitating convection and precipitation (Bluestein 1993; Barlow et al. 1998). Clearly, CAPE maximizes during daytime (Figs. 11a–c) over southern Brazil, with values of about 1300 J kg−1. However, this computation includes all days in November, so that daily CAPE values for rainy days would be expected to be larger. The convective inhibition (CIN, not shown) is close to zero over the regions of maximum CAPE during daytime but achieves values of about 120 J kg−1 at night over northeastern Argentina. Since not only CAPE is small over that region, but also CIN tends to be relatively large, it is unlikely that thermodynamical processes are the primary forcing for the region’s precipitation.

Fig. 11.

(a)–(c) CAPE during nighttime, daytime, and their difference. Units are J kg−1. (d)–(f) The same as (a)–(c) but for vertical velocity (−ω) at 500 hPa; the sign was changed to have ascent as positive values. Units are Pa s−1.

Fig. 11.

(a)–(c) CAPE during nighttime, daytime, and their difference. Units are J kg−1. (d)–(f) The same as (a)–(c) but for vertical velocity (−ω) at 500 hPa; the sign was changed to have ascent as positive values. Units are Pa s−1.

The vertical velocity at 500 hPa is reasonably representative of the larger-scale dynamical processes and can be related to convergence in the lower levels. Figs. 11d–f show a nighttime maximum over northeastern Argentina that is consistent with the collocated maximum of precipitation and vertically integrated moisture flux convergence. Both during night and day, there is a clear orographic effect near the Andes Mountains, with ascent to the west and descent on the eastern slopes.

The maximum nighttime vertical velocity at 30°–40°W over SACZ is at the same location of the maximum nighttime precipitation; in fact, decomposition of the forecast precipitation into its “convective” (the result of convective parameterization) and “large scale” (the result of generalized ascent) components reveals that this maximum is mostly due to the large-scale component (not shown).

The diurnal march of precipitation and the potential forcings are examined in greater detail in the form of regional Hovmoeller diagrams for the latitudinal band between 25° and 30°S (Fig. 12). The Hovmoeller diagram of Eta Model forecast precipitation (Fig. 12a) depicts two distinct maxima, one occurring during the afternoon at 50°W near the coast of southern Brazil and a second one occurring at nighttime between 55° and 60°W. The independently measured satellite estimates of convective cloud frequency (Fig. 12b) show a remarkably similar diurnal evolution, supporting the soundness of the model’s performance and also the concept that the mechanisms producing precipitation during November 1997 are representative of those dominating during springtime. Note that the forecast precipitation at 55° and 60°W has not developed in the first day at the beginning of the integration (0–12-h forecast); instead, it develops toward the second day in the latter part of the integration (27–36-h forecast). Evidently, the model needs some time to adjust and develop the correct circulations. Convective cloud frequency (Fig. 12b) shows that clouds develop over the eastern slopes of the Andes (at 65°W) during the day and propagate toward the east as the evening/night progresses, with a phase speed of about 12.5 m s−1. The model produces weak precipitation at 65°W between 2100 and 2400 UTC, and it develops later in agreement with the convective cloud frequency evolution.

Fig. 12.

The 25°–30°S averaged regional Hovmoeller diagrams of (a) Eta Model precipitation (CI = 0.5 mm h−1; the 0.6 and 0.8 mm h−1 contours are included as dashed lines); (b) convective cloud frequency; (c) convective available potential energy (solid lines and shades; contour interval is 100 J kg−1) and convective inhibition (dashed lines; contour interval is 20 J kg−1); (d) vertical velocity (−ω) at 500 hPa;contour interval is 0.02 Pa s−1 and the sign was changed to have ascent as positive values. All values are for Nov 1997 except (b), which is a Nov 1985–91 climatology.

Fig. 12.

The 25°–30°S averaged regional Hovmoeller diagrams of (a) Eta Model precipitation (CI = 0.5 mm h−1; the 0.6 and 0.8 mm h−1 contours are included as dashed lines); (b) convective cloud frequency; (c) convective available potential energy (solid lines and shades; contour interval is 100 J kg−1) and convective inhibition (dashed lines; contour interval is 20 J kg−1); (d) vertical velocity (−ω) at 500 hPa;contour interval is 0.02 Pa s−1 and the sign was changed to have ascent as positive values. All values are for Nov 1997 except (b), which is a Nov 1985–91 climatology.

