Low-frequency response of the upper Paraná basin

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 28: (2008) Published online 28 August 2007 in Wiley InterScience ( Low-frequency response of the upper Paraná basin Carlos M. Krepper, a Norberto O. García b * and Phil D. Jones c a CONICET- Facultad de Ingeniería y Ciencias Hídricas, Universidad Nacional del Litoral, Santa Fe, Argentina b Facultad de Ingeniería y Ciencias Hídricas, Universidad Nacional del Litoral, Santa Fe, Argentina c Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, England ABSTRACT: The low-frequency response of the upper Paraná River basin is investigated by means of singular spectral analysis (SSA) applied to precipitation rates and runoff contributions of different sub-basins, upstream of the Posadas gauging station. The temporal structure of Posadas discharges is characterized by a trend after 1970 and a modulated signal with a period of T 9 years that produces a maximum in amplitude around ENSO-range signals (T 3 5 years) contribute to fit the highest peaks in the streamflow series. Moderate changes in mean annual precipitation rates are reflected in large increases in mean runoff contributions for each subbasin. Changes in relative mean annual runoff responses before and after the 1970s, downstream of Sao Simao, cannot be only explained by increases in the corresponding mean annual precipitation rates. It is more probable that such changes arise as a consequence of some anthropogenic impact than a climatic one. In particular, the outstanding increase in runoff contribution, after 1970, for the drainage area between Itaipú and Jupiá of around 96% is mainly responsible for a positive trend in Paraná flow. From the detailed analysis of each sub-catchment, we can observe that not all the signals present in precipitation are reflected in runoff contributions and vice versa. Copyright 2007 Royal Meteorological Society KEY WORDS climatic signals; low frequency; rainfall; streamflow; upper Parana River Received 3 January 2006; Revised 22 February 2007; Accepted 4 March Introduction The relationship between the climatic behaviour of the rainfall over a river basin and its hydrologic response, through its streamflow, presents different degrees of complexity according to the physical characteristics of the basin. Chiew et al. (1995) have mentioned that the variations in precipitation are amplified in streamflow, and that, in general, it is easier to detect a change in discharge than directly in the basic climatic variables. This implies that river discharges reflect precipitation climatic signals. This is probably not completely true in large basins, particularly those that encompass different climatic regions. The Paraná River (Figure 1) is the most important component of the Rio de la Plata system and, together with the Paraguay river, constitutes a combined basin that covers an area of around km 2.The Paraná Paraguay basin represents around 84% of the total Rio de la Plata basin and contributes more than 80% to the Rio de la Plata streamflow. There is no agreement about the name of the different parts of the Paraná River system in the literature. According to Camilloni and Barros (2003), the portion of the river upstream of Corrientes City in Argentina, including the Paraguay River * Correspondence to: Norberto O. García, Hydroclimatic Research Unit, Department of Hydrologic Sciences, National University of Litoral, CC 217, (3000) Santa Fe, Argentina. nogarcia@fich1.unl.edu.ar basin, is known as upper Paraná, and downstream of this city to 32 S asmiddle Paraná. The river between this point and the confluence with the Río de la Plata is called lower Paraná. In contrast, Garcia and Vargas (1996) separate the Paraná Paraguay system into three hydrological units: the upper Paraná basin, upstream of Posadas City (Argentina); the lower Paraná basin downstream of this city and the separate Paraguay River basin. We use the second definition here. The Paraná River basin, upstream of the city of Posadas (Figure 1), from now on the upper Paraná basin, covers an area of around km 2, where 89.6% of drainage area is located in Brazil and the rest in Paraguay and Argentina. The sources of the Paraná River are located in the Santa Marta and Das Piloes hills, where it takes the name Paranaíba River. The regional toponymy recognizes the name of the river as Paraná downstream of the confluence of the Paranaíba and Grande rivers (García, 2000a). The main tributaries of the Paraná River come from the east of the basin, such as the Grande, Tieté, Paranápanema, Ivaí and Iguazú rivers; meanwhile, from the West, the basin receives contributions of many smaller rivers. The principal characteristic of the river basins that come from the east is the high slope from the Atlantic coast towards the confluence with the Paraná River. The northern tributary in the east is the Grande River, whose source areas are located in the Mantequeira hills, at heights of 1000 and 1500 m above sea level. The length of the Copyright 2007 Royal Meteorological Society

2 352 C. M. KREPPER ET AL. Figure 1. The upper Paraná River basin. Grande River is around 1050 km, with a mean slope for the sub-catchment of 0.69 m/km and a mean annual flow of 1930 m 3 /s(garcía, 2000a). Further south, we found the Tieté River confluence, upstream of the Jupiá dam (Figure 1), with a mean annual flow of 752 m 3 /s. The Tieté River sources are close to the city of San Pablo, at 750 m above sea level, and the mean slope of the basin is around 0.41 m/km along its 700 km length. The Paranápanema River sources are located in the Paranápiacaba hills, between 700 and 1200 m above the sea level, and, after a course of 750 km, it flows into the Paraná River, downstream to Jupiá, with a mean annual flow of 1384 m 3 /s. The southern tributary that comes from the east is the Iguazú River. The length of the Iguazú River is around 650 km, with a mean annual flow of 1409 m 3 /s and a mean basin slope of 1.02 m/km. During the austral summer in South America, the South Atlantic convergence zone (SACZ) and convective activity in the Amazon basin are the main components of the South American monsoon system (SAMS) (Carvalho et al., 2004), and an important source of precipitation for southeastern South