1. Introduction

The South American continent has a predominantly meridional shape from 12°N to 55°S. It is bathed by the tropical and extratropical waters of the Pacific and Atlantic Oceans, and has a wide variety of landscapes due to the interaction between geological, climatological, and biological processes. The complexity of its geological formation resulted in intercalary mountain regions, as the Andes Cordillera, the Brazilian Shield, and the Guyanese Shield, and subsidence regions, as the Amazon Basin and the La Plata Basin. These geomorpholocial features are shown in Fig. 1. While the Amazon and the La Plata Basins are the two main and largest watersheds of South America, the interaction of the Andes Cordillera, Brazilian Shield, and Guyanese Shield with global circulations define the main pathways of moisture transport and precipitation regimes in this continent (Junquas et al. 2016). Between these two watersheds, the moisture transport is predominantly from the Amazon Basin to the La Plata Basin, while the atmospheric perturbations propagate predominantly from the La Plata Basin to the Amazon Basin (Zanin and Satyamurty 2020a).

View Full Size

Topography of South America, watersheds analyzed, and their subdivisions. B1–B5 constitute the Amazon Basin. B6–B10 constitute the La Plata Basin. B11 is the Tocantins/Araguaia Basin.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

Around and during the austral summer season, the monsoon circulation and precipitation occur over South America (Zhou and Lau 1998; Vera et al. 2006; Marengo et al. 2012; Costa and Satyamurty 2016), called the South American monsoon (SAM). In this period the main mechanisms that transport moisture from the Amazon Basin to the La Plata Basin are the South Atlantic convergence zone (SACZ) and the low-level jet (LLJ) east of the Andes (Silva and Berbery 2006; Boers et al. 2014). The main source to these continental moisture exchange mechanisms during the monsoon season is the North Atlantic (Arraut and Satyamurty 2009; Drumond et al. 2014). Due to evapotranspiration in the Amazon Basin, the oceanic air mass maintains high moisture content even after precipitation along the watershed, supplying moisture to the precipitation in the La Plata Basin (Drumond et al. 2008; Spracklen et al. 2012; Zemp et al. 2014; Staal et al. 2018). In other seasons the SACZ activity disappears, while the LLJ east of the Andes occurs with more frequency (Berbery and Barros 2002; Nascimento et al. 2016; Montini et al. 2019). The LLJ constitutes a meridional aerial river over the South American continent (Arraut and Satyamurty 2009; Arraut et al. 2012) and affects the hydrological balance of the Amazon and the La Plata Basins in opposite ways (Nascimento et al. 2016). The main moisture sources to LLJ in the non-monsoon season are the South Atlantic, through the western branch of the subtropical high over the South Atlantic (SHSA), and the evapotranspiration from the southern Amazon Basin (Arraut et al. 2012). During the austral winter, most of the Amazon Basin receives smaller amount of precipitation (Villar et al. 2009a) and the evapotranspiration is maintained by water stored in the soils during the rainy season (Tomasella et al. 2008; Miguez-Macho and Fan 2012b). The capacity of soil moisture to remember the previous wet or dry seasonal or interannual condition is called hydrological memory (Koster and Suarez 2001; Tomasella et al. 2008; Guedes et al. 2013). According to Figs. 1 and 3 of Dirmeyer et al. (2009), the soil moisture in the southern portion of the Amazon Basin has significant positive correlation with evapotranspiration during winter season. In this season, the mean soil moisture memory in this portion of the basin calculated by these authors with two datasets is around two months. Tomasella et al. (2008) and Guedes et al. (2013) verified interannual effects of soil moisture memory on other surface hydrological processes in the Amazon Basin also.

Some studies quantify the contribution of Amazonian evapotranspiration to precipitation over the La Plata Basin. Martinez and Dominguez (2014) found a value of around 24% along the year. Figure 2c of Staal et al. (2018) shows annual values from ~5% to ~25%. Yang and Dominguez (2019) obtained an annual mean value of 16%, without a discernible seasonal cycle. According to Zemp et al. (2014), the contribution of the Amazonian evapotranspiration to precipitation over the La Plata Basin from December to March (June–September) is around of 18%–23% (21%–25%). Moreover, synthesizing the observations that (i) the largest frequency of LLJ is during the winter months (Nascimento et al. 2016; Montini et al. 2019), (ii) the discharge of aerial river east of the Andes over the La Plata Basin (around 10–23 Gt day⁻¹ during July–August) is comparable to the discharge of the Amazon River into the Atlantic Ocean (Arraut et al. 2012), and (iii) the largest moisture contribution from Amazon to the La Plata Basin occurs during the relatively dry season (Zemp et al. 2014), it is possible to conclude that the evapotranspiration of the Amazon forest has a key role in precipitation over the La Plata Basin during austral winter.

In their critical review, Zanin and Satyamurty (2020a,b) synthesized the hydrological processes that connect the Amazon and the La Plata watersheds in different time scales. A synthesis is shown in a conceptual flowchart of interseasonal and interbasins connections among the components of hydrological system in the southern portion of the Amazon Basin and the subtropical portion of the La Plata Basin (see Fig. 9 of Zanin and Satyamurty 2020a). However, this synthesis based on many isolated papers, needs verification with observational datasets. Zanin and Satyamurty (2020a,b) also formulated many hypotheses for future investigations. Some of these hypotheses are about the interseasonal and interbasins hydrological coupling in South America:

1. During non-SAM season, the hydrological memory of the southern Amazon Basin, due to water storage in deep soils by monsoon rains, plays a key role in the amount of moisture supply to the subtropical portion of the La Plata Basin.

2. During non-SAM season, the hydrological memory in the Pantanal region has also an important effect on the moisture supply to the subtropical La Plata Basin by the aerial river east of the Andes.

3. The intensity of the SAM over the Amazon Basin can be considered as an indicator of rainfall over the subtropical La Plata Basin during post-monsoon season.

Thus, this article has the main objective of verifying with observational data, the conceptual interseasonal and interbasins coupling between the Amazon and La Plata watersheds, and the hypotheses stated above. For this purpose it is first needed to create links between moisture source regions and moisture sink regions within the study area in South America. Then, the mechanisms of coupling, feedbacks, and memory [see these concepts in Zanin and Satyamurty (2020a)] related to hydrological processes between the two basins can be analyzed. This article is divided into six sections. In section 2, the material and methods are described. In section 3, an analysis of each region of the study area is made. In section 4, the results about moisture transport and relationship between regions of source and sink of moisture are shown and discussed. Section 5 analyzes the memory and the coupling between source and sink regions. Last, in section 6 the main conclusions are presented.