The thermodynamical forcing as represented by CAPE (Fig. 12c) shows a maximum that precedes by about 3–6 h the onset of precipitation around 50°W. Note that at the same time, CIN is close to zero. The land–sea contrasts during daytime, plus the orography, develop a local circulation (not shown) that is likely to trigger this convective activity. On the other hand, there is no precipitation at 55°W where although large CAPE values are found, there is also increased CIN with values of about 40 J kg−1. A maximum in CAPE at 65°W (eastern slopes of the Andes) develops between 1800 and 2100 UTC, again preceding the development of convective clouds (Fig. 12b) and light forecast precipitation (Fig. 12a). The results agree with the observational study of Paegle et al. (1978) that shows a maximum of precipitation at about 1900 local time (2000 UTC) for the same region.

The large-scale forcing is represented by the vertical velocity at 500 hPa (Fig. 12d); despite showing some ascent over the southern coast of Brazil (at about 50°W), it does not have a defined daytime maximum concurrent with the maximum convection/precipitation, suggesting that the processes are mainly thermodynamical. Conversely, the nighttime maximum of precipitation over southeastern South America agrees with the maximum in vertical velocity (Fig. 12d), and with a similar maximum in moisture flux convergence in the lower levels (not shown), suggesting the dynamical nature of the forcing. Note that in this region there is not only small CAPE, but also large CIN.

6. Implications for moisture budgets

Cuenca del Plata covers an area of approximately 3.6 × 106 km2, which is slightly larger than the Mississippi River basin (3.1 × 106 km2), and is the water resource for one of the most densely populated regions of South America. Furthermore, several hydroelectric power plants regulate the river flow and, in turn, can affect the navigability of these natural waterways. Last, harvests and livestock are also an important asset to the region. All these elements are greatly affected by precipitation variability and more generally by changes in the hydrological cycle. While research has been conducted to investigate the basin’s precipitation and streamflow variability on interannual timescales (e.g., Mechoso and Pérez Iribarren 1992; Pisciottano et al. 1994; García and Vargas 1997), the results of the previous sections make clear that there is a need to resolve the smaller-scale features of the circulation both in time and space to improve our understanding of the atmospheric component of the hydrological cycle.

In a previous article (Berbery et al. 1996) we examined the feasibility of using Eta Model operational forecasts to estimate the atmospheric hydrologic cycle over the central United States. The results showed that evaporation (estimated as a residual of the moisture budget equation) was consistent with previous studies based in observations (e.g., Rasmusson 1968). The subject was further pursued in Berbery and Rasmusson (1999), where two years of regional analyses and observations were used to evaluate the atmospheric hydrologic cycle for the Mississippi River basin and its subbasins. It was found that consistent estimates of the atmospheric hydrologic cycle relied in the ability of the Eta Model to resolve, both spatially and temporally, mesoscale circulations. One additional advantage of the higher resolution in the regional forecasts is the possibility of defining the basin boundaries closer to the physical boundaries, thus avoiding errors resulting from crossing mountainous areas, or including precipitation regimes and/or oceanic surfaces that are not part of the basin. Likewise, the higher frequency of forecasts (three hourly) should give a fair representation of the regional diurnal cycle of moisture flux convergence and precipitation.

Figure 13 shows the basin-averaged moisture flux convergence as a function height and time. Moisture flux convergence prevails at all times and two maxima are observed. The first one, between 1500 and 1800 UTC, is shallower and of smaller scale than the second between 0600 and 0900 UTC. The maximum at 1500–1800 UTC is consistent with the daytime maximum in precipitation toward the east of the basin (Fig. 12), while the second, between 0600 and 0900 UTC, is associated with the maximum in nighttime/predawn precipitation. Twice daily measurements would miss the two maxima, and probably even four analyses per day would miss some of the details of the diurnal evolution.

Fig. 13.

Diurnal cycle of moisture flux convergence averaged for Cuenca del Plata. Contour interval is 0.025 g kg−1 h−1.

Fig. 13.

Diurnal cycle of moisture flux convergence averaged for Cuenca del Plata. Contour interval is 0.025 g kg−1 h−1.

One month of data is not enough to attempt to compute moisture budgets for Cuenca del Plata, but some preliminary results may be noted. The basin-averaged vertically integrated moisture flux convergence is 4.05 mm day−1, which is about double the spring values for the Mississippi River basin (Berbery and Rasmusson 1999). This larger moisture flux convergence may be related to the stronger LLJ and the particular conditions of November 1997, as discussed in section 4. The remaining problem in moisture budgets is the reliable estimation of basin-averaged precipitation. Values ranged from a too low 2.5 mm day−1 estimated from the model to a more consistent ∼6 mm day−1 estimated from the high-resolution GPROF4.0 satellite products. It was discussed earlier that the model tends to underforecast precipitation, but the real magnitude of the difference with the actual precipitation cannot be ascertained because of the limited network of rain gauges (recall Fig. 3a). On the other hand, while GPROF4.0 appears to reproduce satisfactorily many observed features of the precipitation fields (not shown here, but see Negri et al. 1994), it remains to be determined if the similarity is also quantitative.