America. The northern portion of the Parana basin has the largest precipitation contribution during austral summer with dry conditions during the cold season as a result of the predominant effect of the SAMS (Grimm et al., 1998; Berbery and Barros, 2002). Further south, the annual cycle changes in a transitional zone where no unique seasonal maximum is found. Large contributions in precipitation occur in the late winter and spring seasons. In this region, during the austral cold season, the most relevant forcing is due to frontal penetration, associated with migratory extra tropical cyclones (Vera et al., 2002). Krepper and Sequeira (1998) have studied the lowfrequency variability of the rainfall in the southeastern part of South America. Diaz et al. (1998), Grimm et al. (1998, 2000) and Barros and Silvestri (2002) analyse the relationship betweenprecipitation anomalies in southeastern South America and sea surface temperature (SST) in the Pacific and Atlantic Oceans, or directly with El Niño and La Niña events. Meanwhile, Krepper and García (2004) have explored the space-time structures of the variability displayed by the precipitation over the La Plata basin. Several studies performed over southeastern South America have used river flows as indicators of climatic variability, the extreme discharges and their remote climate forcing. Aceituno (1988), in particular, has mentioned a weak negative correlation between Paraná discharges at Corrientes and the southern oscillation index (SOI). Mechoso and Perez Iribarren (1992) reported that the streamflow in the Negro and Uruguay rivers has a clear tendency to be below average in the period June to December in high SOI (cold events) periods. Meanwhile, from November to the next February in ENSO years, there is a weak tendency to be above average. Anderson et al. (1993) analysed the flooding in the Paraná Paraguay basin for the period and they found that flooding was both more frequent and

3 LOW-FREQUENCY RESPONSE OF THE UPPER PARANÁ BASIN 353 more severe in the latter part of the period than previously. García and Vargas (1998) reported a particular behaviour of the upper Paraná basin, where the mean annual flows at Posadas are almost double of that at Jupiá, with drainage areas following approximately the same relationship. None of the gauged tributaries downstream of Jupiá (Paranápanema and Iguazú rivers) justify such a difference by themselves, so illustrates the role played by the small and nongauged rivers of the region. Amarasekera et al. (1997) found a positive correlation between Paraná discharges and SST anomalies of eastern and central equatorial Pacific Ocean. Robertson and Mechoso (1998) reported that the streamflow of the four major rivers in southeastern South America (Uruguay, Negro, Paraná and Paraguay) have interannual peaks with ENSO timescale and a near-decadal component as well as marked nonlinear trends in the Paraguay and Paraná rivers. Genta et al. (1998) have identified positive trends after the mid-1960s in the four major rivers of southeastern South America. Tucci and Clarke (1998) pointed out that the increasing runoff after the 1970s may be due to rainfall increases, agriculture practices or both. They conclude that if rainfall has been the principal cause of increased runoff, the increase may not be permanent; but if changes in land use have contributed to increased runoff, it is possible that such an effect might be more permanent. Camilloni and Barros (2000) explored the response of Paraná river during El Niño and events. Robertson et al. (2001) analysed the interannual to decadal predictability of the Paraná river, extracting near-cyclic components. Berri et al. (2002) examined the influence of ENSO on monthly streamflows for the upper Paraná River. Krepper et al. (2003) have analysed the interannual variability in rainfall and riverflows on several sub-catchments of the Uruguay River basin. Camilloni and Barros (2003) have studied the relationship between the phases of ENSO and the major discharges of the Paraná River upstream of Corrientes City. They concluded that two-thirds of the major discharge anomalies occurred during El Niño events and none of the major anomalies occurred during the La Niña phase. García and Mechoso (2005) reported changes in trend and interannual quasiperiodicities of Paraná discharges at Posadas and Corrientes. The purpose of this work is to investigate aspects of the temporal structure of time series of Paraná River flow upstream of Posadas, looking for deterministic signals that can explain the long-term response of the river and its interannual variability. At the same time, we study the probable relationships between long-term and quasioscillatory modes of runoff contributions with those of precipitation on the different sub-catchments (Figure 1). In essence, the latter consists of a detailed analysis, at the subbasin scale, of low-frequency and interannual quasioscillatory signals present in relative upstream contributions and precipitation rates and how these modes can explain the behaviour of the streamflow at the exit point of the upper Paraná basin. 2. Data and methodology 2.1. Data used Four gauging stations (Sao Simao, Ilha Solteira, Jupiáand Guairá) on the Paraná River with 70 years of naturalized monthly discharge data ( ) were used in the present study. An additional gauging station, at Posadas city, with 100 years of monthly discharges ( ) was also used. Brazilian monthly discharge series were obtained from the Operador Nacional do Sistema Elétrico (ONS) do Brasil and the monthly data at Posadas came from the Subsecretaría de Recursos Hídricos (SHR) in Argentina. The name of the station, river as well as the drainage areas and annual mean discharges are shown in Table I. In the basin, there are more than 58 dams each with more than 1 km 3 volume (Comision Mixta Paraguayo-Brasileña, 1974) but their effects on mean monthly flows are negligible (Garcia and Vargas, 1996). Precipitation data for the catchment come from the x gridded data set CRU TS 2.1 available at A significant data base for the region of monthly rainfall time series (1192 stations) was collected, reformatted and quality-checked during the project Assessing the impact of future climatic change on the water resources