2. Material and method

b. Methods

To answer the scientific questions, this study needs to analyze large regions of the Amazon and the La Plata watersheds over long periods of time. Moreover, it is also necessary to analyze interseasonal relationships in space. Therefore, a short review of the methods to track atmospheric moisture and establishment of the source–sink relationships (Supplementary Material 1) is made in the online supplemental material at the Journals Online website: https://doi.org/10.1175/JHM-D-20-0080.s1. In this review the main characteristics, limitations, and advantages of each category of existent methods are shown.

Based on a review in Supplementary Material 1, and aiming to track the moisture transport between different regions, in a way that easily allows combined analysis of the surface and atmospheric hydrological variables in different seasons, without demanding large computational processing, a new method derived from atmospheric branch of water balance equation is developed. This new method is a 2D Box Model approach and is named Articulated Box Analysis.

1) Regions analyzed

The whole area covered by the Amazon, La Plata, and the Tocantins/Araguaia watersheds is divided into 11 regions in the form of boxes, as shown in Fig. 1. Over the Amazon Basin there are five boxes. These boxes are distributed over the northeastern (B1), northwestern (B2), southwestern (B3), southeastern (B4), and southernmost (B5) regions of the basin. Boxes B1 and B2 (B3, B4, and B5) constitute the northern (southern) portion of this basin. Over the La Plata Basin there are five boxes. These boxes are distributed over the northernmost (B6), northeastern (B7), central-western (B8), southeastern (B9), and southernmost (B10) regions of the basin. Boxes B6 and B7 (B8, B9, and B10) constitute tropical (subtropical) portion of this basin. B11 covers the central-southern portion of the Tocantins/Araguaia Basin.

The subdivisions of the Amazon Basin are based on the following observations:

1. Villar et al. (2009a) verified that the precipitation seasonality over the Amazon Basin is not homogeneous. Differences exist mainly between northern and southern portions of this basin, and between the northwestern and northeastern regions.

2. da Rocha et al. (2009) verified that the seasonality of evapotranspiration is different in the Amazon forest transition and “Cerrado” biome in southeastern region of the basin from the seasonality of the tropical rain forest and semi-deciduous tropical forest in the northern portion and the southwestern region of the basin.

3. Zemp et al. (2014) verified that the moisture recycling in cascade occurs from east to west within the Amazon Basin.

4. The southernmost region has opposite hydrological behavior from the rest of the basin during different phases of the interannual variability of sea surface temperature in adjacent oceans (Ronchail et al. 2005; Villar et al. 2009b).

5. Marengo (2009) verified that a pressure gradient between the northern and southern portions of the Amazon Basin occurs, and it makes the northern portion a moisture source to southern portion, as was verified by Marengo (2005) for El Niño events.

6. Arraut et al. (2012) identified the southern portion of the Amazon Basin as a moisture source to the La Plata Basin during austral winter months.

7. Martinez and Dominguez (2014) identified the Amazon Basin as a moisture source to the La Plata Basin throughout the year, mainly the southern portion during the austral winter season.

8. The southwestern region of the Amazon Basin acts throughout the year as a direct source of moisture to the La Plata Basin, and in the wet season it also becomes an intermediary region between the moisture of the Amazon Basin as a whole and the La Plata Basin (Zemp et al. 2014).

9. According to Zemp et al. (2014), due to moisture cascade recycling, the moisture transport from the Amazon to the La Plata Basin increases along its trajectory.

The subdivisions of the La Plata Basin are based on the following observations:

1. The annual mean precipitation increases from west to east within the basin, with peak in the southeastern region (Berbery and Barros 2002; Su and Lettenmaier 2009).

2. The tropical and subtropical portions of the basin have different behaviors as moisture source–sink (Rodriguez and Cavalcanti 2006), because while the tropical portion becomes a source region, the subtropical portion becomes a moisture sink, and vice versa (Nascimento et al. 2016).

3. The evapotranspiration increases from subtropical to tropical portions of the basin (Su and Lettenmaier 2009; Doyle and Barros 2011).

4. The large marshland called “Pantanal,” where the highest values of the evapotranspiration in the La Plata Basin occur (Su and Lettenmaier 2009), is located in the northernmost region.

5. The tropical portion of the basin is influenced by the SACZ, while the subtropical portion is influenced by the LLJ (Robertson and Mechoso 2000; Boers et al. 2014).

6. In Fig. 6 of Montini et al. (2019), it is verified that the magnitude and spatial extent of moisture transport by the LLJ has seasonal variability over the subtropical portion of the La Plata Basin.

The Tocantins/Araguaia Basin is included in the study area because (i) Drumond et al. (2008) verified that central Brazil is a moisture source to the La Plata Basin throughout the year, especially in the autumn and winter seasons, and (ii) Martinez and Dominguez (2014) identified this basin as a moisture source to the La Plata Basin throughout the year.

2) Moisture transport

More detailed description about the method developed (Articulated Box Analysis) and used in this study is given in the supplemental material (Supplementary Material 2) at the Journals Online website: https://doi.org/10.1175/JHM-D-20-0080.s2. A brief description is given below. The atmospheric moisture balance in an individual box is given by Eq. (1):

$\hspace{0.17em}{T}_I-P_a+E_{\text{res}}=T_O\hspace{0.17em},$(1)

in which T_I is the inflow into the box, obtained over the borders of the box where water vapor flux is predominantly inward; T_O is outflow obtained over the borders where water vapor flux is predominantly outward; P_a is precipitation due to advection of moisture from the external source regions; and E_res is residual evapotranspiration, which is the moisture from evapotranspiration that does not return as precipitation in the same box, and is available to be transported to other boxes. It is important to mention that all the variables in Eq. (1) are always positive.

The articulated procedure starts when a given box is a moisture source. The source–sink characteristic of a box is determined by the difference between T_O and T_I. If T_O > T_I the region is a moisture source, and if T_O < T_I the region is a moisture sink. This definition is equivalent to the definition of Satyamurty et al. (2012) for moisture source or sink.

Here and in Fig. 2 (this figure is a schematic diagram to help explain the method, being applicable to any spatial configuration of a set of boxes), the boxes are denominated B_i, i = 1, 2, 3, …, 11. The part of E_res of B_i that is transported to its neighboring B_i+1 is represented by DT_i,i+1, and is called Direct Transport (DT). The direct transport occurs at the border between B_i and B_i+1. The border between B_i and B_i+1 is called L_i,i+1. The part of E_res of B_i that is transported to $B_{i+2}$, after traveling all the extension of B_i+1 without precipitating fully, is represented by IT_i,i+2 and is called Indirect Transport (IT). The indirect transport occurs at boundaries between B_i and B_i+1 (L_i,i+1) and between B_i+1 and B_i+2 (L_i+1,i+2). Also, indirect transport may occur over a large number of boxes.

