Precipitation diurnal cycle and summer ... - Wiley Online Library

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, D10108, doi:10.1029/2008JD010212, 2009

Precipitation diurnal cycle and summer climatology assessment over South America: An evaluation of Regional Climate Model version 3 simulations Rosmeri P. da Rocha,1 Carlos A. Morales,1 Santiago V. Cuadra,1 and Te´rcio Ambrizzi1 Received 2 April 2008; revised 26 February 2009; accepted 26 March 2009; published 29 May 2009.

[1] Regional Climate Model version 3 (RegCM3) simulations of 17 summers

(1988–2004) over part of South America south of 5°S were evaluated to identify model systematic errors. Model results were compared to different rainfall data sets (Climate Research Unit (CRU), Climate Prediction Center (CPC), Global Precipitation Climatology Project (GPCP), and National Centers for Environmental Prediction (NCEP) reanalysis), including the five summers mean (1998–2002) precipitation diurnal cycle observed by the Tropical Rainfall Measuring Mission (TRMM)-Precipitation Radar (PR). In spite of regional differences, the RegCM3 simulates the main observed aspects of summer climatology associated with the precipitation (northwest-southeast band of South Atlantic Convergence Zone (SACZ)) and air temperature (warmer air in the central part of the continent and colder in eastern Brazil and the Andes Mountains). At a regional scale, the main RegCM3 failures are the underestimation of the precipitation in the northern branch of the SACZ and some unrealistic intense precipitation around the Andes Mountains. However, the RegCM3 seasonal precipitation is closer to the fine-scale analyses (CPC, CRU, and TRMM-PR) than is the NCEP reanalysis, which presents an incorrect north-south orientation of SACZ and an overestimation of its intensity. The precipitation diurnal cycle observed by TRMM-PR shows pronounced contrasts between Tropics and Extratropics and land and ocean, where most of these features are simulated by RegCM3. The major similarities between the simulation and observation, especially the diurnal cycle phase, are found over the continental tropical and subtropical SACZ regions, which present afternoon maximum (1500–1800 UTC) and morning minimum (0900–1200 UTC). More specifically, over the core of SACZ, the phase and amplitude of the simulated precipitation diurnal cycle are very close to the TRMM-PR observations. Although there are amplitude differences, the RegCM3 simulates the observed nighttime rainfall in the eastern Andes Mountains, over the Atlantic Ocean, and also over northern Argentina. The main simulation deficiencies are found in the Atlantic Ocean and near the Andes Mountains. Over the Atlantic Ocean the convective scheme is not triggered; thus the rainfall arises from the grid-scale scheme and therefore differs from the TRMM-PR. Near the Andes, intense (nighttime and daytime) simulated precipitation could be a response of an incorrect circulation and topographic uplift. Finally, it is important to note that unlike most reported bias of global models, RegCM3 does not trigger the moist convection just after sunrise over the southern part of the Amazon. Citation: da Rocha, R. P., C. A. Morales, S. V. Cuadra, and T. Ambrizzi (2009), Precipitation diurnal cycle and summer climatology assessment over South America: An evaluation of Regional Climate Model version 3 simulations, J. Geophys. Res., 114, D10108, doi:10.1029/2008JD010212.

1. Introduction [2] In the last years, the performance of different regional climate models (RCM) over South America (SA) has been 1 Departamento de Cieˆncias Atmosfe´ricas, Instituto de Astronomia, Geofı´sica e Cieˆncias Atmosfe´ricas, Universidade de Saõ Paulo, Saõ Paulo, Brazil.

Copyright 2009 by the American Geophysical Union. 0148-0227/09/2008JD010212

investigated using reanalyses or Global Climate Models (GCM) results as lateral boundary forcing to drive the simulations. Most of RCM studies forced with GCM have focused on the ability of the RCM to simulate climatology [Nicolini et al., 2002; Misra et al., 2003; Druyan et al., 2002; Chou et al., 2000; Seth et al., 2007]. For this purpose, they have investigated mainly monthly and seasonal means of low-level circulation, air temperature and precipitation. Other studies using the reanalyses from the National Centers for Environmental Prediction (NCEP) [Kalnay et al.,

D10108

1 of 19

D10108

DA ROCHA ET AL.: SIMULATED PRECIPITATION DIURNAL CYCLE

1996; Kanamitsu et al., 2002] or the European Centre for Medium-Range Weather Forecast (ECMWF) [Uppala et al., 2005] as initial and boundary conditions have focused on validating the RCM simulations over SA [Druyan et al., 2002; Seth and Rojas, 2003; Fernandez et al., 2006a; Roads et al., 2003; Pal et al., 2007]. [3] Different time scales have been used to validate RCM simulations over SA. The RCM ability to simulate the interannual variability has been analyzed among others by Seth and Rojas [2003], Rojas and Seth [2003], Fernandez et al. [2006b], Cuadra and da Rocha [2006], Seth et al. [2007], Misra et al. [2002], and Druyan et al. [2002]. Many of these analyses try to identify the RCM abilities in simulating monthly and seasonal climate patterns associated with El Ninõ/La Ninã forcings. [ 4 ] Focusing on the subseasonal rainfall statistics, Rauscher et al. [2007] investigated the daily precipitation intensity, rainy season onset and withdrawal, and the frequency and duration of dry spells using four members of a long-term (20 years) Regional Climate Model version 3 (RegCM3) simulation. They showed that the RegCM3 presents an early bias in the rainfall onset and withdrawal over central Brazil, but in the northeast the interannual variability and rain season onset are better represented than in a GCM. Also investigating the subseasonal variability, Cuadra and da Rocha [2006] showed that the RegCM3 is able to simulate well, dry spells during the austral summer over southeast of Brazil. [5] Another important and less explored feature is how models represent the precipitation diurnal cycle. The failure to reproduce this pattern can affect not only the main climatic characteristics but also the energy budget and the regional circulation. For example, over the Amazon, Betts and Jakob [2002] showed that in the ECMWF model the precipitation onset occurs 2 h after the sunrise, that is, several hours before the observations. According to the authors, this could be associated with an inadequate simulation of the morning development of the nonprecipitant convective boundary layer. A similar problem was also found by Dai et al. [1999], who used a RCM (over the United States), and Dai and Trenberth [2004] with a GCM. Dai et al. [1999] pointed out that it could be due to the weak model criteria to trigger the moist convection, a common problem for three different cumulus convective parameterizations tested. Also over the Amazon, Lin et al. [2000] showed that a GCM shifted the commonly observed afternoon total rainfall peak to early morning. This occurs because the stratiform precipitation is unrealistically strong in the early morning and the convective precipitation is unrealistically weak in the afternoon. Ma and Mechoso [2007] analyzed the precipitation diurnal cycle over the west Amazon and in the central part of Brazil and found that the precipitation peaks about 6 and 4 h, respectively, before the observations. [6] For springtime, Berbery and Collini [2000] showed that the ETA model simulated a nighttime rainfall peak over the Brazilian area. However, this peak is inconsistent with observational evidence of Garreaud and Wallace [1997] (hereinafter referred to as GW), who evaluated the diurnal march of cloudiness over SA using infrared (IR) brightness temperature. However, Berbery and Collini [2000] and

D10108

Nicolini and Saulo [2006] obtained better results over center northern area of Argentina and Paraguay. In this area, the ETA model simulates a well-defined nocturnal maximum of precipitation according to the observational evidence. This nighttime peak is normally attributed to the mesoscale convective systems that are a very common feature in the austral summer over northern Argentina and Paraguay [Velasco and Fritsch, 1987; Salio et al., 2007]. [7] From a climate point of view, the evaluation of the precipitation diurnal cycle over SA is not a trivial task since very few weather stations have precipitation observations with sufficient time frequency to define it. With the recent advances in rainfall remote sensing, and especially the use of radar observations on board satellites, such as the Tropical Rainfall Measuring Mission (TRMM) [Kummerow et al., 1998] satellite, this problem can be reduced because this instrument samples the entire atmosphere over the Tropics (35°N and 35°S) every 1.5 h, with a horizontal resolution of 5 5 km2. Owing to the satellite’s orbit, many months are required to extract a statistical representative diurnal cycle over small regions of the globe [Negri et al., 2002]. [8] As emphasized by Seth et al. [2007] and Rauscher et al. [2007], it is necessary to improve the RCM physical parameterizations in order to better simulate the SA climate characteristics. However, this development requires a better understanding of how the physical parameterizations work in this region. Another aspect pointed out by Misra et al. [2002] is the importance of climate modeling studies at higher spatial and temporal resolutions over SA to improve the simulation of spatial pattern of the precipitation. Based on these discussions, the present study aims to investigate the performance of RegCM3, with medium horizontal resolution (50 km), in simulating the summer conditions and the precipitation diurnal cycle over SA (between 8° and 35°S). [9] Section 2 describes the RegCM3, climate simulation setup, and the data set used. Section 3 presents the summer climatology, while section 4 shows the simulated and observed precipitation diurnal cycle. Section 5 finishes with the summary and the main conclusions.