and the hydrology of the Rio de La Plata Basin, Argentina, , (Conway et al., 1999). The rainfall series were cross checked with those held in CRU global data set (Hulme, 1994) and all the new stations were added to the global data set and new time series of gridded rainfall for constructed for the region (CRU TS 1.0: New et al., 2000). The grids were subsequently updated and extended to 2000 (CRU TS 2.0: Mitchell et al., 2004) and after that to 2002 (CRU TS 2.1: Mitchell and Jones, 2005). The data set used contains monthly precipitation records from January 1901 to December 2000 (1200 time points) for 234 grid points. The data were totalled from monthly to annual precipitation values Method of analysis Singular spectral analysis (SSA) is a statistical method related to principal component analysis (PCA) but it is Table I. Stations, rivers, drainage areas and annual mean discharges. Gauging station River Drainage area (km 2 ) Annual mean discharge for the period (m 3 /s) (% of discharge at Posadas) Sao Simao Paranaíba (18.9%) Ilha Solteira Paraná (42.2%) Jupiá Paraná (50.6%) Itaipú Paraná (79.7%) Posadas Paraná (100%)

4 354 C. M. KREPPER ET AL. applied in the time domain. The objective is to describe the variability of a discrete and finite time series, X(t), in terms of its lagged autocovariance structure (Vautard and Ghil, 1989; Ghil and Vautard (1991); Plaut and Vautard, 1994; Vautard, 1995; Ghil and Yiou, 1996). The eigenvalue decomposition of the lagged autocovariance matrix produces temporal-empirical orthogonal functions (T-EOFs) (eigenvectors), and statistically independent temporal-principal components (T-PCs), with no presumption as their functional form. Each T-PC has a variance λ s (eigenvalue) and represents a filtered version of the original series, which can be classified essentially into nonlinear trends, deterministic quasioscillations and noise (Ghil and Yiou, 1996). A significance test for the singular values, λ s, can be made against a white noise null-hypothesis using the Monte Carlo (MC) method, generating an ensemble of 1000 independent realizations (Preisendorfer, 1988). Some authors have successfully applied SSA in different river basins of South America (Robertson and Mechoso, 1998; Robertson et al., 2001; Krepper et al., 2003, 2006 and Garcia and Mechoso, 2005). The whole upper Paraná Riverbasin( km 2 ) was divided into five different drainage areas or subbasins (Figure 1): P1-basin ( km 2 ) upstream of Sao Simao, on Paraná river; P2-basin ( km 2 ) formed by the drainage area between Ilha Solteira and Sao Simao; P3-basin ( km 2 ) between Jupiá and Ilha Solteira; P4-basin ( km 2 ) between Itaipú and Jupiá; P5-basin ( km 2 ) between Posadas and Itaipú. The runoff contribution of each subbasin, Q(t) j, is determined as the difference of discharges between the corresponding gauging stations, and the temporal structure of the Q(t) j series is analysed by means of SSA. The precipitation grid allows us to calculate annual precipitation rates, PR(t) j, defined as the annual precipitation contribution over different sub-catchments, which we express in m 3 /s, between 1901 and The behaviour of the upper Paraná basin The essential features of the annual evolution of the flow at Posadas are a high water in February and March with low-water periods in August and September. Figure 2 shows the monthly average contributions, for the period, of each subbasin. The P1-basin and P2-basin (northern part of the whole basin) have high contributions during the austral warm season and the minimum contribution during the cold season; meanwhile, the southern part of the basin (P4-basin and P5-basin) has the opposite behaviour. The runoff of the northern part of the basin is mainly responsible for the high-water period at Posadas in February and March; meanwhile, the runoff corresponding to the southern part reduces the effect of low water during August and September. From Table I, taking the mean annual flow at Posadas (Q(t) Posadas ) as reference, for , it is possible to estimate the relative average contribution of Figure 2. Monthly average contribution (m 3 /s) for each subbasin ( ). each drainage area. The P1-basin provides 18.9% of Q(t) Posadas ; P2-basin 23.3%; P3-basin only 8.4%; P4- basin 29.1% and P5-basin 20.3%. Around 50% of the flow that reaches Posadas comes from the drainage area upstream of Jupiá. Nevertheless, the relative contributions of the different sub-basins changed during the second part of the twentieth century. Anderson et al. (1993), García and Vargas (1998) and Tucci and Clarke (1998) have reported increases in Paraná streamflow after the 1970s. Garcia and Vargas (1998), in particular, have studied the changes in the river discharges and have inferred that during the last century the Paraná River at Posadas had two very different periods: a driest period (D-period) between 1931 and 1970 with mean annual streamflow Q(t) Posadas ( ) = m 3 /s and a humid period (Hperiod) after 1970 with Q(t) Posadas ( ) = m 3 /s, characterized by a positive trend in flow that does not seem to revert. This change in river flow before and after 1970 could be analysed from the relative changes in mean annual runoff contributions, Q(t) j (j = 1,...,5), of each subbasin, together with the corresponding mean annual precipitation rates, PR(t) j (Table II). We observe only moderate and quite uniform increases in mean precipitation rates, for all the sub-basins, between periods ( (PR(t) j ) 6 9%). However, the relative runoff responses have strong subbasin dependence. The variations between periods of different contributions range from a minimum change (Q(t) 1 9%, for P1- basin to a maximum one (Q(t) 4 96%, for P4-basin. The changes in mean annual runoff contributions produce a significant increase of 32% in mean annual flow at Posadas, without an equivalent increase in precipitation. In particular, Q(t) 4 for the H-period represents 30.4% of mean annual flow at Posadas, and it must be largely related to Paraná behaviour after On the other hand, the mean annual runoff coefficient for P4-basin, defined as RC 4 = Q(t) 4 /PR(t) 4, changes by around 86% after 1970; meanwhile, PR(t) 4 only changes by 6.8%. This fact could be indicating some drastic impact on the subbasin from one period to the other.