View Full Size

Schematic diagram of Articulated Box Analysis. Boxes with internal fill (color) refer to the land surface and boxes without internal fill (no color) refer to the atmosphere. Dashed lines refer to flows, where the arrows indicate the destination, and the circles indicate the origin and also the division of path in two ways. Atl, moisture from the Atlantic Ocean; Patl, precipitation due to moisture from the Atlantic Ocean; Prec, precipitation recycled; E_res, evapotranspiration residual; PEE, precipitation due to external evapotranspiration; DT, direct transport; IT, indirect transport.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

To quantify direct transport the Internal Moisture Fraction (IMF), which refers to the ratio between E_res and T_O of a given box with moisture source behavior, is calculated first. Then, multiply the IMF of B_i (IMF_i) by transport of moisture into B_i+1 that occurs at L_i,i+1 (from B_i to B_i+1) represented by t_i,i+1 (moisture inward in B_i+1 across its border with neighboring B_i). Thus, it is possible to know how much of E_res from B_i is exported to B_i+1 (DT_i,i+1), according to Eq. (2). The direct transport is calculated in all boundaries between a given box with moisture source behavior and their neighboring boxes:

${\text{DT}}_{i,i+1}={\text{IMF}}_i\times t_{i,i+1}\hspace{0.17em}.$(2)

The amount of evapotranspiration in a given box that precipitates in another box is called Precipitation due to External Evapotranspiration (PEE). The PEE from B_i to B_i+1 (PEE_i,i+1) is obtained by Eq. (3):

${\text{PEE}}_{i,i+1}=P_{a_{i+1}}\times \left({\displaystyle \frac{{\text{DT}}_{i,i+1}}{T_{I_{i+1}}}}\right)\hspace{0.17em},$(3)

in which $P_{a_{i+1}}$ is the mean precipitation due to advection of moisture into B_i+1.

To quantify the indirect transport the principle of the method of Martinez and Dominguez (2014) to quantify moisture from different source regions to the atmospheric transport over a given sink region is used. This principle consists in obtaining the fraction of moisture from a source region to a given sink region that is not lost by precipitation during atmospheric transport (or precipitation over the intermediate regions). In the present study, Eq. (4) is used to obtain the indirect transport from B_i to B_i+2, designated IT_i,i+2. First, it is necessary to obtain the External Moisture (EM), which refers to the difference between T_I and the corresponding P_a in a given box. Then, PEE_i,i+1 is subtracted from DT_i,i+1 and then divided by EM of B_i+1 (EM_i+1). Last, multiplying this fraction by total moisture transported from B_i+1 to B_i+2 (t_i+1,i+2) it is possible to know the E_res from B_i entering B_i+2 (IT_i,i+2). It is important to mention that here the case of the L_i,i+1 and L_i+1,i+2 is presented. However, as shown in Fig. 2, this procedure can also be used on the borders L_i+1,i+2 and L_i+2,i+3, and so on. The indirect transport IT_i,i+2 is given by Eq. (4):

${\text{IT}}_{i,i+2}={\displaystyle \frac{{\text{DT}}_{i,i+1}-{\text{PEE}}_{i,i+1}}{{\text{EM}}_{i+1}}}\times t_{i+1,i+2}\hspace{0.17em}.$(4)

As shown in Fig. 2, the indirect transport may also occur over longer distances. Thus, the procedure used in Eqs. (3) and (4) will be extended for the case of transport of E_res from B_i traveling the extension of B_i+2 without precipitating fully.

PEE is considered significant if it is equal to or greater than twice the median value of the hydrological imbalance from the set of boxes. The hydrological imbalance of the dataset used is given in the supplemental material (Supplementary Material 3) at the Journals Online website: https://doi.org/10.1175/JHM-D-20-0080.s3, and the median hydrological imbalance as percentage of median precipitation from the set of boxes is 5%. Thus, the seasonal PEE equal to or greater than 10% of total seasonal precipitation is significant.

It is important to mention that this method only tracks moisture transport from the regions with divergence of moisture at monthly time scale (source region). Therefore, it tracks mainly the moisture evaporated from water bodies and transpired from trees by deep soil water uptake by roots, as commonly occurs during relatively dry season in the monsoon regions. Therefore, the direct and indirect transports from B_i can be somewhat underestimated during seasons with low frequency of becoming a source of moisture, because of the starting of the tracking restraint, while remaining very robust during seasons in which B_i has a high frequency of moisture source behavior (see Supplementary Material 2 for more details).

3) Memory and Coupling Analyses

This analysis is made with source–sink relationship between two boxes (B_i and B_i+n) in each month identified as described in the previous section. To verify the coupling between the components of surface and atmospheric water balances, the coefficient of correlation (r) according to Montgomery and Runger (2003) is used. The statistical significance of correlations is obtained by a two-tailed Student’s t test.

To verify the interseasonal and interbasins coupling between different boxes of the two basins discussed in Zanin and Satyamurty (2020a, see their Fig. 9), as well as the hypotheses of these authors stated in the introduction, two sets of variables shown in Table 1 are considered. To confirm the existence of the interseasonal and interbasins coupling, the correlations in the two sets of variables must be positive and greater than 0.40 and significant at 0.01.

Table 1.

Sets of variables correlated: P, precipitation; P*, lagged precipitation; SM, soil moisture; SM*, lagged soil moisture; E_res, evapotranspiration residual; X, moisture from evapotranspiration of a source region that reaches a sink region; Y, precipitation in sink region due to evapotranspiration from the source region.

View Table

Considering a given source–sink relationship during a given month m_j, j = 1, 2, 3, …, 180, positive and significant correlations in all pairs of variables of Set 1 show that, the lagged precipitation (P*) over a given source region B_i during month m_j−lag influences the soil moisture (SM*) during the month m_j−lag. The lagged soil moisture (SM*) remembers its dry or wet conditions until the month m_j (SM). The amount of water in the soil during the month m_j (SM) influences the amount of evapotranspiration that does not precipitate over its region of origin (E_res). This residual evapotranspirated moisture is transported downwind and reaches a given sink region B_i+n (X). In this sink region the moisture from the source region (X) contributes to a part of the precipitation in month m_j (Y).