2. Data and Methodology 2.1. RegCM3 [10] This study uses the RegCM3 model to simulate the seasonal climate. This model was initially developed at NCAR (National Center for Atmospheric Research) on the basis of the Mesoscale Model version 4 (MM4) [Anthes et al., 1987] as documented by Giorgi et al. [1993a, 1993b] and Pal et al. [2007]. RegCM3 is a compressible model in finite differences, hydrostatic and in vertical sigma-pressure coordinate. [11] The model surface physics is solved by the Biosphere-Atmosphere Transfer Scheme (BATS) [Dickinson et al., 1993], a scheme that considers the presence of vegetation and soil interaction in the turbulent moment, energy and water vapor exchange between the surface and atmosphere. In the planetary boundary layer, the turbulent transport of heat, moment and moisture are computed as the product between the vertical gradient of these variables

2 of 19

D10108


D10108

Figure 1. Domain and topography (m) used in the RegCM3 simulations. The squares indicate the limits of subdomains investigated. and the turbulent vertical diffusion coefficient, with corrections for nonlocal turbulence proposed by Holtslag et al. [1990]. [12] The radiative transfer in the RegCM3 follows the Community Climate Model 3 (CCM3) [Kiehl et al., 1996] parameterization where the heating rate and surface fluxes for solar and infrared radiation for clear skies and cloudy conditions are computed separately. This parameterization also includes the effect of the greenhouse gases (NO2, CH4, and CFCs), atmospheric aerosols and cloud water. [13] The moist processes are solved according to the convective parameterization of the deep cumulus convection and grid-scale precipitation scheme. The present simulations used the Grell [1993] convective scheme, which is a simplification of the Arakawa and Schubert [1974] parameterization. The Grell scheme considers a single cloud represented by undiluted updraft and downdraft currents that originate, respectively, at the levels of maximum and minimum moist static energy. The convection warms and dries the environment by compensating subsidence and cools and moistens it through the detraining of the water vapor and condensate water in the top and bottom of the cloud. The Grell scheme is initiated when a lifted parcel attains the moist adiabatic. Two closure assumptions are available for Grell in the RegCM3 [Pal et al., 2007]: (1) Arakawa and Schubert [1974] closure considers that the large-scale destabilization processes are in quasi-equilibrium with the convection. (2) Fritsch and Chappell [1980] closure considers that all the available buoyant energy (CAPE) is dissipated during a specified convective time period (30 min), and it was used in this study. The grid-scale precipitation scheme [Pal et al., 2000] solves only the prognostic equation for cloud water, which is then directly used in the radiative transfer evaluations.

2.2. Climate Simulations Setup [14] The simulations were performed during the Austral Summer and started each summer at 0000 UTC of 1 November and ended on 0000 UTC at 1 March of the following year. The summer months are defined as December, January, and February (DJF). The first month, November, was considered as spin-up time, and it was excluded from the analysis. The simulations were performed from 1988 to 2004; therefore, 17 austral summer simulations were generated. [15] The initial and boundary atmospheric conditions (geopotential height, temperature, wind, and relative humidity at 13 vertical levels, from 1000 to 70 hPa) used in the simulation are from the NCEP – Department of Energy (DOE) [Kanamitsu et al., 2002] reanalysis (hereinafter referred to as R2), which is the updated version of the NCEP-NCAR [Kalnay et al., 1996] reanalysis. It should be mentioned that the reanalysis data are a statistical merger of the model first guess field and available observations. The R2 has a horizontal resolution of 2.5° 2.5° of latitude by longitude at every 6 h (0000, 0600, 1200, and 1800 UTC). The R2 precipitation analysis was also used to compare with simulations and other rainfall data sets. [16] The topography and land use cover are from the United States Geological Survey (USGS) and Global Land Cover Characterization (GLCC), respectively, with 10 min horizontal resolution [Loveland et al., 2000]. Over the ocean, the sea surface temperature was obtained from the monthly climatology (1° 1° horizontal resolution) of Reynolds and Smith [1995]. [17] A horizontal resolution of 50 km and 18 sigmapressure levels in the vertical (top of the model at 80 hPa) were used. The domain and the model topography are shown in Figure 1. The simulation domain has 118 by 88

3 of 19

D10108


west-east and north-south grid points, respectively, and only 8 grid points in the lateral boundary. The seasonal climatology and the mean precipitation diurnal cycle were evaluated in the subdomains shown in Figure 1, and also over the whole domain, hereinafter referred as the TOT area. The subdomains are named as TR1-TR4, SB1-SB4, and EX1EX4 and will be referred to, respectively, as tropical (from 15°S to 8°S), subtropical (from 25°S to 15°S), and extratropical (from 35°S to 25°S), and they do not include the lateral boundary points. 2.3. Observational Data [18] The climate simulations of precipitation will be compared with different data sets in this study. For the summer climatology, the analysis of the Climate Research Unit (CRU) [New et al., 1999; Mitchell and Jones, 2005], the Climate Prediction Center (CPC) [Silva et al., 2007], the Global Precipitation Climatology Project (GPCP) [Adler et al., 2003], and R2 are used. However, the mean seasonal precipitation diurnal cycle simulated by the RegCM3 is compared to the TRMM-Precipitation Radar (PR) measurements [Kummerow et al., 1998]. 2.3.1. CRU [19] The CRU data set developed by the University of East Anglia [Mitchell and Jones, 2005] is available from 1901 to 2002. This data set is a monthly mean of several variables (mean, maximum and minimum air temperature, relative humidity, precipitation, etc.) over the continent, with a horizontal resolution of 0.5°. Owing to its fine horizontal resolution, CRU data have been extensively used to validate regional climatology simulated by RCM in several regions of the globe [Vidale et al., 2003; Seth and Rojas, 2003; Martıńez-Castro et al., 2006; Pal et al., 2007]. This work used the air temperature and precipitation from 1988 to 2002, which totals 15 summers of validation. 2.3.2. CPC [20] The CPC precipitation analysis from the National Oceanic and Atmospheric Administration (NOAA) is described by Silva et al. [2007]. This rainfall data set is a gauge-only precipitation data quality control and analysis system that provides daily analysis with 1° horizontal resolution over Brazil from 1948 to 2004. Silva et al. [2007] showed that this data set presented high correlations and low biases when compared to the stations data and that it had a higher quality in the east of Brazil for all periods. The present study used the CPC data set from 1988 to 2004, that is, 17 summers. 2.3.3. GPCP [21] The GPCP gridded precipitation database is a combination of satellite rain estimates with ground-based rain gauge measurements [Adler et al., 2003]. This data set, available since 1979, is used to complement the CRU and CPC (available only over the continent) analyses over the ocean. 2.3.4. TRMM [22] The TRMM satellite was launched in November of 1997 and carried on board for the first time a precipitation radar (PR). In addition to the PR, this satellite also has radiometers that measure radiance in the frequencies of microwave, infrared and visible. On the basis of these measurements, the main goal of the TRMM program was to produce more detailed information about the tridimen-

D10108

sional structure of the tropical precipitation, as well as the latent heat release rate, which can significantly improve the weather prediction models [Kummerow et al., 1998]. In this study, the TRMM-PR measurements were used to analyze the diurnal cycle of precipitation. [23] Once TRMM flies above 350 km (405 km after August 2001), the diurnal sampling becomes a problem because it visits the same point every 46 days [Negri et al., 2002]. In order to adequately capture the precipitation diurnal cycle, Negri et al. [2002] suggests using an area of approximately 10° 10° and more than 3 years of TRMM observations with a minimum of 2 h time intervals [Nesbitt and Zipser, 2003]. [24] In order to describe the precipitation diurnal cycle, the estimated surface rainfall rate of the product TRMM2A25 – version 6 were clustered in areas of 10° 10° and 10° 7° (see Figure 1) in 3 h intervals during DJF months, between 1998 and 2002 (five summers). Afterward, mean rainfall rate for each region and time interval were computed. According to Negri et al. [2002], this procedure was sufficient to describe the diurnal cycle of precipitation with the TRMM satellite. In addition to the surface rainfall rate, the 2A25 product described if the measured precipitation is of convective or stratiform type. Therefore, the mean convective rainfall rate was computed to check if the numerical model was able to represent the distinct types of precipitation.

3. Seasonal Climatology: Large-Scale Features [25] Figure 2 presents the low-level (850 hPa) wind field climatology based on 17 summers (1988– 2004) for R2 and RegCM3 simulation. Over the continent, the low-level circulation simulated by the RegCM3 (Figure 2b) is similar to the R2 field (Figure 2a) that was used as initial and boundary conditions. The simulated northwest flow along the east Andes, usually referred to as the South American Low-Level Jet (LLJ) [Marengo et al., 2004; Liebmann et al., 2004; S. Sugahara et al., Condiço˜es atmosfe´ricas de grande escala associadas a jato de baixos nı´veis na Ame´rica do Sul, paper presented at Anais do VIII Congresso Brasileiro de Meteorologia, Brazilian Meteorological Society, Rio de Janeiro, Brazil, 1994], resembles the R2, though the model tends to show a smaller core with maximum wind intensity closer to the Andes. The displacement of the maximum wind core toward the Andes is probably due to the higher topography resolution present in the model but not in the R2. For example, the summer months of 2002– 2003 in the work of Herdies et al. [2007] show this westward LLJ displacement when local soundings were included in the high horizontal resolution (1°) global analysis. Over northern Argentina and Paraguay, the RegCM3 simulated a more zonal easterly wind pointing toward the Andes when compared to the R2. This feature will have an important impact on the precipitation climatology in the eastern side of the Andes, as discussed later. The anticyclonic flow at lower levels associated with the subtropical South Atlantic High (Figure 2a) is similar to R2 (Figure 2b), both in wind intensity and direction. [26] At upper levels the RegCM3 simulates the Bolivian High [e.g., Gandu and Geisler, 1991; Lenters and Cook, 1997] followed by a downstream trough over northeast Brazil (figure not shown). The Bolivian High is a result

4 of 19

D10108


D10108

700– 600 hPa. Above this layer, the simulated atmosphere is warmer and wetter than the R2. It is important to recall that these differences are not necessarily due simulation error, as the R2 have lower resolution (250 km). Therefore, it is possible that the small-scale features are simulated but they are not adequately represented by the R2. However, Giorgi et al. [2004] also found similar patterns in their experiments, that is, colder/drier atmosphere at low levels and warmer/wetter in the middle troposphere over Europe using the RegCM3 model. The justification of their results is based on RegCM3 physical parameterizations: (1) The nonlocal boundary layer parameterization used in the model tends to enhance vertical heat and moisture transport, thereby leading to a relative cooling and drying of the boundary layer. (2) The net effect of convection is to redistribute vertically, and in particular transport upward, energy and moisture, and the Grell scheme appears efficient in this redistribution, forcing a cold/dry and warm/moist in the low and middle troposphere, respectively. (3) Land surface processes included in BATS scheme may contribute

Figure 2. Seasonal climatology, averaged from 1988 to 2004, of the wind vector and wind intensity (shaded scale in m s 1 at the bottom) at 850 hPa from (a) the NCEP-DOE [Kanamitsu et al., 2002] reanalysis (R2) and (b) RegCM3.

of the latent heating release associated with the convective activity in the Amazon [Figueroa et al., 1995; Lenters and Cook, 1997] and a similar mechanism can explain in part the downstream trough. However, the main core of the simulated Bolivian High is displaced slightly southeastward in comparison to the R2. This displacement can be due to the smaller release of latent heat on the Amazon and larger on the central Andes (see Figure 4) similarly to that obtained in the numerical experiments of Fernandez et al. [2006a]. [27] The spatial patterns of the RegCM3 and R2 climatology of air temperature and specific humidity at 850 hPa are very similar (figures not shown), having spatial correlation coefficients of 0.96 and 0.91 in the TOT area, respectively. Figures 3a and 3b show the vertical cross sections of temperature and specific humidity biases along the east-west band through the latitude of 21°S. It is noted that to the east of 55°W the RegCM3 is colder (Figure 3a) and drier (Figure 3b) than the R2 from the surface up to

Figure 3. East-west vertical cross section along 21°S of latitude showing the climate (average from 1988 to 2004) bias between the RegCM3 and R2 (RegCM3-R2) for (a) temperature (°C) and (b) specific humidity (g kg 1).

5 of 19

D10108


D10108

Figure 4. Precipitation (shaded scale in mm d 1 at the bottom) summer climatology (from 1988 to 2004, except for the CRU, which is from 1988 to 2002) of the (a) CRU data set and GCPC (over the sea) analyses, (b) CPC and GCPC (over the sea) analyses, (c) R2 analysis, and (d) RegCM3 simulation.

for the low-level cold/dry bias (e.g., the treatment of drag coefficient or the transpiration processes). [28] Figure 4 shows the summer precipitation climatology for the RegCM3 simulations and CRU-GPCP (over the sea), CPC-GPCP, and R2 observational analyses. The main climate precipitation feature from CRU and CPC data sets (Figures 4a and 4b) is the South Atlantic Convergence Zone (SACZ), which organizes a precipitating band oriented in a northwest-southeast direction, from the Amazon to southeast Brazil, being responsible for simultaneous precipitation in large areas and for consecutive days [Kodama, 1992; Liebmann et al., 1999; Carvalho et al., 2002]. The RegCM3 is able to simulate most of the SACZ characteristics but not its intensity, which is weaker (about 1 – 2 mm d 1 less) than observed (Figure 4d). The R2 does not present this SACZoriented precipitating band, but it shows an intense (higher than 10 mm d 1) and unrealistic northward rainfall band, from southeast to the central north of Brazil (Figure 4c), which is not present in the CRU and CPC analyses. The rainfall in the oceanic branch of the SACZ simulated by

RegCM3 (Figure 4d) presents reasonable similarities with GPCP analysis (Figures 4a and 4b), with a small underestimation in its southeastern sector. [ 29 ] When compared with the CRU climatology (Figure 4a), the RegCM3 (Figure 4d) shows good skill in simulating the weaker precipitation in northeast Brazil and Argentina. These characteristics are not so evident in the CPC analysis (Figure 4b). However, along the eastern side of the Andes the RegCM3 tends to simulate excessive precipitation, particularly over northern Argentina, which is probably a result of intense orographic uplift due to the easterly wind simulated by the model (Figure 2b). Over the eastern coast and central western Peru the model also simulates some unrealistic rainfall areas that could be due to the RegCM3 sigma vertical coordinate. The employment of ETA vertical coordinates can reduce spurious precipitation near the mountains, but does not eliminate all of them as shown by Fernandez et al. [2006a]. [30] The summer mean rainfall for the continental grid points in the 11 subregions shown in Figure 1 is presented

6 of 19

D10108


Figure 5. Summer climatology for the subdomains (TR1EX4) and over the whole domain (TOT area) of (a) precipitation (mm d 1) and (b) air temperature (°C). The data sets used are indicated in the legend. in Figure 5a. Figure 5a shows that there is great agreement of rainfall intensity between the CRU and CPC in the majority of the subdomains, except in the areas TR1, TR2 and TR3 (tropical subdomains). This difference in the tropical regions may be related to the spatial and temporal scale of the meteorological systems that organize the precipitation over this area, where most of the rainfall comes from mesoscale systems (e.g., cloud cluster, squall lines, and cumulus convection), and due to the very sparse observational rainfall network. On the other hand, in the southeast of SA, the network is denser and presents more regular observations [Silva et al., 2007]. In addition, synoptic systems play an important role in the seasonal precipitation, which also contributes to the better agreement between CPC and CRU analyses in the region. [31] In both Tropics (TR2 to TR4) and Subtropics (SB1 to SB3), R2 presents large differences of seasonal mean precipitation in relation to the CPC and CRU analyses (Figure 5a), mainly over TR4 and SB3. The unrealistic representation of the SACZ in the R2 (Figure 4c) could explain the R2 rainfall overestimation in the TR3, TR4, and

D10108

SB3 regions. Both, RegCM3 and R2 overestimate the rainfall in the SB1 and SB2 areas. As can be seen from Figure 4, rainfall is intense on the east side of the Andes Mountains in the RegCM3 and R2 reanalyses, which can explain the deviation in relation to the CPC and CRU analyses. In any case, it seems that in these areas the RegCM3 is slightly better than the R2 to represent the seasonal values. [32] Except in the TR3 area, the summer precipitation simulated by the RegCM3 is in general closer to the CPC and CRU than R2 analyses, though the RegCM3 had been initialized and driven by R2. In simulations with a medium (or fine) horizontal resolution, the better results may occur because of the internal RegCM3 dynamics associated with local physiographic features (topography, coastline boundary, land use, and vegetation). Therefore, part of these improvements comes from regional details of seasonal precipitation simulated by the RegCM3 that is more comparable to the fine-scale analyses of CPC and CRU. The better rainfall representation by the RegCM3, in comparison to the R2, can also be related to the large-scale dynamics. Near 50°W, between 10° and 20°S, the low-level wind in the R2 weakens and veers southward (Figure 2a). This generates a convergence band in the eastern side of this region, which may contribute to the R2 excessive rainfall in the north-south band shown in Figure 4c. The RegCM3 does not simulate this large-scale pattern (low-level wind convergence/rainfall); therefore the simulated rainfall particularly in the SB3 and TR4 areas is closer to the CPC and CRU analyzes. [33] On the east side of the Andes, between Bolivia and Argentina (20° – 30°S), the RegCM3 overestimation of precipitation is probably due to excessive air uplift from the low levels. This is related to the simulated wind, which is easterly and normal to the mountains (Figure 2b). When the air moves toward the mountains it ascends, condenses and generates the intense precipitation (Figure 4d). Most of the simulated rainfall in this area results from the grid-scale precipitation scheme (see SB1 and SB2 in Figures 7 and 8), with a small contribution from the Grell convective scheme, which is in agreement with the dynamical mechanism discussed. [34] The near surface (2 m) air temperature climatology for the CRU and the simulation are presented in Figure 6. Similar to the CRU analysis (Figure 6a) the RegCM3 simulated (Figure 6b) warmer temperatures over northern Argentina and Paraguay and lower temperatures in the elevated areas of the Andes and southeastern Brazil. On the basis of Figures 6 and 5b, it can be emphasized that the simulated temperature is almost 1° to 2°C colder than the CRU in most parts of Brazil and 1°C or above warmer to the south of SA (EX1 and EX2 areas), with the TOT area presenting a cold bias of only 0.9°C. The model also presents smaller temperature biases over the elevated topography of the Andes Mountains (the SB1 and EX1 regions), where the seasonal rainfall rate is lower (Figures 5a and 5b). The cold bias in the RegCM3 simulations has been observed in other regions of the globe, such as in the Caribbean [Martıńez-Castro et al., 2006], Europe [Giorgi et al., 2004], and SA [Fernandez et al., 2006a], and it is usually attributed to a combination of physical parameterizations

7 of 19

D10108


Figure 6. Air temperature (shaded scale in °C at the bottom) summer climatology (from 1988 to 2002) of the (a) CRU data set and (b) RegCM3 simulation.

(convective, boundary layer, and land surface) deficiencies [Giorgi et al., 2004].

4. Diurnal Cycle [35] The evaluation of the precipitation diurnal cycle simulated by numerical models is very important because it helps to understand the failures and agreements of the moist processes parameterizations used by them. In the last decades, this topic has been investigated in several studies [Lieberman et al., 1994; Lin et al., 2000; Betts and Jakob, 2002; Dai et al., 1999; Dai and Trenberth, 2004; Randall et al., 1991; Ma and Mechoso, 2007]. For example, Betts and Jakob [2002] discussed that the ECMWF model over the Amazon tends to produce precipitation just after sunrise (1200 UTC), when the Planetary Boundary Layer (PBL) becomes unstable, contributing to errors in the forecasting of other variables. A similar problem was found by Dai et al. [1999] and Dai and Trenberth [2004] with a RCM and GCM, respectively, during the summer. More recently, Lin et al. [2000] reported that the diurnal cycle simulated by a

D10108

GCM over the Amazon was quite different from the observation both in the phase and amplitude. [36] Figures 7 and 8 compare the diurnal cycle of the total, convective, and the ratio of convective to total precipitation average for the summers of 1998– 2002 for the TRMM-PR and RegCM3 data. One should recall that the areas TR1-TR4, SB1-SB4, and EX1-EX4 are referred to as tropical, subtropical and extratropical, respectively, and are shown in Figure 7. Furthermore, Figure 9 shows the difference between the nighttime (0600– 1200 UTC) and daytime (1800 – 0000 UTC) precipitation following the terminology adopted by Nicolini and Saulo [2006]. Figures 7– 9 clearly indicate the differences in the precipitation diurnal cycles over various South American areas and part of the Atlantic Ocean. [37] In the tropical region (15° –8°S), from TR1 to TR4, two precipitation peaks of different intensities in the TRMM-PR measurements are evident (Figure 7): nighttime (0600 UTC, 0200– 0300 local standard time (LST)) and daytime (1800 UTC, 1400 – 1500 LST). In general, RegCM3 correctly simulated these two rainfall peaks, though the intensities are different when compared to the TRMM-PR measurements, except in the TR1. [38] In the subdomain TR1 (between 75° and 65°W and 15° – 8°S), which includes the southwestern part of the Amazon and the eastern Andes Mountains, TRMM-PR presents a more intense daytime maximum and a weaker nocturnal peak. The RegCM3 simulated the rainfall in both periods, but the peak only occurs in the afternoon and is overestimated (Figure 7) due the excessive daytime convective rain (Figures 8a and 8b). In spite of the excessive precipitation, from Figure 9 it is clear that the model simulated well the daytime precipitation spatial pattern over the Amazonian Plateau and the Andes. In the afternoon period, the excessive precipitation simulated by the RegCM3 (Figure 7) in the TR1 area is basically of convective origin (Figures 8a and 8b), and according to Pal et al. [2007], it could be related to the excessive topographic uplift associated with the strong model topography curvature in TR1 (see Figure 1). The authors suggest that smoothing the topography in these areas could reduce the excess of precipitation. However, it is important to remember that topography smoothing could modify the local circulation, and then the precipitation diurnal cycle itself. The RegCM3 unrealistic precipitation maximum over Peru (72°W and 15°S) is a daytime peak (Figure 9), and it could be associated with an incorrect low-level convergent circulation simulated (Figure 2b). [39] The nocturnal rainfall (Figure 7) in the subdomain TR1 is probably related to the rainfall in the plain that follows the eastern slopes of the Andes (Figure 9a), which was also suggested by GW, and was correctly simulated by the RegCM3 (Figure 9b). GW speculated that this nocturnal precipitation is the result of the convergence between the descendent mountain streamflow during the night (valley circulation) and the predominant northeasterly wind in the western Amazon basin. Figure 11a depicts the difference between the 1200 UTC (0800 LST) and the daily mean wind in the first model sigma level (995 hPa). The GW hypothesis can be confirmed by Figure 11a that shows the convergence of intense downslope winds along a narrow band in the eastern side of the Andes.

8 of 19

D10108


D10108

Figure 7. Seasonal mean (from 1998 to 2002 summers) of the total precipitation diurnal cycle as observed by the TRMM-PR (solid line) and RegCM3 simulation (dashed line). The values are averaged over the 10° 10° or 10° 7° (TR1-TR4) latitude by longitude square indicated in the background map. The units are mm d 1. [40] Over the southwestern Amazon, in the tropical area TR2, both RegCM3 and TRMM-PR observations present a maximum rainfall peak during the day (Figure 7), and they are of convective origin (Figures 8a and 8b), that represent until 0.7 (TRMM-PR) and 0.9 (RegCM3) of the total rainfall (Figure 8b). The TR2 area includes the Brazilian state of Rondoˆnia where the TRMM-Large-Scale Biosphere Atmosphere (LBA) [Silva Dias et al., 2002] was carried out during 1999 (11°S and 62°W). This observational data set was used by Betts and Jakob [2002] to show that the ECMWF triggers the precipitation too early, in the morning hours. More recently, in this same region, Ma and Mechoso [2007] found a similar too early initiation of the rainfall in the University of California Los Angeles (UCLA) atmospheric global model. This is in contrast with TRMM-PR (Figure 7) and the surface radar observations analyzed by Rickenbach et al. [2002], which show a strong afternoon rainfall maximum. As noted from Figure 9a, in large part of TR2 area, the rainfall occurs during the day, as a result of boundary layer instability associated with the diurnal heating (GW). According to the TRMM-PR, two small areas, one northwest (10° –5°S–65°– 60°W) and another on the east side of the Andes Mountains, contribute to the weak

secondary nighttime peak (Figure 9a). The RegCM3 better simulated the nocturnal rainfall on the east side of Andes (Figures 9a and 9b). [41] Small-scale variability can be found in the rainfall diurnal cycle over the Amazon [Angelis et al., 2004]. For example, the stations of Humaita´ and Ji-Parana´ (both inside TR2) present stronger nighttime and daytime precipitation peaks, respectively [Angelis et al., 2004], which is in accordance with TRMM-PR observations over the TR2 (Figure 7). [42] Figure 10 presents the frequency distribution of the daily precipitation from the RegCM3 and CPC analysis. In general, over TR1 and TR2 areas the RegCM3 simulated the main characteristic of the observed daily rainfall frequency with more events in the 5 to 10 mm d 1 category, and some differences in the other categories. For example, in TR1 area, the RegCM3 tendency to simulate 10– 12.5 mm d 1 or more is probably due to the overestimation of the afternoon and nighttime rainfall (Figure 7). In the TR2, the RegCM3 underestimates the intense (larger than 10 mm d 1) daily events. This could be associated with the lower intensity of the simulated nighttime rainfall coming from the convective and grid-scale precipitation (Figures 7, 8a, and

9 of 19

D10108


D10108

Figure 8. Seasonal mean (from 1998 to 2002 summers) diurnal cycle observed by the TRMM-PR (solid line) and RegCM3 simulation (dashed line) (a) of the convective precipitation (mm d 1) and (b) of the ratio of convective to total precipitation. The values are averaged over the 10° 10° or 10° 7° (TR1TR4) latitude by longitude square indicated in the background map. 8b) and could explain the seasonal negative rainfall bias discussed before (Figure 5a). [43] Over the southeastern Amazon (TR3), an intense daytime rainfall maximum and a secondary nighttime one (0600 UTC, 0300 LST, Figure 7) are observed. The TRMM-PR shows that in most parts of this subdomain the preferred time for precipitation is during the day (Figure 9a) and is probably associated with the boundary layer forcing. However, in the northern and central parts of TR3 area, the TRMM-PR indicates a nocturnal rainfall peak (Figure 9a). The northern part (between 5° and 8°S) could be associated with the propagation of the coastal squall lines. Observational studies [Kousky, 1980; Cohen et al., 1995; GW] suggest that this synoptic system forms at the mouth of the Amazon River during the afternoon and propagates inland with a velocity of approximately 15 m s 1. With this propagation velocity, the precipitation maximum would occur around 500 – 700 km inland, a distance that is very close to the position found in the TRMM-PR (Figure 9a, near 5° S), while in RegCM3 (Figure 9b) there is an overestimation of the area of the nocturnal precipitation over TR3 area. However, another northwest-southeast nighttime rainfall band, centered in 50°W and from 15° to

8°S, appears to be a new feature revealed by TRMM-PR (Figure 9a) where the RegCM3 (Figure 9b) was able to only simulate its south branch (near 11°S and 49°W). [44] In the TR3 area, for the RegCM3 and TRMM-PR data, about 60– 70% of the nighttime total precipitation comes from the grid scale (Figure 8b). During the afternoon, the TRMM-PR presents a maximum of 11 mm d 1, which is less intense in RegCM3 (Figure 7). Most parts of this bias comes from the grid-scale precipitation, which is weaker in the RegCM3 (Figures 7, 8a, and 8b). This afternoon precipitation deficit (Figure 7) is probably responsible for the excessive frequency of weak rain events simulated (Figure 10) and, therefore, the seasonal precipitation underestimation (Figure 5a) in the north sector of SACZ (Figure 4). [45] According to the TRMM-PR observations (Figure 9a) the daytime precipitation dominates over the continental part of TR4 area, while the nocturnal peak (Figure 7) occurs over the ocean and in a small area around 40°W to 10°S. Although the RegCM3 presents a similar spatial pattern of the daytime and nighttime rainfall (Figure 9b), it fails to capture the intensity of nighttime maximum (Figure 7) owing to the less intense convective precipitation simulated

10 of 19

D10108


D10108

Figure 8. (continued) (Figures 8a and 8b). Through the analysis of rain gauge data, Kousky [1980] also found a precipitation nocturnal peak in a narrow band over the east coast of northeast Brazil. However, this signal was not apparent in the infrared (IR) satellite images used by GW (see their Figure 4). According to Kousky [1980], this nighttime peak would be the result of the convergence between the land breeze and the southeast trade winds mean flow (mean flow opposed to the land breeze). In the TR4 area, the model was able to simulate the high (low) frequency of weak (intense) daily rainfall, which agrees with the CPC analysis (Figure 10). [46] Over the Subtropics (Figure 7), between 25° and 15°S, the TRMM-PR shows that in the subdomain SB1 the maximum precipitation occurs at 1800 UTC, daytime peak, and it was simulated by the RegCM3 with a lag of 6 h (0000 UTC) and much more intense. Comparing Figures 7, 8a, and 8b, it is possible to notice that the excessive precipitation simulated during the night and early morning arises from grid-scale scheme, while the convective daytime peak, although more intense, is in phase with TRMM-PR. In this area, the RegCM3 overestimation of the nighttime rainfall comes from a small area at the northern part of SB1 (Figure 9). In the remaining area, which includes the Andean Altiplano, the daytime rainfall is greater than the nocturnal rainfall in the TRMM-PR observations and

RegCM3 (Figures 9a and 9b). According to GW, the cloudiness increases rapidly between 1500 and 2000 UTC in the Andean Altiplano, reaching a maximum after 1800 UTC, which is coherent with the TRMM-PR and RegCM3 data sets (Figures 7 and 8a). The modal frequency of the daily rainfall shows disagreement between the RegCM3 and CPC analysis (Figure 10) in the SB1, which is related to the excessive nighttime grid-scale rainfall in the simulation (Figures 7, 8a, and 8b). This nighttime bias is also responsible for the systematic positive seasonal bias (Figure 5a) over this area. [47] To the east of the Andes (Figure 7), in the SB2 area, the TRMM-PR and the RegCM3 precipitation curves have similar features, although the RegCM3 shows higher rainfall rates. The afternoon maximum (1800 UTC, 1400 LST) in this area is mainly of moist convection origin in both data sets (Figures 8a and 8b) as well as the weaker nighttime peak at 0900 UTC (0500 LST). In this region, the nocturnal precipitation is normally attributed to mesoscale convective complexes (MCCs) that are more frequent during the summer and spring [Velasco and Fritsch, 1987; Salio et al., 2007] and form downstream of the South American LLJ (see Figure 2) in the east of the Andes [Salio et al., 2007; Sugahara et al., presented paper, 1994]. The area of the nocturnal rainfall simulated by the RegCM3 in SB2 is larger than the TRMM-PR observations (Figure 9).

11 of 19

D10108


Figure 9. Nighttime (0600– 1200 UTC) minus daytime (1800 – 0000 UTC) total (convective plus large-scale) precipitation (mm d 1), averaged from 1998 to 2002 summers, for (a) TRMM-PR and (b) RegCM3. The RegCM3 generates less (more) daily precipitation in the 0 – 5 mm d 1 (5 – 12.5 mm d 1) category than the CPC analysis (Figure 10). The larger frequency of the moderated/ intense daily rainfall is a consequence of the more intense grid-scale precipitation at night and convective in the afternoon (Figures 7, 8a, and 8b), which also influenced the large seasonal mean simulated (Figure 5a). However, the RegCM3 shows good agreement with CPC analysis in the simulation of extreme daily precipitation category (higher than 10.0 mm d 1). [48] In the SB3, which includes the most part of the SACZ continental area [Carvalho et al., 2002], the RegCM3 and TRMM-PR only show one total precipitation maximum at 1800 UTC (1500 LST, Figure 7) that it is basically of convective type (Figures 8a and 8b). In this area, the rainfall keeps intense at 2100 UTC (1800 LST) and both the phase and the amplitude are well reproduced by the RegCM3. Figure 9 shows that the spatial pattern of the precipitation difference between the day (1800– 0000 UTC) and night (0600 – 1200 UTC) simulated by the RegCM3 is

D10108

very similar to the TRMM-PR observations, including the strong daytime precipitation in the coastline, near the Atlantic Ocean. The daytime peak is probably the result of the strong solar heating at the surface that produces large sensible and latent heat fluxes from the surface to lower troposphere, making the atmosphere more favorable for convection. To help in this process, the sea-breeze circulation associated with the steep coastline in the region (see Figure 1) contributes to the rainfall intensification during the day (Figures 7 and 9). The simulated difference between 1800 UTC (1500 LST) and diurnal mean low-level wind in first sigma level (995 hPa), in Figure 11b, exemplifies the importance of the sea-breeze circulation on the east coast of Brazil. This circulation and the associated mass convergence are more intense on the coasts of south and southeast of Brazil (between 20° and 30°S), which is supported by previous observational studies [Oliveira et al., 2003] and numerical simulations [Silva Dias et al., 1995]. On the basis of the diurnal cycle of precipitation (Figure 7), one could conclude that the RegCM3 convective parameterization (Figures 8a and 8b) responds to both (daytime solar heating and sea-breeze development) forcings adequately. In the SB3 area the agreement of the simulated and TRMM-PR diurnal cycle (Figures 7 and 8) is linked to the RegCM3 and CPC daily rainfall frequency distribution similarities (Figure 10). [49] In the subtropical east coast of Brazil (SB4 in Figure 7) TRMM-PR presents a maximum at 1500 UTC (1200 LST), which is simulated by the RegCM3 3 h later (1800 UTC, 1500 LST), both are from convective origin (Figures 8a and 8b). A secondary maximum (5 mm d 1) is observed in the TRMM-PR at 0000 UTC, which was not simulated by the RegCM3 (Figure 7). From the comparison of Figures 7 and 8, it is noted from the TRMM-PR observations that the nocturnal precipitation maximum is produced by the convective and grid-scale process (Figure 8b), and RegCM3 failure is associated with the small intensity and contribution of nighttime convective rainfall (Figures 8a and 8b). The nocturnal rain in SB4 is produced in its oceanic part, initiating in the night and continuing during the morning (Figures 7 and 9). The model shows some skill in reproducing the different precipitation regimes between continent (daytime rainfall) and ocean (nighttime rainfall) as observed from the TRMM-PR (Figures 9a and 9b). This transition was not found in other modeling studies that analyzed the rainfall diurnal cycle over SA (see Figure 5a from Nicolini and Saulo [2006] and Figure 9c of Berbery and Collini [2000]). In this area the RegCM3 has daily rainfall modal frequency very similar to the CPC analysis (Figure 10), but it underestimates extreme events (higher than 15 mm d 1). [50] The overestimation of the precipitation produced by the RegCM3 in the Subtropics during the dawn/early morning period (0000 –1000 UTC) over SB1 and SB4 areas is mainly of grid-scale origin (Figures 7, 8a, and 8b). In this last area, this may be related to the proximity of the ocean where the available moisture can easily saturate the atmosphere, even in unfavorable large-scale conditions. However, with regard to the SB1, the low-level easterly wind directed to the Andes Mountains simulated by the RegCM3 (Figure 2b) supports the activation of the grid-

12 of 19

D10108


D10108

Figure 10. Relative frequency distribution of daily precipitation intensity (mm d 1) from CPC analysis (black) and RegCM3 (gray) during the summers from 1998 to 2002 for the areas TR1-TR4, SB1-SB4, and EX1-EX3. scale scheme and, therefore, an excessive precipitation over SB1 (Figure 7). [51] The diurnal cycle variability in the east-west direction in the Extratropics (Figure 7), from EX1 to EX4 areas, is larger than those observed in the Subtropics (25° – 15° S) and Tropics (15° – 8° S). The TRMM-PR observations (Figure 7) show that between the west (Chile) and east (Argentina) Andes (EX1) there is a maximum of precipitation at 0000 UTC (2000 LST) from convective origin (Figures 8a and 8b) and a minimum at 1500 UTC (1100 LST). The diurnal cycle simulated by the RegCM3 is smoothed and the maximum precipitation time is delayed 6 h with relation to the TRMM-PR observations. In this area, the RegCM3 and CPC analysis (Figure 10) show highfrequency occurrence of weak daily rainfall (0– 2.5 mm d 1), with an abrupt reduction of the daily rainfall frequency greater than 5 mm d 1. Although the RegCM3 reproduces the observed frequency distribution, the overestimation of the daily 2.5– 12.5 mm d 1 events generates a positive seasonal bias in this region (Figure 5a). [52] In the EX1 area the precipitation diurnal cycle could be a result of two distinct processes: (1) over the Andes and coast of Chile a daytime peak (1800 – 0000 UTC) is associated with the thermodynamic conditions of the diurnal cycle (GW), and (2) toward the east, between 70°and 65°W,

the nighttime peak (0600 – 1200 UTC) is normally associated with the MCCs activity over the central northern part of Argentina and Paraguay [Velasco and Fritsch, 1987; Salio et al., 2007; Nicolini and Saulo, 2006]. The RegCM3 simulates both processes, though it shows the nocturnal rain extending southward when compared to the TRMM-PR (Figure 9). [53] Over the region EX2, the TRMM-PR (Figures 7, 8a, and 8b) shows a first maximum at 0300 UTC (2300 LST) followed by a more intense nighttime peak (0900 UTC, 0500 LST). In this area, the RegCM3 presents a precipitation increase during the night, which is an indication that it simulates nighttime rain; however, the peak of the rainfall occurs only at 1800 UTC (1400 LST). Although the RegCM3 does not reproduce exactly the intense nocturnal peak of the TRMM-PR observations, Figure 9 shows that the simulated nocturnal rainfall covers most parts of this subdomain, which is similar to TRMM-PR. As for the EX1 area, the nighttime rain in the EX2, and part of EX3, is mainly attributed to the development of the MCCs [Velasco and Fritsch, 1987; Nicolini and Saulo, 2006; Salio et al., 2007]. Over the United States, Anderson et al. [2003] discussed the RCMs difficulty in simulating the MCCs. However, the RegCM3 nighttime precipitation in the EX2 area would be an indication of the presence of these systems

13 of 19

D10108


Figure 11. First sigma level (995 hPa) mean (1998 – 2002 summers) wind differences and the associated mass divergence (10 5 s 1) between the (a) 1200 UTC (0800 LST) minus daily mean and (b) 1800 UTC (1500 LST) minus daily mean.

in the simulations and a detailed analysis of this feature is in course. Although the RegCM3 underestimates the rainfall intensity as shown by the TRMM-PR (Figure 7), the simulated seasonal climatology is similar to the CRU and CPC analyses (Figure 5a). Additionally, the daily rainfall

D10108

frequency distribution (Figure 10) is also comparable to the CPC analysis. [54] In the south of Brazil (EX3 area), the TRMM-PR diurnal cycle is different from the other regions (Figure 7), having one precipitation maximum at 1200 UTC (0900 LST) and another at 2100 UTC (1800 LST), where the precipitation is from convective and grid-scale origin (Figures 7, 8a, and 8b). Although the RegCM3 did not identify the observed peaks, it correctly shows a tendency for total precipitation to increase during the day. Over the continental part of EX3, TRMM-PR (Figure 9a) shows a clear pattern with nighttime precipitation in the southwest half and daytime in the northeast half. The pattern in the southwest can be related to the MCCs activity and in the northeast to thermodynamic heating and topographical influence. Figures 9a and 9b show that the RegCM3 simulates correctly the limit between the two precipitation regimes. [55] Once more, over the South Atlantic Ocean in the subdomain EX4, in the RegCM3 most of the precipitation (Figure 7) arises from the grid-scale scheme where the convective scheme is almost inactive contributing with about 5% of the total rainfall along the day (Figures 8a and 8b). However, this result does not agree with the TRMM-PR that shows a convective precipitation peak in the early morning and 40 –50% of the rainfall is convective (Figures 8a and 8b), which has also been reported in other studies [Randall et al., 1991; Nesbitt and Zipser, 2003]. Nevertheless, the temporal evolution of the total precipitation simulated by the RegCM3 is very similar to the TRMM-PR (Figure 7). [56] Figures 7, 8a, 8b, and 9 indicate that over tropical and subtropical SA the Grell convective scheme used in the RegCM3 is able to capture the main characteristics of the diurnal cycle (i.e., minimum during the morning and maximum at afternoon). The Grell scheme used in this study considers the Fritsch-Chappell closure, which removes the CAPE during a convective time period. The diurnal cycle of convective precipitation indicates that in most parts of the continent, as the CAPE tends to increase in the afternoon, this closure inhibits the morning trigger of convection and produces an afternoon maximum (Figures 8a and 8b). In the central tropical and subtropical subdomains (TR2-TR3 and SB2-SB3) the TRMM-PR presents a secondary nocturnal precipitation maximum, that is equally partitioned between convective and grid scale (Figure 8b). The RegCM3 was able to reproduce the nighttime precipitation peak in the central regions (TR2-TR3 and SB2-SB3), where the gridscale scheme is activated. However, it is not clear why RegCM3 limits the convective rainfall near the coast and oceanic regions (in the TR4, SB1, and SB4 regions the TRMM-PR shows around half precipitation is convective and RegCM3 produces only 20%; Figures 8a and 8b). However, over the coastal regions, the RegCM3 grid-scale scheme is activated, indicating an environment that is near saturation, and as a consequence, the simulated total rainfall diurnal cycle is similar to the TRMM-PR observations. The Grell convective scheme produces very weak rain during the night-morning over the extratropical continental areas (EX1 to EX3; Figures 8a and 8b) and over the ocean (EX4) that could be related with the small CAPE or high convective inhibition. In these regions, the TRMM-PR data