5 LOW-FREQUENCY RESPONSE OF THE UPPER PARANÁ BASIN 355 Table II. The mean annual contributions (runoff and precipitation rates in m 3 /s) of each subbasin for the driest and humid period. The number between parentheses represents percentage of contribution with respect to the Posadas discharge and ( ) the mean annual runoff coefficient. Mean runoff contributions (m 3 /s) P1-basin P2-basin P3-basin P4-bvasin P5-basin D-period ( ) (21.3%) (24.1%) (8.1%) (20.5%) (26.0%) H-period ( ) (17.6%) (23.1%) (7.12%) (30.4%) (21.8%) Mean precipitation rates (m 3 /sec) D-period ( ) H-period ( ) First, we analyse the temporal structure of mean annual discharges at Posadas, Q(t) Posadas, in order to detect any trend and quasiperiodicities. Our main interest is focused on the temporal behaviour of the time series and how the detected signals are evident in the Paraná behaviour after the 1970s. We also analyse shorter signals in the interannual frequency band (IFB) encompassing the ENSO-period range (2- to 5-year period) and how these signals could relate to extreme discharges. Table III shows the modes detected by SSA in Q(t) Posadas using different window length (M = 15, 20 and 25 years) where we have only retained singular values, λ s, with significant levels greater than 95%. The first component T-PC1 is associated with a nonlinear trend and the pair T-PC2 and T-PC3 is associated with a near-decadal oscillatory component with dominant period T 9 years. The periods were obtained by computing the power spectrum of the principal components for each oscillatory pair. There are moderate variance pairs with ENSOlike periodicities (Garcia and Mechoso, 2005) of about 3 5 years, robust to change in M but with low statistical significance. The pairs in the ENSO-range must be analysed after subtracting the long period signals of discharges (Robertson et al., 2001), as we see later. Figure 3 shows the partial reconstruction based on T- EOF1 (REC[1]), corresponding to the nonlinear trend (accounting for 18.7% of variance, for M = 25 year), and the reconstruction based in the pair T-EOF2 and T-EOF3 (REC[23]), corresponding to the near-decadal oscillatory mode (18.5%). REC[1] shows a marked positive trend after the 1970s and REC[23] a modulate oscillation, which increases its amplitude after the 1960s. From Figure 3, we can observe that the general long-term behaviour of the Paraná River, during the H-period, could be explained by the sum of REC[1] and REC[23]. To analyse possible shorter signals in the IFB (ENSOrange), we must detrend the discharge time series by first subtracting REC[123], obtained from a preliminary SSA with M = 25 years, and then applying SSA to residual series minimizing the M-length as possible, in order to obtain more statistical confidence. The results of applying SSA with M = 10 years to residual series are shown in Table IV; where we can see two quasioscillatory modes with T 3.7 and 5.2 years in the ENSO-range, accounting for a total of 37.4% of the total variance of the unfiltered original series. Figure 4 shows the partial reconstruction (REC(tot) = REC[1] + REC[23] + REC[1234](res)), corresponding to the trend and the near decadal oscillatory modes (REC[1] + REC[23]), plus the partial reconstruction of ENSOrange modes (REC[1234](res)) derived from the detrended residual series. From the comparison between Figures 4 and 3, we can observe how ENSO-range signals fit the higher peaks in Table III. Leading quasioscillatory modes detected by SSA in the time series of Posadas discharges ( ). Window length (years) Oscillatory pair (components) Trend or dominant period (years) Explained variance (%) 15 T-PC1 Trend 23.7 T-PC2&T-PC T-PC1 Trend 20.7 T-PC2&T-PC T-PC1 Trend 18.7 T-PC2&T-PC Figure 3. Partial reconstruction off annual discharges at Posadas (m 3 /s) (light solid curve), REC[123] = REC[1] + REC[23] (heavy solid); where REC[1] is the partial reconstruction for the nonlinear trend and REC[23] the corresponding to a quasioscillatory mode with T 9.2 years, derived from SSA, using a window length M = 25 years.

