As the tropical regions are intense heat sources for the atmosphere, the convective activity annual cycle of such regions has been studied through the outgoing longwave radiation (OLR) data in a number of papers (Janowiak et al., 1985; Horel et al., 1989; Mitchell and Wallace, 1992; Wang, 1994; Hastenrath, 1997; Matsumoto and Murakami, 2000). Partitioning the OLR into symmetric and antisymmetric (AOLR) components in relation to the equator, Murakami and Nakazawa (1985) found that the AOLR component mainly reflects the seasonal alternation between the Northern and Southern hemisphere (NH and SH) summers, whereas the symmetric component does not show the monsoon character. Thus, these results suggest that the South American Monsoon System (SAMS) convection features might be better identified in the AOLR component.
Since the SAMS cycle evolves in response to the seasonal changes of the continent–ocean thermal gradient in low latitudes and is the major system modulating the summer rainfall over tropical South America (Vera et al., 2006), a variety of methods have been used to identify the onset and demise dates of the rainy season (ONR and DER, respectively). Using the pentad OLR data over South America, Kousky (1988) defined the ONR (DER) pentad when the OLR is lower (greater) than 240 W m−2. But in relation to this pentad, at least 10 of the 12 preceding pentads have OLR greater (lower) than 240 W m−2, and 10 of the 12 following pentads have OLR lower (greater) than 240 W m−2. He found that the ONR progresses from northwest to southeast, while the DER progresses in the opposite direction. In this context, several authors found that the SAMS-related convective band and precipitation progress southward rapidly and abruptly during the SAMS onset, and gradually following the Sun's seasonal march during the SAMS demise (Horel et al., 1989; Marengo et al., 2001; Tanimoto et al., 2010).
Using precipitation and 850 hPa zonal wind data in Western-Central Brazil (WCB), Gan et al. (2004) identified the ONR and DER dates for this region. According to them, the higher variability of the ONR dates compared to the DER dates might be due to the influences of the transient systems in starting and organising the convection during the rainy period. They confirmed that the convection associated with the SAMS is organized in a northwest-southeast (NW-SE) orientation as noted in other papers that used different methods to determine the ONR and DER dates (Horel et al., 1989; Marengo et al., 2001). In addition, Gan et al. (2006) using four monsoon indices, including that used by Gan et al. (2004), determined the ONR and DER dates for the WCB region. These indices were based on the dynamical aspects of the atmosphere (low and upper level winds and vertical wind shear). They compared the ONR and DER dates obtained from these indices with those determined from an index based only on the precipitation data. They concluded that the index also used by Gan et al. (2004) is the one that gives a better representation of these dates compared to the other three indices.
Furthermore, Silva and Carvalho (2007) proposed a large-scale index for the South American Monsoon (LISAM) to determine the ONR and DER dates. This index is the time series that gives the temporal variations of the first combined empirical orthogonal function of the anomalies of the precipitation, specific humidity, 850 hPa zonal and meridional winds, and 850 hPa air temperature in the region bounded at the equator, 40°S, 65°W and 20°W. They defined the ONR (DER) date when the three-pentad running mean becomes positive (negative). The small differences between the dates obtained using LISAM, and by Gan et al. (2004), are likely due to the fact that the region used to compute LISAM is not constrained to the WCB region.
It is noteworthy that most of the papers use threshold values of a given variable to detect the ONR dates (Kousky, 1988; Marengo et al., 2001; Gan et al., 2004; Gan et al., 2006). However, Marengo et al. (2001) stated that the ONR dates are dependent on the threshold chosen. Indeed, they showed that the southeastward progression of the onset is reversed when the threshold value is doubled from 4 mm day−1 to 8 mm day−1. In this context, detailed analyses on the ONR date detection methods developed by Kousky (1988), Marengo et al. (2001), and Liebmann and Marengo (2001) were done by González et al. (2007). Liebmann and Marengo (2001) obtained the ONR and DER dates based on the evolution of local precipitation accumulated in the SAMS region. González et al. (2007) concluded that the method based on the OLR developed by Kousky (1988) represents better the ONR date in continental scale because those based on the precipitation use threshold values, making them regionally dependent.