14 of 19

D10108


Figure 12. (a) Mean seasonal (from 1998 to 2002 summers) air temperature (K) diurnal cycle as simulated by RegCM3. (b) Mean seasonal (from 2000 to 2004 summers) air temperature (K) in some stations in the east of southeast Brazil, as indicated by the circle in Figure 12a. 15 of 19

D10108

D10108


(Figure 8) presents more intense convective rain but the RegCM3 is not able to capture this feature. [57] Despite the reported biases, the Grell scheme coupled in the RegCM3 indicates better agreement with the observations than do the schemes of Betts and Miller [1986] (hereinafter referred to as BM) and Kain and Fritsch [1990] (hereinafter referred to as KF) used in the MM5 simulations from Tadross et al. [2006] that did not adequately reproduce the time of precipitation peak during two summer seasons in South Africa. In this region, the precipitation maximum for BM (KF) scheme occurred 3 h after (before) the observations. This KF result is similar to that of Betts and Jakob [2002], Dai and Trenberth [2004], and Ma and Mechoso [2007]; that is, the model precipitation is initiated at the same time that the surface heating begins. [58] Over the United States, Liang et al. [2004] compared the KF and Grell schemes in the MM5 model. From their Figure 2, the time of the rainfall peak in the KF scheme is always the same, independent of the region investigated; that is, the precipitation peak is around 1600 – 1700 LST, diverging considerably from the observations. However, Liang et al. [2004] obtained that the Grell scheme gets closer to the observations in each of the areas analyzed, except over the southeastern United States. They attributed these results to the fact that the Grell scheme responds better to the large-scale systems while the KF scheme is most influenced by the PBL forcing. Comparing those finding with our results, it seems that the Grell scheme in the RegCM3 responds well to the different forcings. For example, the precipitation over the SB3 region, where the RegCM3 results is close of the TRMM-PR, is composed by different scale systems, that is, the local forcing (daytime heating), the regional (sea-breeze circulation), and the large scale (frontal systems and the SACZ, which is a large-scale convergence system [Kodama, 1992]). [59] Even without enough observations to validate these results, the simulated air temperature diurnal cycle is presented in Figure 12a. It is noted that over the continent the maximum temperature occurs at 1800 UTC (between 1400 and 1500 LST), but it occurs earlier at 1500 UTC (between 1100 and 1200 LST) toward the coastline regions. This effect could be explained by the sea-breeze circulation that penetrates the continental area along the coastline, bringing the oceanic cold air and thus slowing the temperature increase or even advancing the maximum temperature time to 1500 UTC. To illustrate this effect, Figure 12b presents the mean temperature diurnal cycle for the summers of 2000– 2004 (observations were unavailable for all the period of 1998– 2002 of Figure 12a) from some stations in the eastern part of southeast Brazil. It is noted that near the shoreline the maximum temperature occurs at 1500 UTC, while inland it is observed at 1800 UTC, showing a good agreement between RegCM3 and observations. This result suggests that the RegCM3 is able to correctly simulate the precipitation diurnal cycle and the circulation features over the region. Further inland, the 1800 and 1500 UTC temperature difference increases slightly, but it is still small and could be explained by the precipitation diurnal cycle (before the rain onset, the cloudiness coverage increases, decreasing the solar heating and consequently constraining the temperature increase). In Figure 12, the maximum temperature

D10108

amplitude is found over the Extratropical inland areas owing to the large insolation during the Austral summer.

5. Summary and Conclusions [60] This study analyzed the summer season systematic errors of the RegCM3 model nested in the reanalysis [Kanamitsu et al., 2002; R2] over SA south of 5°S. One of the main objectives was to evaluate the model performance by comparing the simulated rainfall with different observed precipitation data sets, particularly the diurnal cycle obtained from the TRMM-PR. [61] The R2 large-scale characteristics of 17 mean summers (1988 – 2004) were well simulated by the RegCM3. The simulated low-level wind velocity and direction was similar to the R2, though RegCM3 underestimated the wind intensity in the core of the LLJ east of the Andes. At low levels, the RegCM3 was in general colder ( 1.0°C) and drier than the R2, while in the middle troposphere it was slightly warmer (+0.5°C). The RegCM3 bias vertical distribution may be associated with the model physical parameterizations (planetary boundary layer, precipitation parameterizations, etc.) and was also identified in several regions of the globe [Martıńez-Castro et al., 2006; Giorgi et al., 2004]. [62] The spatial pattern of the precipitation climatology simulated by the RegCM3 shows many features of the CRU and CPC analyses, including the lowest rain rates over northeast Brazil and northern Argentina, and a precipitation maximum in the northwest-southeast band, similar to the SACZ pattern [Kodama, 1992], although with an intensity smaller than the observed one. The R2 does not accurately represent the SACZ, showing an incorrect excessive northsouth rain band from southeast to the northeast of Brazil. Apparently, this precipitation band is associated with intense low-level wind convergence in the R2. It seems that the RegCM3 improves the circulation/precipitation regime over the central part of the SACZ and, as a consequence, it reduces the seasonal mean R2 rainfall overestimation over part of northeast Brazil (TR4 region). The main failures in the RegCM3 simulations are intense spurious precipitation maximums around the Andes Mountains and underestimation over the northern sector of the continental SACZ. [63] The objective analyses of the 17 summers showed that, in general, the rainfall simulated by the RegCM3 has a regional pattern more similar to the CPC and CRU analyses than the R2, although RegCM3 had been initialized and driven by R2. This result may be related to the internal RegCM3 dynamics associated with local features (topography, coastline boundary, land use and vegetation), producing a seasonal precipitation that is more comparable with the fine-scale analyses of the CPC and CRU. [64] The mean TRMM-PR observation from five (1998– 2002) summers showed a pronounced contrast of the precipitation diurnal cycle between parts of the Tropics (up to 8°S) and Extratropics of SA. The use of the Grell [1993] convective scheme associated with the grid-scale scheme [Pal et al., 2000] allowed the RegCM3 model to reproduce the main characteristics of the observed precipitation diurnal cycle. However, over the Tropics (TR1-TR4) and Subtropics (SB1-SB4) compared to the TRMM-PR it was apparent that RegCM3 tended to overestimate the

16 of 19

D10108


convective rainfall rates during the daytime and the opposite was obtained during the nighttime. The strongest similarities between the simulations and observations, especially related to the phase, were found over the continental SACZ region in the Tropics and Subtropics, which showed an afternoon maximum (1500– 1800 UTC) and a morning minimum (0900 – 1200 UTC) rainfall, consistent with observational studies [Lin et al., 2000; Betts and Jakob, 2002; Kousky, 1980]. Physically, the afternoon maximum is influenced by the strong solar heating at the surface that produces large sensible and latent heat flux from surface to lower troposphere creating atmospheric instability, which turns into convection. This process controls the greatest part of precipitation diurnal cycle in the tropical and subtropical areas from the Andean Altiplano to the east coast of Brazil [GW; Kousky, 1980]. Close to the Andean steep topography, the diurnal heating induces a mountain-valley circulation, intensifying the convergence of the trade winds from the Amazon basin to the Andes, and intensifying daytime rainfall on its eastern side. Over the southeast coast of Brazil there is a coupling of the diurnal heating and the seabreeze circulation, which increases the daytime precipitation intensity. RegCM3 was able to simulate those characteristics, although it overestimated the precipitation in some areas near the Andes (TR1, SB1, and SB2), where the positive biases from the nighttime and afternoon rainfall were probably due to the steep topography. Particularly, in SB1 the model positive bias was associated with an incorrect simulation of the low-level circulation from the Pacific Ocean to the Andes Mountains. Additionally, in most of the regions analyzed the RegCM3 captured the main aspects of the daily rainfall frequency distribution when compared to the CPC analysis. [65] In some regions of SA, besides the daytime peak the TRMM-PR showed a less intense (secondary) nighttime rainfall peak. Although with differences in intensity, some of this nocturnal precipitation was simulated by the RegCM3, especially in the Tropics. The RegCM3 also reproduced the nighttime rainfall along the eastern Andes Mountains, over the Atlantic Ocean, northeast Brazil, and over northern Argentina and Paraguay as shown by TRMMPR. The northern Argentina and Paraguay nighttime rainfall is normally attributed to the MCCs activity [Velasco and Fritsch, 1987; Salio et al., 2007; Nicolini and Saulo, 2006]. However, more detailed analysis is necessary to understand if RegCM3 simulates this kind of mesoscale systems and this analysis is been carried out. [66] Over the Extratropical Atlantic Ocean (EX4 area), during the night-morning, the TRMM-PR data presented convective rain, which was in agreement with the results found by Nesbitt and Zipser [2003]. However, it seems that the Grell convective scheme coupled in the RegCM3 had some difficulties in trigging the convection. Nevertheless, the grid-scale scheme was activated and thus the total rainfall diurnal cycle was closer to the TRMM-PR observation. [67] Despite the reported biases, the Grell scheme coupled in the RegCM3 showed better agreement with observation than found by Tadross et al. [2006] with BM and KF schemes coupled in the MM5 model. In that case, both schemes did not adequately reproduce the time of precipi-

D10108

tation peak during two summer seasons in South Africa. Over SA, the simulation of the nighttime rainfall over Brazil produced by Berbery and Collini [2000], using the BM scheme coupled in the ETA model, was inconsistent with observations. In the Amazon region, many studies using global models have identified a morning initiation of the precipitation that was also in disagreement with the observations [Betts and Jakob, 2002; Dai et al., 1999; Dai and Trenberth, 2004; Liang et al., 2004; Ma and Mechoso, 2007]. Using MM5 model over the United States, Liang et al. [2004] showed that the KF scheme was more influenced by the PBL forcings while the Grell scheme responds better to the large-scale forcings. Additionally, the results presented here indicate that the RegCM3 moist parameterizations (convective plus grid scale) responded to different forcing scales, that is, local (diurnal heating), mesoscale (sea-breeze and mountain-valley circulations), and large scale (SACZ, where the convective scheme was triggered at the right time and the intensity of the rain was close to the observed). [68] Contrary to the results obtained in other studies [Betts and Jakob, 2002; Dai et al., 1999; Dai and Trenberth, 2004; Tadross et al., 2006; Liang et al., 2004; Ma and Mechoso, 2007], the moist processes parameterizations in the RegCM3 do not trigger the convective precipitation after the PBL becomes unstable, that could be due the Fritsch-Chappell closure. This feature generates a response much closer to the observations. For example, the model simulates a coherent air temperature diurnal cycle that shows the main observed features: 1500 – 1800 UTC maximum and 0900 UTC minimum temperatures. [69] In spite of regional bias, RegCM3 was able to capture many of the important features related to the intensity and phase of the precipitation diurnal cycle over part of SA, being an encouraging result in terms of simulation. Of course, there are still many problems in the simulated precipitation diurnal cycle and they are currently being investigated. For instance, tests to trigger the Grell convective scheme over the ocean are being analyzed and will be presented elsewhere. [70] Acknowledgments. The present study was supported by CNPq (grant 475281/2003-9), FAPESP (grants 2001/13925-5 and 2003/01271-6), and CAPES. We thank the Weather and Climate Physics Group of the International Centre for Theoretical Physics (ICTP) for providing the RegCM3. Thanks also to NCEP, CPC/NOAA, and CRU for data sets that are available in the public domain. Finally, the authors would like to thank the Goddard Distributed Active Archive Center (DAAC), for making available the TRMM-PR data set, and the Companhia de Tecnologia de Saneamento Ambiental (CETESB), for providing the observations of air temperature in the Sao Paulo State. T. A. also had support from CNPq, FAPESP, and CAPES. The authors wish to thank Becky Burke for the English proofing and wish to thank the anonymous reviewers.

References Adler, R. F., et al. (2003), The version-2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979 – present), J. Hydrometeorol., 4, 1147 – 1167, doi:10.1175/1525-7541(2003)0042.0.CO;2. Anderson, C. J., et al. (2003), Hydrological processes in regional climate model simulations of the central United States flood of June – July 1993, J. Hydrometeorol., 4, 584 – 598, doi:10.1175/1525-7541(2003)004 2.0.CO;2. Angelis, C. F., G. R. McGregor, and C. Kidd (2004), Diurnal cycle of rainfall over the Brazilian Amazon, Clim. Res., 26, 139 – 149, doi:10.3354/cr026139.

17 of 19

D10108


Anthes, R. A., E.-Y. Hsu, and Y.-H. Kuo (1987), Description of the Penn State/NCAR Mesoscale Model Version 4 (MM4), NCAR Tech. Note, NCAR/TN-282+STR, 66 pp. Arakawa, A., and W. H. Schubert (1974), Interactions of a cumulus cloud ensemble with the large-scale environment, part I, J. Atmos. Sci., 31, 674 – 701, doi:10.1175/1520-0469(1974)0312.0.CO;2. Berbery, E. H., and E. A. Collini (2000), Springtime precipitation and water flux over southeastern South America, Mon. Weather Rev., 128, 1328 – 1346, doi:10.1175/1520-0493(2000)1282.0.CO;2. Betts, A. K., and C. Jakob (2002), Evaluation of the diurnal cycle of precipitation, surface thermodynamics, and surface fluxes in the ECMWF model using LBA data, J. Geophys. Res., 107(D20), 8045, doi:10.1029/ 2001JD000427. Betts, A. K., and M. J. Miller (1986), A new convective adjustment scheme. 2. Single column tests using GATE WAVEVE, BOMEX, ATEX and ARCTIC AIR-MASS data sets, Q. J. R. Meteorol. Soc., 112, 693 – 709. Carvalho, L. M. V., C. Jones, and B. Liebmann (2002), Extreme precipitation events in Southern South America and large-scale convective patterns in South Atlantic Convergence Zone, J. Clim., 15, 2377 – 2394, doi:10.1175/1520-0442(2002)0152.0.CO;2. Chou, S. C., A. M. B. Nunes, and I. F. A. Cavalcanti (2000), Extended range forecasts over South America using regional eta model, J. Geophys. Res., 105, 10,147 – 10,160, doi:10.1029/1999JD901137. Cohen, J. C. P., A. F. S. M.Dias, and C. A. Nobre (1995), Environmental conditions associated with Amazonian squall lines: A case study, Mon. Weather Rev., 123, 3163 – 3174, doi:10.1175/1520-0493(1995)1232.0.CO;2. Cuadra, S. V., and R. P. da Rocha (2006), Simulaçaõ Nume´rica do Clima de Veraõ Sobre o Sudeste do Brasil e sua Variabilidade, Rev. Bras. Meteorol., 21, 271 – 282. Dai, A., and K. E. Trenberth (2004), The diurnal cycle and its depiction in the Community Climate System Model, J. Clim., 17, 930 – 951, doi:10.1175/1520-0442(2004)0172.0.CO;2. Dai, A., F. Giorgi, and K. E. Trenberth (1999), Observed and modelsimulated diurnal cycle of precipitation over the contiguous United States, J. Geophys. Res., 104, 6377 – 6402, doi:10.1029/98JD02720. Dickinson, R. E., A. Henderson-Sellers, and P. J. Kennedy (1993), Biosphere-atmosphere transfer scheme (BATS) version 1E as coupled to the NCAR Community Climate Model, NCAR Tech. Note, NCAR/TN-387, 72 pp. Druyan, L. M., M. Fulakeza, and P. Lonergan (2002), Dynamic downscaling of seasonal climate predictions over Brazil, J. Clim., 15, 3411 – 3426, doi:10.1175/1520-0442(2002)0152.0.CO;2. Fernandez, J. P. R., S. H. Franchito, and V. B. Rao (2006a), Simulation of summer circulation over South American by two regional climate models. part I: Mean climatology, Theor. Appl. Climatol., 86, 247 – 260, doi:10.1007/s00704-005-0212-6. Fernandez, J. P. R., S. H. Franchito, and V. B. Rao (2006b), Simulation of summer circulation over South American by two regional climate models. part II: A comparison between 1997/1998 El Ninõ and 1998/1999 La Ninã events, Theor. Appl. Climatol., 86, 261 – 270, doi:10.1007/s00704005-0213-5. Figueroa, S. N., P. Satyamurty, and P. L. Silva Dias (1995), Simulations of the summer circulation over the South American region with an Eta coordinate model, Mon. Weather Rev., 52, 1573 – 1584. Fritsch, J. M., and C. F. Chappell (1980), Numerical prediction of convectively driven mesoscale pressure systems. part I: Convective parameterization, J. Atmos. Sci., 37, 1722 – 1733, doi:10.1175/15200469(1980)0372.0.CO;2. Gandu, A. W., and J. E. Geisler (1991), A primitive equations model study of the effect of topography on the summer circulation over tropical South America, J. Atmos. Sci., 48, 1822 – 1836, doi:10.1175/15200469(1991)0482.0.CO;2. Garreaud, R. D., and J. M. Wallace (1997), The diurnal of convective cloudiness over the Americas, Mon. Weather Rev., 125, 3157 – 3171, doi:10.1175/1520-0493(1997)1252.0.CO;2. Giorgi, F., M. R. Marinucci, and G. T. Bates (1993a), Development of a second-generation regional climate model (RegCM2). part I: Boundarylayer and radiative transfer processes, Mon. Weather Rev., 121, 2749 – 2813. Giorgi, F., M. R. Marinucci, and G. T. Bates (1993b), Development of a second-generation regional climate model (RegCM2). part II: Convective processes and assimilation of lateral boundary conditions, Mon. Weather Rev., 121, 2814 – 2832, doi:10.1175/1520-0493(1993)1212.0.CO;2. Giorgi, F., X. I. Bi, and J. S. Pal (2004), Mean, inter-annual variability and trends in a regional climate change experiment over Europe. I. Present – day climate (1961 – 1990), Clim. Dyn., 22, 733 – 756, doi:10.1007/ s00382-004-0409-x.

D10108

Grell, G. A. (1993), Prognostic evaluation of assumptions used by cumulus parameterization, Mon. Weather Rev., 121, 764 – 787, doi:10.1175/15200493(1993)1212.0.CO;2. Herdies, D. L., V. E. Kousky, and W. Ebisuzaki (2007), The impact of highresolution SALLJEX data on global NCEP analyses, J. Clim., 20, 5765 – 5783, doi:10.1175/2007JCLI1375.1. Holtslag, A. A. M., E. I. F. De Bruijn, and H.-L. Pan (1990), A high resolution air mass transformation model for short-range weather forecasting, Mon. Weather Rev., 118, 1561 – 1575, doi:10.1175/15200493(1990)1182.0.CO;2. Kain, J. S., and J. M. Fritsch (1990), A one-dimensional entraining/detraining plume model and its application in convective parameterization, J. Atmos. Sci., 47, 2784 – 2802, doi:10.1175/1520-0469(1990)047 2.0.CO;2. Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437 – 471, doi:10.1175/1520-0477(1996) 0772.0.CO;2. Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter (2002), NCEP-DOE AMIP-II Reanalysis (R2), Bull. Am. Meteorol. Soc., 83, 1631 – 1643, doi:10.1175/BAMS-8311-1631(2002)0832.3.CO;2. Kiehl, J. T., J. J. Bonan, B. A. Boville, B. P. Briegleb, D. L. Williamson, and P. J. Rasch (1996), Description of the NCAR Community Climate Model (CCM3), NCAR Tech. Note, NCAR/TN-420+STR, 143 pp. Kodama, Y. (1992), Large-scale common features of subtropical precipitation zones (the Baiu frontal zone, the SPCZ, and the SACZ). part I: Characteristics of subtropical frontal zones, J. Meteorol. Soc. Jpn., 70, 813 – 836. Kousky, E. V. (1980), Diurnal rainfall variation in northeast Brazil, Mon. Weather Rev., 108, 488 – 498, doi:10.1175/1520-0493(1980)108 2.0.CO;2. Kummerow, C., W. Barnes, T. Kozu, J. Shiu, and J. Simpson (1998), The tropical rainfall measuring mission (TRMM) sensor package, J. Atmos. Oceanic Technol., 15, 809 – 817, doi:10.1175/1520-0426(1998)015 2.0.CO;2. Lenters, J. D., and K. H. Cook (1997), On the origin of the Bolivian high and related circulation features of the South American climate, J. Atmos. Sci., 54, 656 – 677, doi:10.1175/1520-0469(1997)0542.0.CO;2. Liang, X.-Z., L. Li, A. Dai, and K. E. Kunkel (2004), Regional climate model simulation of summer precipitation diurnal cycle over the United States, Geophys. Res. Lett., 31, L24208, doi:10.1029/2004GL021054. Lieberman, R. S., C. B. Leovy, B. A. Boville, and B. P. Briegleb (1994), Diurnal heating and cloudiness in the NCAR Community Climate Model (CCM2), J. Clim., 7, 869 – 889, doi:10.1175/1520-0442(1994) 0072.0.CO;2. Liebmann, B., G. N. Kilads, J. A. Marengo, T. Ambrizzi, and J. D. Glick (1999), Submonthly convective variability over South America and South Atlantic Convergence Zone, J. Clim., 12, 1877 – 1891, doi:10.1175/1520-0442(1999)0122.0.CO;2. Liebmann, B., G. N. Kiladis, C. S. Vera, C. Saulo, and L. M. V. Carvalho (2004), Subseasonal variations of rainfall in South America in the vicinity of the low-level jet east of the Andes and comparison to those in the South Atlantic Convergence Zone, J. Clim., 17, 3829 – 3842, doi:10.1175/1520-0442(2004)0172.0.CO;2. Lin, X., D. A. Randall, and L. D. Fowler (2000), Diurnal variability of the hydrologic cycle and radiative fluxes: Comparisons between observations and a GCM, J. Clim., 13, 4159 – 4179, doi:10.1175/1520-0442 (2000)0132.0.CO;2. Loveland, T. R., B. C. Reed, J. F. Brown, D. O. Ohlen, J. Zhu, L. Yang, and J. W. Merchant (2000), Development of a global land cover characteristics database and IGBP DISCover from 1-km AVHRR data, Int. J. Remote Sens., 21, 1303 – 1330, doi:10.1080/014311600210191. Ma, H. Y., and R. Mechoso (2007), Submonthly variability in the South American monsoon system, J. Meteorol. Soc. Jpn., 85A, 381 – 401. Marengo, J., W. R. Soares, C. Soulo, and M. Nicolini (2004), Climatology of the low-level jet east of the Andes as derived from the NCEP-NCAR reanalyses: Characteristics and temporal variability, J. Clim., 17, 2261 – 2280, doi:10.1175/1520-0442(2004)0172.0.CO;2. Martıńez-Castro, D., R. P. da Rocha, A. Bezanilla-Morlot, L. AlvarezEscudero, J. P. Reyes-Fernańdez, Y. Silva-Vidal, and R. W. Arritt (2006), Sensitivity studies of the RegCM-3 simulation of summer precipitation, temperature and local wind field in the Caribbean Region, Theor. Appl. Climatol., 86, 5 – 22, doi:10.1007/s00704-005-0201-9. Misra, V., P. A. Dirmeyer, B. P. Kirtman, H.-M. H. Juang, and M. Kanamitsu (2002), Regional simulation of inter-annual variability over South America, J. Geophys. Res., 107(D20), 8036, doi:10.1029/2001JD900216. Misra, V., P. A. Dirmeyer, and B. P. Kirtman (2003), Dynamic downscaling of seasonal simulations over South America, J. Clim., 16, 103 – 117, doi:10.1175/1520-0442(2003)0162.0.CO;2.

18 of 19

D10108


Mitchell, T. D., and P. D. Jones (2005), An improved method of constructing a database of monthly climate observations and associated highresolution grids, Int. J. Climatol., 25, 693 – 712, doi:10.1002/joc.1181. Negri, A. J., T. L. Bell, and L. Xu (2002), Sampling of the diurnal cycle of precipitation using TRMM, J. Atmos. Oceanic Technol., 19, 1333 – 1344, doi:10.1175/1520-0426(2002)0192.0.CO;2. Nesbitt, S. W., and E. J. Zipser (2003), The diurnal cycle of rainfall and convective intensity according to three years of TRMM measurements, J. Clim., 16, 1456 – 1475. New, M., H. Mike, and J. Phil (1999), Representing twentieth-century space-time climate variability. part I: Development of a 1961 – 90 mean monthly terrestrial climatology, J. Clim., 12, 829 – 856, doi:10.1175/ 1520-0442(1999)0122.0.CO;2. Nicolini, M., and A. C. Saulo (2006), Modeled Chaco low-level jets and related precipitation patterns during the 1997 – 1998 warm season, Meteorol. Atmos. Phys., 94, 129 – 143, doi:10.1007/s00703-006-0186-7. Nicolini, M., P. Salio, J. J. Katzfey, J. L. McGregor, and A. C. Saulo (2002), January and July regional climate simulation over South America, J. Geophys. Res., 107(D22), 4637, doi:10.1029/2001JD000736. Oliveira, A. P., R. Bornstein, and J. Soares (2003), Annual and diurnal wind patterns in the city of Saõ Paulo, Water Air Soil Pollut., 3, 3 – 15, doi:10.1023/A:1026090103764. Pal, J. S., E. E. Small, and E. A. B. Eltahir (2000), Simulation of regional scale water and energy budgets: Influence of a new moist physics scheme within RegCM, J. Geophys. Res., 105, 29,579 – 29,594, doi:10.1029/ 2000JD900415. Pal, J. S., et al. (2007), Regional climate modeling for the developing world: The ICTP RegCM3 and RegCNET, Bull. Am. Meteorol. Soc., 88, 1395 – 1409, doi:10.1175/BAMS-88-9-1395. Randall, D. A., and D. A. Dazlich (1991), Diurnal variability of the hydrologic cycle in a general circulation model, J. Atmos. Sci., 48, 40 – 62, doi:10.1175/1520-0469(1991)0482.0.CO;2. Rauscher, S. A., A. Seth, B. Liebmann, J.-H. Qian, and S. J. Camargo (2007), Regional climate model – simulated timing and character of seasonal rains in South America, Mon. Weather Rev., 135, 2642 – 2657, doi:10.1175/MWR3424.1. Reynolds, R. W., and T. M. Smith (1995), A high-resolution global sea surface temperature climatology, J. Clim., 8, 1571 – 1583, doi:10.1175/ 1520-0442(1995)0082.0.CO;2. Rickenbach, T. M., R. N. Ferreira, J. Halverson, D. L. Herdies, and M. A. F. Silva Dias (2002), Modulation of convection in the southwestern Amazon basin by extratropical stationary fronts, J. Geophys. Res., 107(D20), 8040, doi:10.1029/2000JD000263. Roads, J., S. Chen, S. Cocke, L. Druyan, M. Fulakeza, T. LaRow, P. Lonergan, J.-H. Qian, and S. Zebiak (2003), International Research Institute/Applied Research Centers (IRI/ARCs) regional model intercom-

D10108

parison over South America, J. Geophys. Res., 108(D14), 4425, doi:10.1029/2002JD003201. Rojas, M., and A. Seth (2003), Simulation and sensitivity in nested modeling system for South America. part II: GCM boundary forcing, J. Clim., 16, 2454 – 2471, doi:10.1175/1520-0442(2003)0162.0. CO;2. Salio, P., M. Nicolini, and E. J. Zipser (2007), Mesoscale convective systems over southeastern South American low-level jet, Mon. Weather Rev., 135, 1290 – 1309, doi:10.1175/MWR3305.1. Seth, A., and M. Rojas (2003), Simulation and sensitivity in nested modeling system for South America. part I: Reanalyses boundary forcing, J. Clim., 16, 2437 – 2453, doi:10.1175/1520-0442(2003)0162.0.CO;2. Seth, A., S. A. Rauscher, S. J. Camargo, J.-H. Qian, and J. S. Pal (2007), RegCM3 regional climatologies for South America using reanalysis and ECHAM global model driving fields, Clim. Dyn., 28, 461 – 480, doi:10.1007/s00382-006-0191-z. Silva, V. B. S., V. E. Kousky, W. Shi, and R. W. Higgins (2007), An improved historical daily precipitation analysis for Brazil, J. Hydrometeorol., 8, 847 – 861, doi:10.1175/JHM598.1. Silva Dias, M. A., P. L. Vidale, and C. M. R. Blanco (1995), Case study and numerical simulation of the summer regional circulation in Saõ Paulo, Brazil, Boundary Layer Meteorol., 74, 371 – 388, doi:10.1007/ BF00712378. Silva Dias, M. A., et al. (2002), Cloud and rain processes in a biosphereatmosphere interaction context in the Amazon region, J. Geophys. Res., 107(D20), 8072, doi:10.1029/2001JD000335. Tadross, M. A., W. J. Gutowski, B. C. Hewison, C. Jack, and M. New (2006), MM5 simulation of interannual change and the diurnal cycle of southern Africa regional climate, Theor. Appl. Climatol., 86, 63 – 80, doi:10.1007/s00704-005-0208-2. Uppala, S. M., et al. (2005), The ERA-40 re-analysis, Q. J. R. Meteorol. Soc., 131, 2961 – 3012, doi:10.1256/qj.04.176. Velasco, I., and J. M. Fritsch (1987), Mesoscale convective complexes in the Americas, J. Geophys. Res., 92, 9591 – 9613, doi:10.1029/ JD092iD08p09591. Vidale, P. L., D. Lu¨thi, C. Frei, S. I. Seneviratne, and C. Scha¨r (2003), Predictability and uncertainty in a regional climate model, J. Geophys. Res., 108(D18), 4586, doi:10.1029/2002JD002810. T. Ambrizzi, S. V. Cuadra, R. P. da Rocha, and C. A. Morales, Departamento de Cieˆncias Atmosfe´ricas, Instituto de Astronomia, Geofı´sica e Cieˆncias Atmosfe´ricas, Universidade de Saõ Paulo, Saõ Paulo 05508-090, Brazil. ([email protected])

19 of 19