6 356 C. M. KREPPER ET AL. Table IV. Leading quasioscillatory modes detected by SSA with M = 10 years applied to the residual time series of Posadas discharges ( ). Table V. Largest peaks in mean annual discharges at Posadas and the reconstructed value (REC(tot)) for each peak. The third column shows the relative error (%) in the reconstruction. Window length (years) Oscillatory pair (components) Trend or dominant period (years) Explained variance of unfiltered original series (%) Year Mean annual peak (m 3 /s) The reconstructed value (REC(tot)) (m 3 /s) Relative error (%) 10 T-PC1&T-PC T-PC3&T-PC Figure 4. Annual streamflow at Posadas (m 3 /s) (light solid curve) and the partial reconstruction REC(tot) = REC[1] + REC[23] + REC[1234](res) (heavy solid), corresponding to the nonlinear trend, a modulated quasioscillatory modes with T 9.2(REC[1] + REC[23]); plus the partial reconstruction REC[1234](res) using ENSO-range modes derived from the residual series (as is shown in Table IV). the streamflow series, even during the unusual flood of Table V shows a list of the first 20 largest peaks in mean annual discharges at Posadas for the period, the reconstructed value (REC(tot)) for each peak and the relative fit error. In general, REC(tot) underestimates the peak values except for 1946, 1973, 1989 and To analyse the runoff response of the upper Paraná basin at subbasin scales, we study the temporal structure of the relative precipitation rates, PR(t) j, and runoff contributions series, Q(t) j, for the period in fivedrainageareas(j = 1,...,5). In particular, Table VI shows the results obtained from the SSA applied to the different series with M = 25 years. At this point, we can ask ourselves the following question: Are the runoff contributions (Q(t) j ) reflecting the input signals present in precipitation (P R(t) j )? The answer is that not all the signals present in precipitation are reflected in runoff contributions and vice versa. The relationship between rainfall variability and the behaviour of the runoff contribution in each drainage area is complex and particularly basin dependent. Generally, the relative runoff responses have the tendency to generate large signals (long-term and some interdecadal signals). From Table VI, we can observe that only PR(t) 1 and PR(t) 2, on the northern part of the upper Paraná basin (P1-basin and P2-basin), present small trends together with periodicities in the ENSO-range. In the rest of the drainage areas, the rainfall contributions (PR(t) 3 to PR(t) 5 ) only present periodicities in the ENSO-range. Even then, the relative contributions to streamflow for each subbasin contain trends in all cases, as well as interdecadal and interannual modes on some sub-catchments. We must consider that each component of the hydrological cycle can introduce proper signals in the system, and the water balance in the basin is a non-linear process. We note that interdecadal modes with T yrs in Q(t) 1 and Q(t) 2 never reach the Posadas gauging station. Applying SSA to streamflow series of the different gauging stations upstream to Posadas (Table I), it is possible to detect where the interdecadal modes introduced by Q(t) 1 and Q(t) 2 to the river flow disappear. With an SSA with M = 25 years (not shown here), we found that the streamflow at Sao Simao shows an interdecadal mode (T 21 yrs) accounting for 26.8% of variance and similar modes with T 22 and 21 years, explaining the variance of 26.0 and 22.3% detected at Ilha Solteira and Jupiá respectively. In an SSA applied to the streamflow series of Itaipú, the interdecadal mode still appears, but with low significance and explaining less than 10% of the total variance. In other words, the interdecadal modes, introduced in the river flow by Q(t) 1 and Q(t) 2, become attenuated along the river and disappear completely downstream of Itaipú. Nevertheless, interannual and near-decadal modes introduced by relative upstream runoff contributions are detected in the Posadas discharges (Tables III and IV).

7 LOW-FREQUENCY RESPONSE OF THE UPPER PARANÁ BASIN 357 Table VI. Leading quasioscillatory modes obtained from SSA with M = 25 years for the precipitation rates, PR(t), and runoff contribution series, Q(t), in each of the sub-basins. The ( ) means that detrended series, after subtracting the nonlinear trend and near decadal oscillation, were used with M = 10 years. Precipitation rate (PR(t)) Runoff contribution (Q(t)) Oscillatory pair (components) Trend or dominant period (years) Explained variance (%) Oscillatory pair (components) Trend or dominant period (years) Explained variance (%) P1-basin T-PC1&T-PC T-PC1&T-PC T-PC3 Trend 7.8 T-PC3 Trend 9.4 T-PC4&T-PC T-PC4&T-PC P2-basin T-PC1&T-PC T-PC1 Trend 13.9 T-PC3&T-PC T-PC2&T-PC T-PC6 Trend 6.0 T-PC4&T-PC T-PC6&T-PC P3-basin T-PC1&T-PC T-PC1 Trend 50.7 (T-PC1&T-PC2) P4-basin T-PC1&T-PC T-PC1 Trend 40.3 (T-PC1&T-PC2) P5-basin T-PC1&T-PC T-PC1 Trend 10.5 T-PC4&T-PC A prominent feature is the peculiar low-frequency response to precipitation of the whole upper Paranábasin, particularly downstream of Ilha Solteira (P3, P4 and P5- basin). The significant trends in the Q(t) 3,Q(t) 4 and Q(t) 5 represent 50.7, 40.3 and 10.5% of the total variance respectively, without any equivalent signal in PR(t) 3, PR(t) 4 and PR(t) 5. Figure 5 shows the partial trend reconstruction series derived from SSA with M = 25 years for each gauging station (Table I) along the river. We can observe the relative magnitude of different trends in comparison with the trend at Posadas (REC[1] Posadas ). Upstream of Itaipú, the low-frequency behaviour of discharges does not show a clear increase in the Paraná streamflow after the 1970s, as was pointed by Anderson et al. (1993), García and Vargas (1996) and Figure 5. Reconstructed time series for the corresponding trends, derived from an SSA with M = 25 years, for Posadas (REC[1] Posadas ), Itaipú (REC[1] Itaipu ), Jupiá (REC[1] Jupia ), Ilha Solterira (REC[1] Ilha Solteira ) and Sao Simao (REC[3] Sao Simao ). Tucci and Clarke (1998). Nevertheless, the