The moisture from evapotranspiration of a source region that reaches a sink region, represented by symbol X, is obtained by DT and IT, described previously. The precipitation in a sink region due to evapotranspiration from the source region, represented by symbol Y, is PEE, also described previously. The lag of the variables P* and SM* is determined according to soil moisture memory of each source region. According to Dirmeyer et al. (2009), the soil moisture memory is measured as the average time required for the soil moisture autocorrelation function to fall below 99% confidence level. The autocorrelation function is obtained according to Wey (2006), and the statistical significance at 0.01 is obtained by a two-tailed Student’s t test. The autocorrelation function is calculated with daily data, and then the monthly value is obtained.

To validate the relationship of Set 1 of variables, it is also needed that correlation in the complementary set of variables (Set 2) shown in Table 1 must be positive and greater than 0.40 and significant at 0.01. The Set 2 of variables skip one pair of variables in the sequence of hydrological parameters represented by Set 1.

3. Individual box analysis

The precipitation recycling in the boxes during all months of 2003–17 (Supplementary Material 4: https://doi.org/10.1175/JHM-D-20-0080.s4), in the tropical region of the study area, which includes the Amazon Basin [with the exception of the southernmost region (B5)] and the Tocantins/Araguaia Basin (B11), increases from east to west, in agreement with Eltahir and Bras (1994), and from north to south, in agreement with Martinez and Dominguez (2014). Moreover, the recycling values obtained here in all regions of the Amazon Basin vary from 15% to 25% and are near the recycling values for the whole Amazon Basin found in recent studies (20%–35%) shown in the review of Rocha et al. (2015). It is known that the precipitation recycling ratio obtained by box models are influenced by size of study region (van der Ent et al. 2010; van der Ent and Savejine 2011). In this study the linear correlation between the areas of the 11 boxes analyzed and their precipitations recycled is 0.64. Thus, the regional values obtained in this study agree with values for the whole watershed shown in the review of Rocha et al. (2015).

The values of precipitation recycling over the La Plata Basin (12%–21%) do not show a definite spatial pattern (Supplementary Material 4). Possibly this is due to spatial definition of boxes in this basin. The values over the La Plata Basin obtained here are larger than the recycling values of 0%–7% obtained by Su and Lettenmaier (2009) and agree with recycling values of 10%–20% obtained by Trenberth (1999). On the other hand, they are lower than the value of 23.5% for the whole La Plata Basin obtained by Martinez and Dominguez (2014). In the Tocantins/Araguaia Basin the precipitation recycling is 14%.

Figure 3 shows the seasonality of precipitation recycling in the study area. It is verified that the peak of recycling has northward propagation from summer to spring seasons. This shows an inverse relation between monsoon rain cycle over South America (Vera et al. 2006) and precipitation recycling over the Amazon Basin. On the other hand, over the La Plata Basin the values of recycling are higher in summer season than in winter season. This pattern agrees with Martinez and Dominguez (2014). Moreover, there is an inverse relationship between tropical (B6, B7) and subtropical (B8, B9, B10) portions of this basin, during winter and summer seasons. During monsoon rains in summer season, the recycled precipitation is larger in the subtropical portion than in the tropical portion of the La Plata Basin, while during non-monsoon rains in winter season, the recycled precipitation is larger in the subtropical portion than in the tropical portion of the La Plata Basin.

View Full Size

Spatial and seasonal distribution of precipitation recycling (2003–17).

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

The southern portion of the Amazon Basin (B3, B4, B5) and the tropical portion of the La Plata Basin (B6, B7) have strong seasonality of precipitation associated with South American monsoon (Villar et al. 2009a; Su and Lettenmaier 2009). In the southwestern (B3) and the southeastern (B4) regions of the Amazon Basin the precipitation recycling increases during non-monsoon season, as shown in previous studies (Zemp et al. 2014, 2017; Staal et al. 2018). Over the tropical portion of the La Plata Basin it is possible to verify that the largest dependency of precipitation on recycling occurs in autumn season, which corresponds to immediate post-phase of the SAM.

During all months of 2003–17 the regions of more frequent moisture source behavior (Supplementary Material 5: https://doi.org/10.1175/JHM-D-20-0080.s5) are the Tocantins/Araguaia Basin (B11) and the northernmost region of the La Plata Basin (B6), followed by the southern portion of the Amazon Basin (B3, B4, B5) and the northeastern region of the La Plata Basin (B7). These regions are moisture sources in more than 25% of the months during the period analyzed. Thus, in the annual mean, the Amazon and the La Plata Basins are moisture sink regions. This is in accordance with previous studies (Su and Lettenmaier 2009; Arraut et al. 2012; Satyamurty et al. 2012; Nascimento et al. 2016). It is interesting to note that, only in two months the northwestern region of the Amazon Basin (B2) is a moisture source. Both the months (February 2007 and September 2015) coincide with El Niño events.

Figure 4 shows the seasonality of monthly frequency of moisture source behavior in the study area. It is important to note that whenever a given region is not a source of moisture, it is a sink of moisture. As is verified in the spatial and seasonal distribution of precipitation recycling, the peak of monthly frequency of moisture source behavior in the boxes also has a northward propagation from summer season to spring season. This shows that the seasonality of moisture convergence, due to annual cycle of SAM, is the main reason for the seasonality of moisture source regions in the continent. However, with the exception of summer season, the boxes with largest values of moisture source frequency are not the same as the boxes with largest values of precipitation recycling.

View Full Size

Spatial and seasonal frequency of becoming moisture source (2003–17).

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

In summer season, the Tocantins/Araguaia (B11) and the Amazon (B1, B2, B3, B4, B5) basins are always moisture sinks, with the exception of one atypical month in the northwestern region of the Amazon Basin (B2). The tropical portion of the La Plata Basin (B6, B7) rarely becomes moisture source in this season, while the southernmost region of this basin (B10) is a moisture source in almost half of the summer months.

In the immediate post-monsoon phase in autumn season, the peak of moisture source behavior occurs over the central part of South America, more precisely in the southernmost region of the Amazon Basin (B5) and the northernmost region of the La Plata Basin (B6). The northern portion of the Amazon Basin (B1, B2) is always a moisture sink in all months of this season. In winter season, the southeastern region of the Amazon Basin (B4) is always a moisture source. Except for a few atypical months, the Tocantins/Araguaia Basin (B11), the southwestern region of the Amazon Basin (B3), and the northernmost region of the La Plata Basin (B6) also are always moisture sources in this season. These results agree with Arraut et al. (2012) and Nascimento et al. (2016). According to Arraut et al. (2012) the southern portion of the Amazon Basin is a moisture source during July and August. On the other hand, Nascimento et al. (2016) verify that the whole Amazon Basin becomes moisture source from July to October.