7. Conclusions

This study uses short-term forecasts from the regional Eta Model to add information in regions and timescales that are unavailable in the current large-scale analyses. That the model adds useful information is supported by the fact that correlations between observed precipitation and forecasts increase after the model develops detailed regional circulations. Thus, by avoiding the period of adjustments in the model and allowing time for the parameterizations to act, there is a notable improvement in the forecasts.

The central goal of this study is to use Eta Model forecasts to examine subtropical South American rainfall and related dynamical fields such as the LLJ and moisture transports. The simulations have several successful aspects. In particular, the Eta Model appears to correctly simulate precipitation maxima away from the Tropics. Notably, spatial correlations between the forecast and observed precipitation over Cuenca del Plata are almost as high as those obtained for the Mississippi River basin using NCEP’s operational Eta Model. The model also appears capable of simulating the nocturnal precipitation maximum observed over Cuenca del Plata, although it may be somewhat deficient with respect to magnitude. Other local and global models often produce spurious precipitation maxima near the Andes, but that was not apparent in our integrations.

The Eta Model shares a deficiency common to many other models in its inability to simulate relatively dry conditions, in terms of precipitation, over northeast Brazil and the wet conditions over central/northwestern Brazil; additionally, the Atlantic ITCZ is missing. Parameterization schemes, including those of convection and surface processes, may be partially responsible for the problem over land. The lack of precipitation over the Atlantic may be due to the proximity of the northern and eastern model boundary, which in turn may also relate to precipitation overprediction over northeast Brazil. It is also possible that the heat source due to the excessive precipitation over northeast Brazil could have an associated subsidence on the neighboring areas of central Brazil and the Atlantic ITCZ, as suggested by Gandu and Silva Dias (1998). Clearly, more research is needed in this area.

Integrations for November 1997 suggest that the South American LLJ is potentially stronger than the Great Plains LLJ during May over North America. Structural differences between the two LLJs are observed as well, such as their width and their relative vertical extent in relation to the height of the mountains to the west, suggesting that the topographical control is stronger on the South American LLJ. The exit region of the LLJ has an associated region of vertically integrated moisture flux convergence over southeastern South America, where large precipitation is observed.

The collocated patterns of moisture flux convergence and precipitation over southeastern South America depict a coherent diurnal cycle with nighttime maxima. The nighttime precipitation is also consistent with a dynamical forcing related to the nighttime increase of the LLJ, which in turn favors increased moisture flux convergence (mostly in the lower levels) and, as a result, widespread nighttime ascent is attained. A second regime of precipitation is noticed toward the east, where precipitation is largest during the day. In this case, the forcing does not appear to be dynamical; instead, it appears to be the result of convective instability that is triggered by a sea–land circulation enhanced by the topographic effects of the low mountains in southern Brazil.

The current study provides a preliminary analysis of the needs toward describing the atmospheric water cycle for Cuenca del Plata. The complex structure of the moisture flux convergence and its diurnal variability are reasonably resolved in the model 3-hourly forecasts, but additional work must include a longer-term period and better precipitation estimates. It is likely that a blend of observations and high-resolution satellite estimates may be necessary; in this regard, data from a promising new satellite, NASA’s Tropical Rainfall Measuring Mission, may play an important role toward better estimates of precipitation in the region.

Acknowledgments

Comments by Prof. H. Bluestein are much appreciated. The comments of Dr. Pedro Silva Dias and an anonymous reviewer helped clarify several aspects of the article; to them, our sincere thanks. Thanks also to R. Garreaud, who provided the convective cloud frequency climatic data, and to P. Aceituno, C. Nobre, G. Sampaio, M. Marino, M. Caffera, and M. Bidegain, who made available the observed precipitation. Credit should also go to E. Rogers, F. Mesinger, M. Baldwin, and K. Mitchell who gave support with the Eta Model. Finally, our thanks to G. Huffman and E. Nelkin, who provided NASA’s GPROF4.0 dataset. This work was partially supported by NOAA Grants NA76GP0479 (PACS) and NA76GP0291 (GCIP).

REFERENCES

REFERENCES
Barlow, M., S. Nigam, and E. H. Berbery, 1998: Evolution of the North American monsoon system. J. Climate, 11, 2238–2257.
Berbery, E. H., and E. M. Rasmusson, 1999: Mississippi moisture budgets on regional scales. Mon. Wea. Rev., 127, 2654–2673.
——, ——, and K. E. Mitchell, 1996: Studies of North American continental scale hydrology using Eta model forecast products. J. Geophys. Res., 101 (D3), 7305–7319.
Berri, G. J., and B. J. Inzunza, 1993: The effect of the low-level jet on the poleward water vapor transport in the central region of South America. Atmos. Environ., 27A, 335–341.
Betts, A. K., and M. J. Miller, 1986: A new convective adjustment scheme. Part II: Single column tests using GATE wave, BOMEX, ATEX and Arctic air-mass data sets. Quart. J. Roy. Meteor. Soc., 112, 693–710.
——, F. Chen, K. E. Mitchell, and Z. I. Janjić, 1997: Assessment of the land surface and boundary layer models in the two operational versions of the NCEP Eta Model using FIFE data. Mon. Wea. Rev., 125, 2896–2916.
Black, T. L., 1994: The new NMC mesoscale Eta model: Description and forecast examples. Wea. Forecasting, 9, 265–278.
Bluestein, H. B., 1993: Synoptic–Dynamic Meteorology in Midlatitudes. Vol 2. Oxford University Press, 594 pp.
Bonner, W. D., 1968: Climatology of the low-level jet. Mon. Wea. Rev., 96, 833–850.
Chen, F., K. Mitchell, and J. Schaake, 1996: Modeling of land surface evaporation by four schemes and comparison with FIFE observations J. Geophys. Res., 101 (D3), 7251–7268.
——, Z. Janjić, and K. Mitchell, 1997: Impact of the atmospheric surface-layer parameterizations in the new land-surface scheme of the NCEP mesoscale Eta model. Bound.-Layer Meteor., 85, 391–421.
Dunn, L. B., and J. D. Horel, 1994: Prediction of central Arizona convection. Part II: Further examination of the Eta model forecasts. Wea. Forecasting, 9, 508–521.
Fels, S. B., and M. D. Schwarztkopf, 1975: The simplified exchange approximation: A new method for radiative transfer calculations. J. Atmos. Sci., 32, 1475–1488.
Figueroa, S. N., P. Satyamurty, and P. L. da Silva Dias, 1995: Simulations of the summer circulation over the South American region with an Eta coordinate model. J. Atmos. Sci., 52, 1573–1584.
Gandu, A. W., and P. L. Silva Dias, 1998: Impact of tropical heat sources on the South American tropospheric upper circulation and subsidence. J. Geophys. Res., 103 (D6), 6001–6015.
García, N. O., and W. M. Vargas, 1998: The temporal climate variability in the “Río de la Plata” basin displayed by the river discharges. Climatic Change, 38, 359–379.
Garreaud, R. D., 1999: Multiscale analysis of the summertime precipitation over the central Andes. Mon. Wea. Rev., 127, 901–921.
——, and J. M. Wallace, 1997: The diurnal march of the convective cloudiness over the Americas. Mon. Wea. Rev., 125, 3157–3171.
Grimm, A. M., S. E. T. Ferraz, and J. Gomes, 1998: Precipitation anomalies in southern Brazil associated with El Niño and La Niña events. J. Climate, 11, 2863–2880.
Hendon, H. H., and K. Woodberry, 1993: The diurnal cycle of tropical convection. J. Geophys. Res., 98 (D9), 16 623–16 637.
Higgins, R. W., K. C. Mo, and S. D. Schubert, 1996: The moisture budget of the central United States in spring as evaluated in the NCEP/NCAR and the NASA/DAO reanalyses. Mon. Wea. Rev., 124, 939–963.
Horel, J. D., A. N. Hahmann, and J. Geisler, 1989: An investigation of the annual cycle of convective activity over the tropical Americas. J. Climate, 2, 1388–1403.
Janjić, Z. I., 1990: The step-mountain coordinate: Physical package. Mon. Wea. Rev., 118, 1429–1443.
——, 1994: The step mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kousky, V. E., 1980: Diurnal rainfall variation in northeast Brazil. Mon. Wea. Rev., 108, 488–498.
Lacis, A. A., and J. E. Hansen, 1974: A parameterization of the absorption of solar radiation in the earth’s atmosphere. J. Atmos. Sci., 31, 118–133.
Lenters, J. D., and K. H. Cook, 1995: Simulation and diagnosis of the regional summertime precipitation climatology of South America. J. Climate, 8, 2988–3005.
Manikin, G. F., K. E. Mitchell, D. J. Stensrud, J. S. Kain, J. Gerrity, and M. E. Baldwin, 1998: Convective scheme tests on the coastal precipitation bias in the NCEP Eta model. Preprints, 12th Conf. on Numerical Weather Prediction, Phoenix, AZ, Amer. Meteor. Soc., 107–110.
Marques, J., J. M. dos Santos, and E. Salati, 1979a: O campo do fluxo de vapor dágua atmosférico sobre a região Amazônica. Acta Amazônica, 9, 701–713.
——, ——, and ——, 1979b: O armazenamento atmosférico de vapor dágua sobre a região Amazônica. Acta Amazônica, 9, 715–721.
McGregor, J. L., 1997: Regional climate modeling. Meteor. Atmos. Phys., 63, 105–117.
Mechoso, C. R., and G. Pérez Iribarren, 1992: Streamflow in southeastern South America and the Southern Oscillation. J. Climate, 5, 1535–1539.
Mesinger, F., 1997: Dynamics of limited-area models: Formulation and numerical methods. Meteor. Atmos. Phys., 63, 3–14.
——, Z. I. Janjić, S. Ničković, D. Gavrilov, and D. G. Deaven, 1988:The step-mountain coordinate: Model description and performance for cases of alpine lee cyclogenesis and for a case of Appalachian redevelopment. Mon. Wea. Rev., 116, 1493–1518.
Min, W., and S. Schubert, 1997: The climate signal in regional moisture fluxes: A comparison of three global data assimilation products. J. Climate, 10, 2623–2642.
Negri, A. J., R. F. Adler, E. J. Nelkin, and G. J. Huffman, 1994: Regional rainfall climatologies derived from Special Sensor Microwave/Imager (SSM/I) data. Bull. Amer. Meteor. Soc., 75, 1165–1182.
Nogués-Paegle, J., and K.-C. Mo, 1997: Alternating wet and dry conditions over South America during summer. Mon. Wea. Rev., 125, 279–291.
Paegle, J., C. E. Ereño, and E. A. Collini, 1978: Variaciones diurnas de tormentas y convergencia en la capa límite sobre Argentina. Meteorologica, 8/9, 456–463.
Pielke, R. A., 1984: Mesoscale Meteorological Modeling. Academic Press, 612 pp.
Pisciottano, G., A. Díaz, G. Cazes, and C. R. Mechoso, 1994: El Niño–Southern Oscillation impact on rainfall in Uruguay. J. Climate, 7, 1286–1302.
Rasmusson, E. M., 1967: Atmospheric water vapor transport and the water balance of North America: Part I. Characteristics of the water vapor flux field. Mon. Wea. Rev., 95, 403–426.
——, 1968: Atmospheric water vapor transport and the water balance of North America: Part II. Large-scale water balance investigations. Mon. Wea. Rev., 96, 720–734.
Rogers, E., T. L. Black, D. G. Deaven, and G. J. DiMego, 1996: Changes to the operational early Eta analysis/forecast system at the National Centers for Environmental Prediction. Wea. Forecasting, 11, 391–413.
Ropelewski, C. F., and S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626.
Silva Dias, P. L., J. P. Bonatti, and V. E. Kousky, 1987: Diurnally forced tropical circulation over South America. Mon. Wea. Rev., 115, 1465–1478.
Staniforth, A., 1997: Regional modeling: A theoretical discussion. Meteor. Atmos. Phys., 63, 15–29.
Velasco, I., and J. M. Fritsch, 1987: Mesoscale convective complexes in the Americas. J. Geophys. Res., 92 (D8), 9591–9613.
Virji, H., 1981: A preliminary study of summertime tropospheric circulation patterns over South America estimated from cloud winds. Mon. Wea. Rev., 109, 599–610.
Wang, M., and J. Paegle, 1996: Impact of analysis uncertainty upon regional atmospheric moisture flux. J. Geophys. Res., 101 (D3), 7291–7303.
Warner, T. T., R. A. Peterson, and R. E. Treadon, 1997: A tutorial on lateral boundary conditions as a basic and potentially serious limitation to regional numerical weather prediction. Bull. Amer. Meteor. Soc., 78, 2599–2617.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 2539–2558.
Zhao, Q., T. L. Black, and M. E. Baldwin, 1997: Implementation of the cloud prediction scheme in the Eta Model at NCEP. Wea. Forecasting, 12, 697–712.
Zhou, J., and K.-M. Lau, 1998: Does a monsoon climate exist over South America? J. Climate, 11, 1020–1040.

Footnotes

Corresponding author address: Ernesto H. Berbery, Department of Meteorology/CICS, 3411 Computer and Space Sciences Building, University of Maryland at College Park, College Park, MD 20742-2425.