partial reconstructed trends at Itaipú (REC[1] Itaipu ) and REC[1] Posadas show the characteristic behaviour of the Paraná River during the H-period as the principal consequence of lowfrequency contribution of Q(t) Summary and discussion It was found that changes in the long-term behaviour of streamflow at Posadas after the 1970s, as shown by several authors, could be explained from an analysis of time series by two deterministic signals: a positive nonlinear trend and a modulated near-decadal oscillatory mode. Meanwhile, ENSO-range signals (T 3.7 and 5.2 yrs) contribute to fit the higher peaks in the streamflow series, even during the unusual flood corresponding to the ENSO event. Robertson et al. (2001), in a seminal paper, explore the interannual predictability of summer (January to March) Paraná streamflow at Corrientes, downstream of confluence with the Paraguay River, using two quasioscillatory modes with dominant periods of T 8 and 17yrs. The authors emphasize the role of decadal-scale streamflow climatic fluctuation, and indicate that these may be partially predictable. Nevertheless, they found that the predictability due to ENSO-range fluctuations is negligible at interannual lead times. In the mean annual flow at Posadas, we found that ENSO-range fluctuations account for an important fraction of variance (37.4%) and could be potentially important in the flood peak prediction. García and Mechoso (2005) have found that the streamflow of all major rivers of South America has experienced a positive trend since the early 1970s. Furthermore, they mention that the synchronicity of events

8 358 C. M. KREPPER ET AL. across the continent suggests that the mechanism of generation has a large spatial scale. However, the relative responses of different sub-catchments (Table II) to a moderate and quite uniform increase in precipitation, from one period to another, have strong subbasin dependence. Changes in mean annual runoff responses, (Q(t) j ), could only be explained for increases in the corresponding mean annual precipitation rates, (PR(t) j ), downstream of Sao Simao. Except for the P1-basin, in the rest of the sub-basins, (Q(t) j )(j = 2,...,4) exceeded the corresponding (PR(t) j ). Note the increases in mean runoff coefficients, RC j = Q(t) j /PR(t) j, from one period to another in Table II. In particular, Q(t) 4 for the H-period represents 30.4% of the mean annual flow at Posadas and must be largely related to the long-term behaviour of the Paraná streamflow. We can appreciate that (RC(t) 4 ) between periods is around 86%, while for (PR(t) 4 ) it is only 6.8%. At this point, we can ask ourselves the following questions: What can produce an outstanding change in Q(t) 4 after the 1970s while the changes in the remainder sub-catchments are much smaller? Obviously, an increase in precipitation has not been the principal cause of such a runoff response. The upper Paraná basin represents 11% of Brazilian territory and within this area around 32% of the total population lives. There is a mean discharge of m 3 /s (Table I) and a water demand of around 515 m 3 /s (ANA, 2002). The situation of the basin, in 1998, indicated that almost the whole region was subject to a high level of anthropogenic influence (ANA, 2002). Therefore, we must focus our attention on modifications in the sub-catchments, particularly any strong change in land use. Several authors have mentioned that significant regional changes in deforestation and crop substitution have taken place in the Paraná basin (Anderson et al., 1993; Tucci and Clarke, 1998; Garcia, 2000b and Schnepf et al., 2001). According to Tucci and Clarke (1998), if changes in land use have contributed to increase runoff, probably such an effect is semipermanent, or at least a long-term effect. All the subbasin runoff responses present trends, and Figure 5 shows the relative contribution of each along the river. In particular, we can appreciate the difference between nonlinear trends at Itaipú and Jupiá, as a consequence of long-term changes in (Q(t) 4 ). In consequence, we expect that anthropogenic impacts, like changes in land use, must be greater in the P4-basin than in the rest of the subbasins. On the other hand, the P4-basin encompasses territories of three Brazilian states, Mato Grosso do Sul, Paraná and Sao Paulo. In the state of Paraná andsao Paulo, coffee production with permanent vegetation cover was drastically substituted by annual crops (corn and soybean) at the end of the 1960s (Tucci and Clarke, 1998; Garcia, 2000b). Brazil rapidly expanded their soybean and other field crops production driven predominantly by area expansion (Schnepf et al., 2003). In particular, during , the soybean area expansion grew by 8.5% per year, reaching more than 10 million hectares in The runoff coefficient (RC(t)) is a useful tool for studying the relationships between land cover and land use change, and impacts on runoff. The change in runoff coefficient may be a consequence of physical changes in the sub-catchment runoff response through several mechanisms, such as changes in land cover due to human activities, in moisture available, in the evapotranspiration rate, etc. (Conway, 2001). Figure 6 shows the temporal evolution of runoff coefficient in the P4-basin with linear trends before 1970 (D-period) and after 1970 (H-period). The tests of Mann (Sneyers, 1975; Garcia and Vargas, 1998), Accumulated Deviation (Q-test) and Worsley (Buishand, 1982) were applied to RC(t) 4 in order to detect changes in trend and/or jumps in the mean. Mann s test indicates a change in trend around 1965; meanwhile, the Worsley and Q-test show a jump in the mean in Although deforestation, especially in the states of Paraná and Sao Paulo, started early in the 1970s (Tucci and Clarke, 1998), the important change in Brazilian crop production occurred after 1970 (Graham et al., 1987; Conway, 2001). Figure 6. Temporal evolution of runoff coefficient (RC(t) 4 ) in the P4-basin (thin line) and the corresponding precipitation rate (P R(t) 4 ) (thick line).