It is interesting to note that in the southernmost region of the Amazon Basin (B5) and the northernmost region of the La Plata Basin (B6), those receiving largest amount of their annual precipitation during summer season (Berbery and Barros 2002; Molina-Carpio et al. 2017), the frequency of becoming moisture sources is higher during autumn and winter seasons (increasing from autumn to winter) than in other seasons. Around southwestern (B3) and the southernmost (B5) regions of the Amazon Basin there is a large floodplain called the “Lhanos de Mojos.” This floodplain causes a damped effect of 2 months from peak of rain to peak of the streamflow in the upper Madeira subbasin (Molina-Carpio et al. 2017). The northernmost region of the La Plata Basin (B6) behaves like a huge marshland, the “Pantanal.” This causes a damped effect of 5–10 months from the peak rain to the peak streamflow of the Paraguay subbasin (Berbery and Barros 2002, and references therein; Su and Lettenmaier 2009). That is, these seasonal water bodies are large areas, supplying the evaporative demand of the atmosphere during periods of reduced oceanic moisture convergence over these regions, due to non-monsoon circulation.

In the spring season the peak frequency of moisture source behavior occurs over the northeastern region of the Amazon Basin (B1), followed by the Tocantins/Araguaia Basin (B11). This behavior in the northeastern region of the Amazon Basin coincides with the relatively dry season, when this region receives only 10% of its annual precipitation (Villar et al. 2009a). Over the La Plata Basin (B6, B7, B8, B9, B10) there is no well-defined spatial pattern, but the frequency of becoming moisture source in the regions of this basin is less than 25% during spring months. The central-western region (B8) is always a moisture sink in this season.

In Fig. 4 it is possible to observe that an opposite spatial pattern occurs between tropical (B6, B7) and subtropical (B8, B9, B10) portions of the La Plata Basin, from summer season to winter season. This is in agreement with Rodriguez and Cavalcanti (2006) and Nascimento et al. (2016). These authors verified that these portions of the basin have opposite behaviors. While one portion is a moisture sink the other is a moisture source. According to the results of Herdies et al. (2002) it is possible to verify that this behavior is due to reversal of the direction of moisture transport in the Chaco region from episodes of SACZ to episodes of LLJ east of the Andes. The seasonality of moisture source frequency, largest in tropical portion during winter season and largest in subtropical portion during summer season, also agrees with the results of Nascimento et al. (2016). Su and Lettenmaier (2009) also verified that tropical portion of the La Plata Basin is a moisture source during winter, while the region around the southeastern region becomes a moisture sink throughout the year.

It is important to note in Figs. 3 and 4 during winter season, that while the southeastern (B4) and the southwestern (B3) regions of the Amazon Basin have the largest values of precipitation recycling and frequency as moisture sources, the subtropical portion of the La Plata Basin (B8, B9, B10) has the smallest values of precipitation recycling and covers three of the four regions with the largest frequency of moisture sink. Thus, it is possible to verify that the southern portion of the Amazon Basin (B3, B4, B5) may be an important source of moisture to the subtropical portion of the La Plata Basin (B8, B9, B10). This agrees with the results of Drumond et al. (2008) and Martinez and Dominguez (2014), which verifies a significant importance of this portion of South America as moisture source to the La Plata Basin, mainly in the winter season.

4. Interrelationship analysis among boxes

The PEE by direct transport (Supplementary Material 6: https://doi.org/10.1175/JHM-D-20-0080.s6) is significant in the seasonal precipitation (≥10% of seasonal precipitation; see Supplementary Material 3) only in the southernmost (B5) and southwestern (B3) regions of the Amazon Basin during winter season. On the other hand, PEE by indirect transport (Supplementary Material 7: https://doi.org/10.1175/JHM-D-20-0080.s7) is significant in seasonal precipitation over the whole La Plata Basin (B6, B7, B8, B9, B10) and over the southernmost region of the Amazon Basin (B5) during the winter season, reaching ~30% of seasonal precipitation in the subtropical portion of the La Plata Basin (B8, B9, B10).

The total contribution of evapotranspiration from all the boxes of the study area to the seasonal precipitation in a given box (sum of precipitation recycled, direct and indirect transports) is called total terrestrial moisture contribution (TTMC) to the box. The spatial distribution of TTMC is shown in Fig. 5. The largest amount of TTMC occurs over the northwestern region of the Amazon Basin (B2) during autumn, winter, and spring seasons, with values larger than 199 mm per season. Large amounts also occur in the southwestern region of the Amazon Basin (B3) during summer and autumn seasons, and in southeastern region of this basin (B4) during summer season. It is interesting to note that the TTMC to seasonal precipitation is larger during summer season than in other seasons over the southeastern (B4) and southernmost (B5) regions of the Amazon Basin, the Tocantins/Araguaia Basin (B11), and the northeastern (B7), northernmost (B6), central-western (B8), and southernmost (B10) regions of the La Plata Basin. With the exception of the Tocantins/Araguaia Basin (B11) during the winter season, all the other regions have significant contribution of TTMC to the precipitation in all seasons. During winter season, the continental contribution to precipitation over the southwestern region of the Amazon Basin (B3) reaches ~50%, while in the central-western (B8) and northernmost (B6) regions of the La Plata Basin it reaches ~40%.

View Full Size

Total terrestrial moisture contribution (precipitation recycling + direct transport + indirect transport) in each box of study area (2003–17). B1–B5 are regions of the Amazon Basin. B6–B10 are regions of the La Plata Basin. B11 is the Tocantins/Araguaia Basin. The units of the bottom (top) columns are shown on the left (right) side. The percentages values are in relation to the total seasonal precipitation in each box.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

The PEE exclusively from the La Plata Basin (B6, B7, B8, B9, B10) by both direct and indirect transports (Supplementary Material 8: https://doi.org/10.1175/JHM-D-20-0080.s8) in all boxes are not significant. The largest amounts of PEE (>20 mm per season) occur over the southeastern region of the La Plata Basin (B9) during autumn and winter seasons, followed by the southernmost region of this basin (B10) during the autumn season. The maximum PEE from the La Plata Basin to the seasonal precipitation over the Amazon Basin reaches ~5 mm per season and only occurs over the southernmost region of the Amazon Basin (B5) during summer and autumn seasons. This occurs due to cold fronts that occasionally reach the Amazon Basin (Neto et al. 2015) and transport moisture from the subtropical portion of continent (Drumond et al. 2014) to southern Amazon Basin.

The PEE exclusively from the Tocantins/Araguaia Basin (B11) by both direct and indirect transports (Supplementary Material 9: https://doi.org/10.1175/JHM-D-20-0080.s9) occurs in all regions of the Amazon and the La Plata Basins during autumn, winter, and spring seasons, with less than 16 mm per season. However, the PEE from the Tocantins/Araguaia Basin is not significant in the seasonal precipitations over the boxes in the study area.

About the northernmost region of the La Plata Basin (B6), that covers the large marshland area of Pantanal, its PEE by both the direct and indirect transports (Supplementary Material 10: https://doi.org/10.1175/JHM-D-20-0080.s10) occurs over all regions of the La Plata Basin (B6, B7, B8, B9, B10). The largest amounts occur during winter and autumn seasons, but with less than 16 mm per season. Moreover, the PEE from this region is not significant in seasonal precipitations.

Figure 6 shows the PEE exclusively from the Amazon Basin (B1, B2, B3, B4, B5) by both direct and indirect transports. The largest values of PEE occur during winter season over the northwestern region of the Amazon Basin (B2) and over the subtropical portion of the La Plata Basin (B8, B9, B10), reaching more than 70 mm per season in the southeastern region of this basin (B9). However, the PEE from the Amazon Basin is significant in the seasonal precipitation during winter season over the whole La Plata Basin (B6, B7, B8, B9, B10) and over the southwestern (B3) and southernmost (B5) regions of the Amazon Basin. Over the northernmost region (B6) and the subtropical portion of the La Plata Basin (B8, B9, B10) and over the southernmost region of the Amazon Basin (B5), the PEE from the Amazon Basin represents more than 20% of winter season precipitation. This seasonal relative value agrees with value of contribution of evapotranspiration from the Amazon Basin to the whole La Plata Basin precipitation obtained by Zemp et al. (2014) during June–September. However, the relative values of summer season (Fig. 6) disagree with those of Zemp et al. (2014) during December–March. This is because the method of Zemp et al. (2014) considers that the transport of evapotranspirated moisture downwind occurs immediately after rainfall events, which are more frequent in summer season in the most parts of the Amazon Basin (Villar et al. 2009a), while the present study considers that the transport downwind occurs only when a given region is a source of moisture (T_O > T_I), more frequent during winter season according to Fig. 4. Therefore, the method of this study is more representative of transpiration of trees and evaporation from the soil during relatively dry season, while the method of Zemp et al. (2014), according to Zanin and Satyamurty (2020a), is more representative of the interception and fast transpiration, and not of transpiration of trees influenced by deep root uptake in the relatively dry season. On the other hand, the agreement during winter season between the two studies is due to the method of Zemp et al. (2014), different from this study, considering more than one cycle of moisture recycling, and therefore the evapotranspiration predominantly transported downwind is the interception and fast recycling from the regions with rainy season during winter season (mainly northwestern region of the Amazon Basin according to Villar et al. 2009a).

View Full Size

Direct and indirect transports from the Amazon Basin (2003–17). See explanation in Fig. 5.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

The mean annual PEE from the Amazon Basin (B1, B2, B3, B4, B5) to the whole La Plata Basin (B6, B7, B8, B9, B10) obtained in this study is 12% of the annual mean precipitation. This value is lower than the annual value obtained by Martinez and Dominguez (2014), but it agrees with the range of mean annual values obtained by Staal et al. (2018) along the La Plata Basin, and is close to mean annual value obtained by Yang and Dominguez (2019). This is due to the method of Martinez and Dominguez (2014) which considers the portion of evapotranspiration in the precipitable water at each grid cell at daily time scale, not regarding the occurrences of moisture divergence in the region, different from the method here. On the other hand, the study of Staal et al. (2018) also used a different method from this study and analyzed only the transpiration of trees. Therefore, they analyzed the predominant process of Amazonian evapotranspiration occurring in relatively dry season, similar to this study. Regarding the study of Yang and Dominguez (2019), they used a regional climate model embedded with a water vapor tracer. This method is also different from the method used in this study. However, the simulated precipitation over the Amazon and La Plata Basins strongly agrees with precipitation by TRMM product, and this satellite precipitation product is used in merged precipitation product used in this study (MERGE/CPTEC). Moreover, the simulated evapotranspiration is in good agreement with GLEAM product, which is the same evapotranspiration product used in this study. Therefore, both studies have a similar water balance in the Amazon and the La Plata Basins. On the other hand, while results of Yang and Dominguez (2019) show a very small difference between summer and winter seasons, in this study the difference between these seasons is large. This is because Yang and Dominguez (2019) calculated the moisture transport at 6-h intervals during a 10-yr period, while this study aggregates the moisture transport calculated at 3-h intervals to obtain monthly totals during a 15-yr period, and considers the moisture transport from the Amazon Basin to the La Plata Basin whenever divergence of water vapor occurs in any region of the Amazon Basin. However, both the studies have larger contribution of evapotranspiration from the Amazon Basin to the precipitation over the La Plata Basin in winter season than in summer season.

The results of the present study also disagree with the results of Drumond et al. (2014), in which the Amazon Basin is a moisture source to the southeastern region of Brazil during austral summer and spring seasons. The southeastern region of Brazil is located within box 7 analyzed in this study. As explained previously, the method of this study only tracks the moisture when a given region is a moisture source at monthly time scale, while the method used by Drumond et al. (2014) considers the difference between evapotranspiration and precipitation in each grid cell at 6-hourly time scale. Thus, the study of Drumond et al. (2014) also considers process that occurs when a region is not a moisture source at monthly time scale, and as interception and fast transpiration occur more frequently during rainy season, their result is different from the result obtained here.

The PEE from the Amazon Basin (B1, B2, B3, B4, B5) does not contribute to the northeastern region of the Amazon Basin (B1) and the Tocantins/Araguaia Basin (B11) in any season. Significant seasonal PEE over the La Plata Basin from the Amazon Basin is predominantly originates in the southwestern (B3) and southeastern (B4) regions. Figure 7 shows the PEE exclusively from the southwestern region of the Amazon Basin (B3) by both direct and indirect transports. Largest values of the PEE occur during winter season over the subtropical portion of the La Plata Basin (B8, B9, B10) and the southernmost region of the Amazon Basin (B5), reaching more than 25 mm per season in the southeastern region of the La Plata Basin (B9). On the other hand, the PEE is only significant to seasonal precipitation in the southernmost region of the Amazon Basin (B5).

View Full Size

Direct and indirect transports from the southwestern region of the Amazon Basin (2003–17). See explanation in Fig. 5.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

Regarding the southeastern region of the Amazon Basin (B4), Fig. 8 shows the PEE exclusively from this region by both direct and indirect transports. The largest values of PEE occur during winter season over the southwestern region of the Amazon Basin (B3) and the subtropical portion of the La Plata Basin (B8, B9, B10), reaching more than 35 mm per season in the southeastern region of the La Plata Basin (B9). The PEE values over these boxes are significant in this season. PEE is also significant in the southernmost region of the Amazon Basin (B5) and the northernmost region of the La Plata Basin (B6).

View Full Size

Direct and indirect transports from the southeastern region of the Amazon Basin (2003–17). See explanation in Fig. 5.

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

It is important to mention that the relation between the moisture sink regions (B8, B9, and B10) and the moisture source region (B4) during winter season is in agreement with atmospheric moisture transport patterns during the occurrences of the LLJ east of the Andes (Montini et al. 2019). The seasonal frequency of the LLJ is the highest during winter season (Nascimento et al. 2016; Montini et al. 2019), and according to Arraut et al. (2012) the moisture from the evapotranspiration of the southern portion of the Amazon Basin is transported by LLJ east of the Andes to the La Plata Basin during non-monsoon season.

5. Memory and coupling analyses

As shown in previous sections, there is a source–sink relationship between the southeastern region of the Amazon Basin (B4) and the subtropical portion (B8, B9, B10) and northernmost region (B6) of the La Plata Basin (see Fig. 8). The relations between these regions in the La Plata Basin and the soil moisture memory of the southeastern region of the Amazon Basin are analyzed in this section.

According to Fig. 9, the autocorrelation function of soil moisture has sinusoidal decay with time lag, and falls to below 99% confidence level at 91 days. Thus, the soil moisture memory is 3 months in the southeastern region of the Amazon Basin (B4), agreeing with Dirmeyer et al. (2009). In their Fig. 1, soil moisture memory calculated with daily data from GOLD-2 (version 2 of the Global Offline Land Surface Dataset), shows predominance of values around 2–3 months in the region of the Amazon Basin.

View Full Size

Student’s t test of autocorrelation function of soil moisture from the southeastern region of the Amazon Basin (2003–17).

Citation: Journal of Hydrometeorology 22, 6; 10.1175/JHM-D-20-0080.1

• Download Figure
• Download figure as PowerPoint slide

It is also verified in Fig. 9 that the autocorrelation function of soil moisture has a random behavior. This is due to the strong seasonality of precipitation in this region (see Fig. 7 in Villar et al. 2009a), with rainy season occurring in austral summer and relatively dry season occurring in the rest of the year, but mainly remaining dry during winter season (see Figs. 4 and 5 in Villar et al. 2009a).

It is important to mention that the 2 m of soil depth in soil moisture data from GLDAS-2 used in this study is less than the depth of uptake of soil water by roots of the Amazonian trees (Bruno et al. 2006; Broedel et al. 2017). According to Bruno et al. (2006) ~28% of the water used in evapotranspiration comes from the upper 2-m soil layer during relatively dry season, while in the rainy season it increases to ~56%. Thus, the soil moisture data used are a limitation in this study, and the soil moisture memory considering deeper soil can be longer than the value showed above.

For the following analysis, a 3-month time lag for precipitation and soil moisture is adopted. According to Table 2, the interseasonal and interbasins coupling occurs only between southeastern region of the Amazon Basin (B4) and the central-western region of the La Plata Basin (B8). In this case, the main set of variables (Set 1) has correlation coefficients higher than 0.50 and the complementary set of variables (Set 2) has correlation coefficients higher than 0.40.

Table 2.

Correlations between southeastern region of the Amazon Basin (B3) and the northernmost, central-western, southeastern, and southernmost regions of the La Plata Basin (B6, B8, B9, and B10, respectively). Correlations greater than 0.40 and significant at 0.01 are highlighted in bold. Variable t is two-tailed Student’s t test and n is sample size. For other abbreviations, see explanations in Table 1.

View Table

The very strong and significant positive correlations between lagged precipitation (variable P*) and lagged soil moisture (variable SM*), as well as between lagged soil moisture (variable SM*) and soil moisture without lag (variable SM), shows that the 3 months of hydrological memory found in the southeastern region of the Amazon Basin (B4) in Fig. 9 is robust. The strong and significant positive correlation between soil moisture (variable SM) and residual evapotranspiration (variable E_res) shows that soil moisture controls the evapotranspiration. According to Dirmeyer et al. (2009) when incident radiation availability is more than the availability of soil moisture, the correlation between soil moisture and evapotranspiration is positive, because the variability of soil moisture controls the evapotranspiration. Moreover, this positive correlation agrees with Fig. 3 of Dirmeyer et al. (2009) and Fig. 2 of Dirmeyer (2011), which show positive significant correlation between soil moisture and evapotranspiration from different data sources during austral winter season in the location of the southeastern region of the Amazon Basin.

The positive and significant correlation between residual evapotranspiration (variable E_res) from the southeastern region of the Amazon Basin and the moisture from evapotranspiration of this same region that reaches the central-western region of the La Plata Basin (variable X) is 0.54. This shows that substantial portion of the residual evapotranspiration from the source region (B4) reaches the sink region (B8). Moreover, the positive and significant correlation between variable X and the precipitation in sink region (B8) due to evapotranspiration from this source region (B4), represented by variable Y, is 0.64. This high correlation shows that most part of moisture from the southeastern region of the Amazon Basin that reaches the central-western region of the La Plata Basin generates precipitation during winter season. This agrees with the results of Martinez and Dominguez (2014), which verifies that moisture from the Amazon Basin is easily converted into precipitation over La Plata Basin during winter season due to the activity of transient systems. On the other hand, it is possible that the correlations obtained with variables X and Y could be due, in part, to oscillations caused by atmospheric wave trains. However, the influences of surface processes in the Amazon Basin are very important in these correlations.

On the average the retreat of SAM occurs at the end of April (de Silva and Carvalho 2007), and the results above show that, due to temporal permanence of soil moisture anomalies during 3 months in the southeastern region of the Amazon Basin (B4), the evapotranspiration during winter season supplies the lower troposphere with water from precipitations that occurred at the end of the SAM during autumn season. Thus, 13% of the precipitation in winter season (Fig. 8) over the central-western region of the La Plata Basin (B8) is influenced by the intensity of preceding monsoon rain over the southeastern region of the Amazon Basin (B4).

The modeling results showed in Fig. 5 of Collini et al. (2008) agree with results showed above. In this figure it is possible to observe that reduction in early soil moisture results in reduction in precipitation during following months over the locations of southeastern region of the Amazon Basin (B4) and central-western region of the La Plata Basin (B8). On the other hand, the increases in early soil moisture results in an indeterminate pattern in precipitation during following months over the location of the southeastern region of the Amazon Basin (B4), where precipitation predominantly increases in northern part and reduces in southern part. On the other hand, precipitation increases in the central-western region of the La Plata Basin (B8). Moreover, Collini et al. (2008) also verified that increases (decreases) in previous soil moisture results in more (less) availability of moisture to be transported by the LLJ east of the Andes. Therefore, considering the results in Table 2 and Fig. 3, it is possible to verify that less previous soil moisture concentration in the source region (B4) results in less precipitation recycling in this region and less residual evapotranspiration to be transported by LLJ to the sink region (B8), and thus the precipitation reduces in both regions. On the other hand, greater previous soil moisture in source region (B4) results in greater evapotranspiration, where precipitation increases in the northern part due to precipitation recycling and reduces in southern part due to moisture transported by the LLJ to the sink region (B4), where precipitation increases. However, it is important to mention that increased and reduced early soil moisture modeling experiments from Collini et al. (2008) are made over the whole South America, and therefore, the changes in precipitation at the location of the southeastern region of the Amazon Basin (B4) obtained by these authors also has remote effects, and the changes of precipitation in the central-western region of the La Plata Basin (B8) also has local effects. Moreover, the changes in soil moisture in these experiments are made in October, and precipitation is evaluated in the following monsoon months, and thus their results are different from this study.

6. Conclusions

The objective of this study is to evaluate the interseasonal and interbasins hydrological coupling between the two largest watersheds of South America. The new method to track the atmospheric moisture from a given source region to a given sink region developed here is efficient for this purpose. This method, Articulated Box Analysis, is based on a 2D Box Model approach and quantifies the moisture from evapotranspiration transported between neighboring regions (direct transport) and between remote regions (indirect transport). Some principles of Brubaker et al. (1993), Satyamurty et al. (2012), and Martinez and Dominguez (2014) are utilized in this method. It is important to mention that this method only tracks moisture transport from the regions with divergence of moisture on monthly time scale (source region). Therefore, it tracks mainly the moisture evaporated from water bodies and transpired by trees drawing soil water by their roots, as commonly occurs during relatively dry season in the monsoon regions. Moreover, while this method is very robust during seasons in which a box has high frequency of being a moisture source, it can underestimate somewhat the results during seasons in which a box has low frequency of moisture source behavior. Anyway, this method is found suitable for the purposes of this study.

It is also important to mention that the validity of results of this method depends on the magnitude of water balance nonclosure, and the median hydrological imbalance as percentage of median precipitation from the set of boxes in this study is small (5%). Moreover, the precipitation due to evapotranspirated moisture (PEE) is only considered significant if it is equal to or greater than twice the median value of the hydrological imbalance from the set of boxes.

The precipitation recycling ranges from 12% to 25% in the study area and, in annual mean, all the regions analyzed are sinks of atmospheric moisture. However, the southern portion of the Amazon Basin, tropical portion of the La Plata Basin, and the Tocantins/Araguaia Basin are sources of moisture to the atmosphere during more than 25% of the months analyzed. In the winter season, the precipitation recycling reaches 30% in the southwestern region of the Amazon Basin, and the southeastern region of the Amazon Basin is always a moisture source. The seasonal and spatial propagation of the peak of recycling and moisture source behavior show an inverse relation with monsoon rains. The displacement of moisture convergence over the continent, due to annual cycle of South American monsoon, is the main reason for these seasonalities in the study area.

The main sink regions to the total terrestrial moisture contribution (TTMC) from the whole study area are the northwestern region of the Amazon Basin during the autumn, winter and spring seasons, the southwestern region of the Amazon Basin during summer and autumn seasons, and the southeastern region of the Amazon Basin during summer season. However, the regions more dependent on TTMC to their seasonal precipitations are the southwestern region of the Amazon Basin and the northernmost and central-western regions of the La Plata Basin. The PEE from the Amazon Basin, during winter season is significant over its southwestern and southernmost regions and over the whole La Plata Basin. The annual value of PEE from the whole Amazon Basin represents 12% of annual precipitation over the whole La Plata Basin, while it represents more than 20% of winter precipitation over the northernmost region and the subtropical portion of the La Plata Basin. The southwestern and southeastern regions of the Amazon Basin are the main sources of moisture to the La Plata Basin. However, only the PEE from the southeastern region of the Amazon Basin is significant to winter precipitation over the whole La Plata Basin, with the exception of its northeastern region.

The hydrological memory of soil moisture in upper 2-m soil layer in the southeastern region of the Amazon Basin is about 3 months. Considering this time of hydrological memory, it is verified that the interseasonal and interbasins coupling occurs between southeastern region of the Amazon Basin and the central-western region of the La Plata Basin. It is shown that the amount of water precipitated over the southeastern region of the Amazon Basin at the end of the South American monsoon during autumn season influences the amount of precipitation during winter season over the central-western region of the La Plata Basin.

Therefore, the first hypothesis stating that “during non-SAM season, the hydrological memory of the southern Amazon Basin, due to water storage in deep soils by monsoon rains, plays a key role in the amount of moisture supply to the subtropical portion of the La Plata Basin,” and the third hypothesis stating that “the intensity of the SAM over the Amazon Basin can be considered an indicator of rainfall over the subtropical La Plata Basin during post-monsoon season,” are true, especially for the southeastern region of the Amazon Basin and the central-western region of the La Plata Basin. On the other hand, the second hypothesis stating that “during non-SAM season, the hydrological memory in the Pantanal region has also an important effect on the moisture supply to the subtropical La Plata Basin by the aerial river east of the Andes,” is not true, because the PEE from the Pantanal region is not significant to seasonal precipitation in other regions.

Regarding the first hypothesis, it is important to note that this refers to hydrological memory due to water storage in deep soils of the Amazon Basin, and the present study used a gridded data with soil depth of 2 m only. Thus, modeling studies considering deeper soil in parameterizations of the surface processes, associated with the method of analysis developed in this study, are recommended in future investigations. Regarding the third hypothesis, it is important to mention that the PEE over the central-western region of the La Plata Basin, influenced by the hydrological memory in the southeastern region of the Amazon Basin, is only 13% of winter precipitation. Therefore, the indicator mentioned in this hypothesis has to be modified to take into consideration the influence of other sources and mechanisms responsible for the other 87% of winter precipitation over the central-western region of the La Plata Basin.

It is also important to mention that the southeastern region of the Amazon Basin is located in the arc of deforestation in the Amazon forest (Costa and Pires 2010). Thus, the increase of deforestation in the southeastern region of the Amazon Basin can reduce the water supply to the population and economic activities of the Pantanal region and the subtropical portion of the La Plata Basin.