Garcia and Kayano (2009) (hereinafter referred to as GK09) developed a new method to determine the ONR dates for the SAMS regions which is not based on threshold values. Following Murakami and Nakazawa (1985) who noted that the monsoon systems can be found in the AOLR component, GK09 established the AOLR signal change from positive (reduced convection) to negative (enhanced convection) as the criterion to define the ONR date in the Central Amazon (CAM) area. Using the composites of some variables, they documented physically coherent evolving dynamic and thermodynamic features during the ONR period. They also used the Tropical Rainfall Measuring Mission (TRMM) precipitation data and showed that this method gives a reasonable indication of when wet conditions are established in the CAM area. It is worth noting that Silva and Carvalho's (2007) method does not use threshold values. However, they use a more complex technique and a large number of variables, including the precipitation, which is not available in near-real time. Thus, considering this aspect, the method developed by GK09 is more advantageous.
The antisymmetric area to the CAM relative to the equator is situated within the SAMS area (GK09). However, the antisymmetric area to the WCB relative to the equator is localized in the North Atlantic Ocean not within the North American monsoon system (NAMS) area. Furthermore, the NAMS onset starts at 98°W and propagates northwestward (Vera et al., 2006). In addition, Matsumoto and Murakami (2002) found that the trajectory of the convective activity of the American monsoon has the NW-SE orientation during the autumn and spring seasons. Thus, for the WCB, the GK09 method has to be adjusted to the NW-SE orientation of the monsoon-related convection displacement, which follows the orientation of the American continents.
The present paper examines the method developed by GK09 for determining the ONR dates for the WCB area. The results are compared to onset dates for the WCB area obtained by Gan et al. (2006) who used only the precipitation averaged over this area. The data and methodology are described in Section 2. The results of the AOLR time series to detect the ONR dates for the WCB area and the precipitation fields are presented and discussed in Section 3. Conclusions are given in Section 4.
2. Data and methodology
Daily OLR records used in this work are obtained from the National Oceanic and Atmospheric Administration's (NOAA) polar-orbiting satellites. This data is used as a proxy for deep convection in the tropical South American region and has a 2.5° latitude and longitude horizontal resolution with interpolated missing values (Liebmann and Smith, 1996). The OLR data is available at http://www.cdc.noaa.gov and is obtained for the 1979–2008 period in the boxes illustrated in Figure 1. These boxes are bounded at 20°S, 10°S, 60°W, and 50°W for the WCB area, and bounded at 20°N, 10°N, 107.5°W, and 97.5°W identified as NW-SE.
TRMM-based 3-hourly precipitation estimates are also used. This data is selected for the area limited at 40°S, 30°N, 110°W and 30°W for the period from 14 August 2007 to 16 December 2008. The estimated precipitation data are in a grid with 0.25° resolution in latitude and longitude, and are available at http://disc.sci.gsfc.nasa.gov/data/datapool/TRMM/. The dataset used is the 3B42. The combined instrument rain calibration algorithm (3B42) uses an optimal combination of 2B-31, 2A-12, SSMI, AMSR and AMSU precipitation estimates, to adjust infrared (IR) estimates from geostationary IR observations. For more details about the algorithm, go to http://trmm.gsfc.nasa.gov/3b42.html; the description of the TRMM data can be found in Huffman et al. (2007).
The influences of the high-frequency transients associated with individual weather events are removed by using the five-day (pentad) non-overlapping means for the OLR data. According to Murakami and Nakazawa (1985), the AOLR is obtained as half the difference between OLR averaged in the southern box and in the antisymmetric northern box. Here, the NW-SE box illustrated in Figure 1 is the antisymmetric box to the WCB area. The AOLR obtained from the WCB and the NW-SE box is given by
where x is the longitude and y is the latitudinal distance to the equator, and y = 0 at the equator. The difference in Equation (1) is southern minus northern values. This calculation is referred to as the NW-SE method. As in GK09, the AOLR is obtained in each grid point in the WCB area for the 1979–2005 period.
The ONR dates for the WCB determined by Gan et al. (2006) using the precipitation data are considered the reference dates, and the mean ONR pentad of these dates, the central pentad of the ONR period. Thus, the ONR period spans from pentad 46 to 70. Twenty-seven ONR periods are taken for the years from 1979 to 2005. Spatial averages of the AOLR over the WCB for the ONR periods are obtained. The AOLR sign changing from positive to negative indicates the transition date from dry to wet season. Comparisons of the NW-SE method and Gan et al. (2006) precipitation index are made using lag composites of OLR for the − 4, − 3, − 2, − 1, 0 and + 1 pentads, with the pentad 0 corresponding to ONR pentad. These composites are done separately for the NW-SE and Gan et al. (2006) ONR dates. In order to accentuate the differences, only the years in which the ONR dates differ between the two analyses are included in the composites.
In order to examine the practical application of the method proposed here, the TRMM precipitation data accumulated every 5 non-overlapping days from pentad 46 to 70 of the years 2007 and 2008 are also obtained. The AOLR2007 and AOLR2008 time series using the NW-SE method are also calculated.
3.1. AOLR time series
The ONR date in the WCB region is determined through the AOLR time series from pentad 46 to 70 of each year of the 1979–2005 period (Figure 2). Positive (negative) AOLR values represent reduced (enhanced) convection in the WCB, and enhanced (reduced) convection in the antisymmetric area. The ONR date might be inferred from the AOLR time series, as the first pentad with negative AOLR that interrupts the sequence of dominantly positive AOLR values. Using this criterion, the ONR dates for the WCB are obtained and listed in Table I. For sake of comparison, this table also lists the ONR dates obtained by Gan et al. (2006).
Table I. ONR pentads of the WCB obtained from the NW-SE method and by Gan et al. (2006)
The transitions from dry to wet season, given by the sign alternations from positive to negative, occur in dates varying within the period of study (Figure 2). These dates correspond to those when the monsoon rainy season is triggered in the WCB. The sign alternation from positive to negative occurs almost as definitive and, in most cases, no positive AOLR values occur afterwards (Figure 2). The mean ONR date in the WCB obtained here, pentad 57, is closer to that obtained by Gan et al. (2006) (pentad 58), who used only the precipitation data than to that obtained by Kousky (1988), who used the OLR data (pentad 54).
Previous papers did not find a relationship between the El Niño–Southern Oscillation (ENSO) and the ONR dates in the WCB area (Marengo et al., 2001; Gan et al., 2004). However, the ONR dates obtained here show a relationship with the El Niño events. Except for the El Niño years of 1982–1983 and 1987–1988, the ONR date during the other El Niño years of the period of analysis (1991–1992, 1992–1993, 1994–1995, 1997–1998, and 2002–2003) is delayed in relation to the mean ONR date. In other words, considering 7 El Niño years, the ONR date was delayed in 5 of them. The few La Niña events during the period of analysis have not allowed us to establish any relationship between the ONR dates and these events. Indeed, a delayed ONR occurred during the 1998–1999 La Niña, while ONR was observed in the mean pentad for the 1988–1989 and 2000–2001 La Niña years.
On average, the ONR date found here, and that obtained by Gan et al. (2006), are close, however, they differ in some individual years. These years correspond to the rainy seasons from 1979–1980 to 1992–1993. In these rainy seasons, lag composites of OLR for the − 4, − 3, − 2, − 1, 0 and + 1 pentads are obtained independently for the ONR dates found here (Figure 3) and those found by Gan et al. (2006) (Figure 4).
In Figure 3, the convective activity, indicated by OLR lower than 240 W m−2, are observed in Central America, northwestern Amazon, and along the Intertropical Convergence Zone (ITCZ) over the southeastern North Pacific and North Atlantic Oceans, during pentad − 4. Gradually, the convective activity intensifies and moves southeastward from pentad − 4 to − 1. At the ONR pentad, the convective activity intensifies in the central part of the continent and OLR lower than 220 W m−2 are noted in the WCB region. At this pentad, a convective band NW-SE oriented extends from the Amazon region towards southeastern South America and the adjacent Atlantic Ocean which features the South Atlantic Convergence Zone (SACZ) previously documented by Kodama (1992, 1993). At the following pentad, a well-established and enhanced convection remains in the study region.
Figure 4 shows similar evolving OLR patterns. However, some important differences are noted. The OLR values lower than 240 W m−2 in WCB, and even lower than 230 W m−2 are observed in part of the region at pentad − 1. Therefore, intense convective activity occurs in the study area one pentad before the ONR date. During the two subsequent pentads, the convective activity intensifies and extends southeastward establishing the SACZ pattern. In contrast, the establishment of the intense convective activity in the WCB occurs only at pentad 0 for the dates of the NW-SE method (Figure 3). Therefore, these composites give more consistent results than the composites using Gan et al.'s (2006) dates (Figures 3 and 4).
Since the present work aims to show that the method developed by GK09 and used here for WCB has advantages over other methods, it would be interesting to apply it to regions where other methods have not succeed in determining the ONR dates. For instance, the northwestern Amazon where Kousky's (1988) method failed because the OLR in that region is normally lower than the threshold value of the method. Nevertheless, the method using the AOLR also cannot be applied because this region is between 2.5°N and 2.5°S, and at the equator the AOLR is null.
In order to confirm that the NW-SE orientation of the monsoon convection displacement should be considered when choosing the antisymmetric area, the spatial averages of the AOLR from pentad 46 to 70 of 2007 and 2008 (AOLR2007 and AOLR2008) in the WCB are calculated with the NW-SE method and compared with the TRMM-based precipitation fields for these pentads.
3.2. AOLR2007, AOLR2008 and TRMM maps
The AOLR2007 time series during the ONR period of 2007 for the WCB area shows that the ONR date is pentad 59 (Figure 5). Figure 6 shows maps of TRMM-based precipitation from pentad 55 to 64 of 2007. At pentad 55, precipitation above 6 mm/pentad occurs in the northwest Amazon, along the ITCZ over the eastern Pacific and Atlantic Oceans, and in Central America. Coherently, the positive AOLR2007 in this pentad indicates convection activity in the antisymmetric area to the WCB. During the two subsequent pentads, the rainfall is enhanced in the Amazon Basin as well as in Central America. Consistently, the AOLR2007 remains positive, but its magnitude decreases at pentad 56 and increases at pentad 57. This indicates that dry conditions remain in the WCB area, but wetting starts there. At pentad 58, the AOLR2007 is still positive, but its magnitude, lower than 5 W m−2 (Figure 5), is coherent with reduced (enhanced) convective activity in the NAMS (WCB) area. Precipitation above 6 mm/pentad is found in most of the area extending from the northwest Amazon to southeast South America, including the WCB region at pentad 59 (ONR). This is consistent with the small magnitude negative AOLR2007 values. Gradually, a well-established SACZ pattern is configured, which is characteristic of the SAMS region (Kousky, 1988; Horel et al., 1989; Marengo et al., 2001; Gan et al., 2004; GK09). Indeed, the highest precipitation in the Amazon Basin with values above 6 mm/pentad extends southeastward over southeast South America in pentads 63 and 64. Rainfall is absent in the NAMS area from pentad 62 to 64. Coherently, the negative AOLR2007 values smaller than − 20 W m−2 (Figure 5) are indicative of the wet conditions in the WCB area from pentad 59 to 64.
As in 2007, the ONR date in 2008 is pentad 59 (Figure 7). Maps of TRMM-based precipitation from pentad 55 to 64 of 2008 are shown in Figure 8. The AOLR2008 shows sign and/or magnitude variations which are consistent with convection variations in the NAMS and WCB areas (Figures 7 and 8). The positive AOLR2008 at pentad 55 is consistent with precipitation above 6 mm/pentad noted in the northwest sector of South America, along the ITCZ over the eastern Pacific and Atlantic Oceans, and in Central America (Figures 7 and 8). During the three subsequent pentads, the gradual rainfall reduction over South America as well as in the NAMS region, is consistent with reduction in the magnitude of the positive AOLR2008 (Figures 7 and 8). The negative AOLR2008 at pentad 59 is consistent with precipitation above 6 mm/pentad extending over most of the area from the northwest Amazon to southeast South America, including the WCB region (Figures 7 and 8). The precipitation reduction noted in the WCB area at pentad 60 is also indicated in the AOLR2008 which shows positive value at this pentad. During the four subsequent pentads, precipitation above 6 mm/pentad extends from the Amazon Basin to southeast South America, resulting in a well-established SACZ pattern, characteristic of the SAMS region (Kousky, 1988; Horel et al., 1989; Marengo et al., 2001; Gan et al., 2004; GK09), while rainfall is absent in the NAMS area. For these pentads, the AOLR2008 shows negative values (most of them smaller than − 25 W m−2). Thus, wet conditions prevail in the WCB area from pentad 59 to 64.
Hence, the analyses of these two years confirm that the AOLR time series obtained from the NW-SE method is useful in determining the establishment of wet (dry) conditions in the WCB (NAMS) area. Thus, our method can identify reductions and enhancements in monsoon-related convection.
A new methodology developed previously to determine the ONR in the SAMS is based on the antisymmetric in relation to the equator pentad outgoing AOLR. This method is adjusted and used here to determine the ONR dates for the 1979–2005 period in WCB, the area is limited at 10°S, 20°S, 60°W and 50°W. The antisymmetric area to the WCB is bounded at 20°N, 10°N, 107.5°W and 97.5°W, and is located 47.5° west of the WCB area. This adjustment is needed to take into account the NW-SE orientation of the American monsoon-related convection displacement. The ONR date is indicated by the sign change from positive to negative of the spatial averages of the AOLR during the ONR period, the interval from pentad 46 to 70.
Concerning the mean ONR date, Gan et al. (2006), who used precipitation data, got the pentad 58; Kousky (1988), who used the OLR data, the pentad 54; and here we found the pentad 57. Kousky's (1988) and our method are both based on OLR data. The difference is that Kousky (1988) used threshold values for OLR and we used AOLR. The fact that our result is better than that of Kousky (1988) is because his method is based on local convection information. On the other hand, the AOLR combines convection information over the northern and southern boxes, therefore, it contains information on the seasonal see-saw of the American monsoon convection between the Northern and Southern Hemispheres. The seasonal monsoon convection see-saw might also involve the meridional displacement of the local Hadley circulation because its ascending branch organizes the convective activity in large portions of the tropics. Thus, our method indirectly includes some dynamics involved in this see-saw.
One of the main advantages of the method presented here in relation to the index based on the precipitation (Gan et al., 2006) is due to the fact that the OLR data can easily be obtained, and the AOLR calculation is very simple. Concerning the mean ONR dates, as discussed above, these methods have equivalent results. It is, however, interesting to discuss if relevant differences arise for individual years. These differences are analysed using lag composites of OLR during the ONR periods of the years when the ONR dates determined here and by Gan et al. (2006) are different. The main difference is that the intense convective activity in the WCB is established at the ONR date for our method and one pentad earlier for the Gan et al. (2006) method. This indicates a small advantage for our method. However, it should not be considered a relevant advantage because one pentad lead or lag for the ONR date might be within the data observation error. It is worth noting that the difference of the ONR dates for individual years is not small. Comparing individual years, this difference varies from − 7 (for 1979) to + 7 (for 1992).
Analyses of the AOLR for the ONR period of individual years (2007 and 2008) together with the TRMM-based precipitation maps show that AOLR time series is useful not only to identify the ONR dates but also as a measure of the convection fluctuations in the study area.
The method to identify the ONR dates based on the AOLR has great potential for monitoring purposes because the OLR data can easily be obtained, the AOLR calculation is very simple, and produces results equivalent to the method based on precipitation data. It is, however, noteworthy that this method should be tested in other areas of the SAMS, because the present study is limited to only one area.
The authors were partially supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico of Brazil. This work is part of the first author's doctoral thesis. Thanks are due to two anonymous reviewers for their useful comments on an earlier version of this work.