9 LOW-FREQUENCY RESPONSE OF THE UPPER PARANÁ BASIN 359 To prove that a specific change in land use practice is the main factor behind (Q(t) 4 ) is a complex problem that needs additional information and exceeds our objectives. However, according to our results, it is more probable that, changes in Q(t) j, specially (Q(t) 4 ),ariseas a consequence of some anthropogenic impacts over the basin than a climatic one. Whatever the nonclimatic origin of the outstanding runoff response, this change in runoff contribution, Q(t) 4, is largely related to the flow increase at Posadas( 32%). From the inspection of Table V, we can appreciate another consequence of long-term behaviour of river flow. Fourteen of the top 20 largest peaks in annual discharges at Posadas occurred after 1970, showing a noteworthy increase in the frequency of extreme streamflow occurrence during the H-period. The analysis on each sub-catchment of temporal structures of PR(t) j and Q(t) j series (j = 1,...,5), along the period , shows that not all the signals present in precipitation are reflected in runoff contributions and vice versa. The sub-catchment runoff responses to precipitation variations are dominated by strong subbasin dependence. Generally, the relative runoff response has a tendency to generate long-term and some interdecadal signals. In particular, PR(t) 1 and PR(t) 2 present weak trends and ENSO-range signals. The corresponding runoff responses (Q(t) 1 and Q(t) 2 ) reflect the trends and partially filter the shortest signals present in precipitation. Downstream of Ilha Solteira, the noisy precipitation contributions only indicate the shortest signals in the ENSOrange (T 3.5 year). Q(t) 3 and Q(t) 4 reflect these signals but, at the same time, generate strong trends in runoff contributions. In contrast, P5-basin filters completely the short signal present in PR(t) 5 and the corresponding runoff generates a trend with a near decadal oscillation. As a consequence of the strong subbasin dependence, it is not possible to extract any general relationship between precipitation and runoff. In contrast, the upper Paraná River is not a conservative environment for some signals. In particular, we have shown that interdecadal signals (T yrs), introduced in river flow by Q(t) 1 and Q(t) 2, get attenuated along the river and disappear completely downstream of the Itaipú gauging station. 5. Conclusions The general long-term behaviour of the Paraná River after the 1970s (H-period) could be explained by the sum of a nonlinear trend and a modulated near-decadal oscillation. ENSO-range signals (T 3 5 yrs) contribute to fit the highest peaks in streamflow series even during the unusual flood of The other consequence of the long-term behaviour of the river flow is a noteworthy increase in frequency of extreme streamflow occurrence during the H-period. Changes in relative mean annual runoff responses before and after the 1970s, downstream of Sao Simao; especially (Q(t) 4 ) 96%, cannot be simply explained by increases in the corresponding mean annual precipitation rates. It is more probable that such changes arise as a consequence of some anthropogenic impact than a climatic one. Such possible anthropogenic impacts must be stronger on some subctachments, such as the P4-basin, than on the others. From the detailed analysis of each sub-catchment, we can observe that not all the signals present in precipitation are reflected in runoff contributions and vice versa. The sub-catchment runoff responses to precipitation variations have strong subbasin dependence. Generally, the relative runoff response has a tendency to generate longterm signals in the form of nonlinear trends, especially downstream of Jupiá. The upper Paraná River is not a conservative environment for some signals. Interdecadal oscillations witht 21 22yrs, introduced in river flow by Q(t) 1 and Q(t) 2, are attenuated along the river and disappear downstream of the Itaipú gauging station. Acknowledgements The authors are grateful to the anonymous reviewers for their helpful comments on the earlier version of this paper. This research was partially supported by ANCyT-Universidad Nacional del Litoral, Argentina Grant PICTO-UNL Contract BID 1728/OC- AR. References Aceituno P On the fluctioning of the Southern Oscillation in the South America Sector. Monthly Weather Review 116: Amarasekera KN, Lee RF, Williams ER, Eltahir EAB ENSO and the natural variability in the flow of tropical rivers. Journal of Hydrology 200: ANA The evolution of water resources management in Brazil, Agencia Nacional de Aguas (ANA), Also available at Anderson RJ, da Franca Ribbeiro dos Santos N, Diaz H-F An análisis of Flooding in the Paraná/Paraguay River Basin. LATEN Dissemination Note #5, The World Bank, Latin America & the Caribbean Technical Department, Environment Division. Barros VR, Silvestri GE The relation between sea surface temperature at subtropical south-central Pacific and precipitation in Southeastern South America. Journal of Climate 15: Berbery EH, Barros VR The hydrologic cycle of the La Plata Basin in South America. Journal of Hydrometeorology 3: BerriJG,GhiettoMA,García NO The influence of ENSO in the flows of the Upper Paraná River of South America over the past 100 years. Journal of Hydrometeorology 3: Buishand TA Some methods for testing the homogeneity of rainfall records. Journal of Hydrology 58: Camilloni I, Barros VR The Paraná River response to El-Niño and events. Journal of Hydrometeorology 1: Camilloni IA, Barros VR Extreme discharge events in Paraná River and their climate forcing. Journal of Hydrology 278: Carvalho LMV, Jones C, Libmann B The South Atlantic Convergence Zone: persistence, intensity, form, extreme precipitation and relationships with intraseasonal activity. Journal of Climate 17: Chiew F, Whetoton P, McMahon T Detection of climate changes in recorder runoff volumes in South-east Australian Rivers. International Journal of Climatology 13:

10 360 C. M. KREPPER ET AL. Comisión Mixta Paraguayo-Brasileña Estudiode estudio del río Paraná: Informe final, factibilidad. Apéndice A: Hidrología y Meteorología. Ande-Electrobras, Asunción-Rio de Janeiro, Brazil. Conway D Understanding the hydrological impacts of landcover and land-use change, IHDP Update Issue, 1, article 2, Also available at Conway D, Jones PD, García NO, Vargas WM Assesing the impact of future climatic change on the water resources and the hydrology of the Rio de La Plata Basin, Argentina. Final Report of work under contract ARG/B7-3011/94/25a. Commision of European Communities; Norwich: UK - Santa Fe: Argentina. Diaz AF, Studzinski CD, Mechoso CR Relationships between precipitation anomalies in Uruguay and southern Brazil and sea surface temperature in the Pacific and Atlantic Oceans. Journal of Climate 11: García NO. 2000a. Análisis de la variabilidad climática de la Cuenca del Río de la Plata, a través de los caudales de sus principales ríos. Thesis Doctoral, Universidad Nacional de Córdoba, Argentina; 130. García NO. 2000b. About the climate variability and runoff in the Río de la Plata basin. In Preprints of 6 th International Conference on Southern Hemisphere Meteorology and Oceanography, Santiago, Chile, 3 7 April, 279/281. García NO, Vargas WM The spatial variability of runoff and precipitation in the Rio de la Plata basin. Hydrological Sciences Journal 41: García NO, Vargas WM The temporal climatic variability in the Río de La Plata basin displayed by the river discharges. Climatic Change 38: García NO, Mechoso CR Variability in the discharges of South American rivers and in climate. Hydrological Sciences Journal 50(3): Genta JL, Pérez Iribarren G, Mechoso CR A recent increasing trend in the streamflow of rivers in Southeastern South America. Journal of Climate 11: Ghil M, Vautard R Interdecadal oscillations and the warming trend in global temperature time series. Nature 350: Ghil M, Yiou P Spectral methods: what they can and cannot do for climatic time series. In Decadal Climate Variability: Dynamics and Predictability, Anderson D, Willebrand J (eds). Elsevier: Amstredam. Graham DH, Gauthier H, Mendonca de Barros JR Thirty years of agricultural growth in Brazil: Crop performance, regional profile and recent policy review. Economic Development and Cultural Change 36(1): Grimm AM, Ferraz SE, Gomes J Precipitation anomalies in southern Brazil associated with El Niño and La Niña events. Journal of Climate 11: Grimm AM, Barros VR, Doyle ME Climate variability in southern South America associated with El Niño and La Niña events. Journal of Climate 13: Hulme M Validation of large-scale precipitation fields in global circulation models. In Global Precipitation and Climate Change, Dasbois M, Desalmand F (eds). NATO ASI Series I, Vol 26. Springer -Verlag: Berlin; Krepper CM, Sequeira ME Low frequency variability of rainfall in Southeastern South America. Theoretical and Applied Climatology 61: Krepper CM, García NO Spatial and temporal structure of trends and interannual variability of precipitation over La Plata Basin. Quaternary International 114: Krepper CM, García NO, Jones PD Interannual variability in the Uruguay River basin. International Journal of Climatology 23: Krepper CM, Garcia NO, Jones PD Paraguay river basin response to seasonal rainfall. International Journal of Climatology 26: Mechoso CR, Pérez Iribarren G Streamflow in Southeastern America and the Southern Oscillation. Journal of Climate 5: Mitchell TD, Jones PD An improved methods of constructing a data base of monthly climate observations and associated highresolution grids. International Journal of Climatology 25: Mitchell TD, Carter TR, Jones PD, Hulme M, New M A comprehensive set of high-resolution grids of monthly climate for Europe and the globe: the observed record ( ) and 16 scenarios ( ), Tyndal Centre Working Paper 55, 25. New M, Hulme M, Jones PD Representing twentieth century space-time climate variability. Part 2: development of monthly grids of terrestrial surface climate. Journal of Climate 13: Plaut G, Vautard R Spells of low-frequency oscillation and weather regimes in the Northern Hemisphere. Journal of Atmospheric Sciences 51: Preisendorfer RW Principal Components Analysis in Meteorology and Oceanography. Elsevier: Amsterdam; 425. Robertson AW, Mechoso CR Interannual and decadal cycles in River flows of Southeastern South America. Journal of Climate 11: Robertson AW, Mechoso CR, García NO Interannual prediction of the Paraná river. Geophysical Research Letters 22: Schnepf RD, Dphlman E, Bolling C Agricultural in Brazil and Argentina: Developments and prospects for major field crops. ERSAgriculture and Trade Report No. WRSO13; 85. Sneyers R Sur l analyse statistique des series d observations. Technical Note No WMO: Geneve; 192. Tucci CEM, Clarke RT Environmental issues in the la Plata Basin. Water Resources Development 14: Vautard R Patterns in Time: SSA and MSSA. In Chapter 14 of Analysis of Climate Variability: Applications of Statistical Techniques, von Storch H, Navarra A (eds). Springer Verlag: Berlin; 327. Vautard R, Ghil M Singular spectrum analysis in non-linear dynamics with applications to paleoclimatic time series. Physica D 35: Vera CS, Vigliarolo PK, Berbery EH Cold season synoptic scale waves over subtropical South America. Monthly Weather Review 130: