Abstract
The extremely high-streamflow events of the Paranaíba River basin are found to be associated with La Niña phenomenon during December–February (DJF). Extreme events are identified based on their persistent flow for seven days and more after taking retention time into consideration. The extremely high-streamflow events are associated with the La Niña years; 80% of the high-streamflow events have occurred during La Niña phases. Therefore, a very-significant 80% and above correspondence of the La Niña events and the seasonal streamflow anomalies are found in DJF. Although climate variations have direct relationship with the rainfall, streamflow variations are considered as the surrogates to rainfalls. However, apart from climate variations the anthropogenic and land-use changes also influence streamflow variations. In this study, we have applied multivelocity TOPMODEL approach and residual trend analysis to examine the impact of land-use to the streamflow at the Fazenda Santa Maria gauge stations. However, the model residual trend analysis of the TOPMODEL approach cannot quantify the extent of land-use impact. Thus, La Niña phase is important components to understand and predict the streamflow variations in the Paranaíba River basin.
1. Introduction
Streamflow plays a major role in the livelihood of the people in a river catchment. Hence, the scientific analysis of streamflow is very essential for the present and future generations. The influences of climate variability on the streamflows have been studied by Sahu et al., [1–3] in their previous studies of Indonesia, and found very good correlation of the impact of climate variability on the streamflow. Several studies performed on southeastern South America have used streamflows as indicators of climatic variability from the interannual to the seasonal scale [4–6]. It is stated that the climate variability and changes can be studied by analyzing river flows as a surrogate to rainfall, under the assumption that changes in the rainfall are reflected and likely amplified in streamflows [7, 8]. Moreover, it is easier to detect a change in streamflow than to directly observe changes in the basic climatic variables [9].
The Paranaíba River flows in the Rio Paranaíba of Brazil and in the state of Minas Gerais of the Mata da Corda Mountains (19°13′21′′S and 46°10′28′′W). The river is flowing at an altitude of 1,148 meters. The length of the river is approximately 1,000 kilometers. The Paranaíba and the Grande River both confluence and then form the second largest Parana River of Brazil, at the point to make the borders between the states of Sao Paulo and Minas Gerais [10]. The catchment area of the Paranaíba is approximately 36,000 km2. However, Fazenda Santa Maria gauge station (17°58′51′′S and 50°14′49′′W, Figure 1) is in the Upper Paranaíba River catchment, having a catchment area of about 16,750 km2. The Upper catchment is not artificially regulated; thus, it is best suited for our analysis to minimize anthropogenic influences on streamflow [1]. This river is the primary source of water to the Parana River. The water resources of this basin sustain one of the most densely populated regions of South America, where harvests and livestock are among the region’s most important assets [10].
The physical characteristics of a river basin and the relationship between the climatic behavior of rainfall and its hydrologic response, through streamflow, can present different degrees of complexity [1, 11, 12]. Streamflow is a synthesis of precipitation and evapotranspiration and various components of the hydrologic cycle together with possible anthropogenic influences [13]. The rainfall variation in Northeast Brazil is shown to be influenced by variability in the tropical Atlantic besides El Niño/La Niña [14, 15]. In this study we investigate the ENSO (El Niño and Southern Oscillation) particularly La Niña relationship at the basin scale. The signature of La Niña is found in the extremely high discharge events of December-February (DJF) in the Paranaiba River basin. This paper also applies the hydrological model TOPMODEL [16] with a multivelocity approach in order to investigate the land-use change on discharges.
2. Data and Methods
2.1. Model Input Data
The topographic data used in this study were extracted by using ETOPO1 elevations global data from National Geophysical Data Center (NGDC), National Oceanic and Atmospheric Administration (NOAA). The topographic data were composed of basin boundary, slopes, cells distances (distance to the next downward cell), cells areas, and cumulative areas. Precipitation data were obtained from ANA (Brazilian National Agency of Water Resources) in two stations: Fazenda Aliança and Maurilândia. Meteorological data (radiation and temperature) were extracted from Hirabayashi et al.’s [17] reanalysis. They developed and assessed a global 0.5 degree near-surface atmospheric data from 1948 to 2006 at daily time scale; we used data from 1978–2006.
Potential evapotranspiration was estimated through the Priestley-Taylor radiation method [18]. As TOPMODEL is a lumped hydrological model, an aerial average daily precipitation (Figure 2) and evapotranspiration (Figure 3) data were used as input. For this period (1978–2006) the mean precipitation value was 3.94 mm with a maximum value of 108.95 mm, whereas the mean evapotranspiration value was 4.11 mm with a maximum value of 6.37 mm and minimum value of 2.34 mm. Daily discharges data were acquired from ANA at Fazenda Santa Maria station. They encompass the period from 1978 to 2006. The last six years (2001–2006) of this time series were used for model calibration purpose and the entire time series was used for model validation purpose.
2.2. Climatology and Composite Index Data
Daily climatology and anomalies of river discharge are computed from the 29-year data. Extremely high discharge events were cataloged based on a threshold; 1.5 ( stands for standard deviation) was set as threshold for extremely high discharges events. The NCEP/NCAR (National Centers for Environmental Prediction/National Center for Atmospheric Research) global atmospheric reanalysis-1 zonal wind (850 hPa) dataset [19] is used from January 1, 1979 to December 31, 2008. The other major dataset used in this study is the global coverage NOAA interpolated of daily averages of outgoing longwave radiation anomalies (here after OLR) data on a grid at standard pressure levels from 1 January 1979 to 31 December 2008 [20]. In addition to these, the SST anomalies are used from the daily OISST analysis version 2 AVHRR-AMSR (Advanced Very High Resolution Radiometer-Advanced Microwave Scanning Radiometer) products from National Climate Data Center (NCDC) from 1981 to 2008 [21].
3. Paranaíba Streamflow Characteristics
The climatology of streamflow (Figure 4(a)) at the Fazenda Santa Maria gauge station of the Paranaíba River in Brazil shows significant flow from November to May and very little flow from June to October. The variation in this seasonal streamflow significantly affects the human population [10]. A linear trend is seen in the streamflow at the Santa Maria stations. During the season, we have found that the El Niño Modoki influence reduces the streamflow to nearly half of the average streamflow of the whole time series for extremely low-discharge events [3]. However, in this study we have investigated the influences of La Niña for extremely high-streamflow events (Figure 4(b)).
(a)
(b)
It is important to understand the underlying mechanisms that cause the variation of streamflows due to the influences of La Niña on the Paranaíba streamflows. A scientific analysis is made to link the streamflow variability with the rainfall and SST and OLR variations on daily time scale like the previous studies [1, 2]. Apart from the climate variability impact, in this study we have applied multivelocity approach TOPMODEL to examine the land-use influences on the streamflow, because the river streamflows, unlike the rainfall, are affected by morphological and anthropogenic factors including soil and forestry recharge, sediment deposit, topography and land-use changes.
4. Hydrological Model Approach
The multivelocity model approach, which is consistent with field observations carried out by Leopold et al. [22], consists in deriving a time-area function from a distance-area function using the following equation: where () is the time of concentration of a determined distance-area function class ; is a proportionality constant (); is a power law exponent (–); is the plan flow path length from a class area to the basin outlet; () is the cumulative area of the class and is the total number of classes which the distance-area function is composed. Details about this approach and its implementation may be seen in the work of Silva et al. [23, 24].
In order to evaluate the model performance, Nash coefficient [25] and log Nash coefficient were chosen as follows: where is the observed discharge at the time , is the calculated discharge at the time given the parameter set , is the observed discharge average, and is the number of time steps. Thereby, the model performance (Em) is determined by the product of these two coefficients, that is, by the product of (1) and (2). This is an attempt to search for simulations that try to fit the observed discharge data at high and low discharges simultaneously.
The methodology consists basically of model calibration against a period of six years, model validation over thirty-one years, and model residual trend analysis.
4.1. Model Performance
In the calibration period, the model obtained a performance coefficient Em of 0.54 (6 years) and, in the validation period, Em was equal to 0.32. From Figure 6 it is possible to see that most observed discharges lay inside the uncertainty bounds of 90% and inside the max/min interval. Therefore, the model was validated for the entire time series. The model residuals analysis (Figure 5) does not provide a clear upward trend in the discharges. This means that there may be very little difference between observed and calculated discharge increased along the time. However, a statistical test was carried out to find the significance of the trend on model residual. Kruskal-Wallis test [26] was applied to identify significant difference among the first six years and the last six years (Figure 6). The test showed little difference between the groups (group 1 and group 2, Figure 7) at . It is probably due to the flux in the form of heat or mass transfers. Nevertheless, the land-use does not have very significant influences on the streamflow characteristics.
5. Impact of La Niña on Austral Summer
To examine the possible other component impacts on streamflow of the Paranaíba River, we investigate the climate variability influences on the streamflow at Fazenda Santa Maria gauge station. In this study we found that the La Niña has significant influence on Paranaíba streamflow during austral summer (DJF). As shown in Table 1, 7 out of the total 9 extremely low-discharge events are associated with La Niña during the austral summer season.
Moreover, 80% of extremely high discharge events are found in the La Niña phase of austral summer (Table 1). Out of the 9 extremely high discharge events during the austral summer season, 7 events are associated with La Niña and only one event is associated with La Niña Modoki. The composite anomalies of SST, wind and OLR for all the events during the DJF extremely high streamflow depict a La Niña condition when the eastern Pacific is colder than normal (Figure 8). Unlike the El Nino Modoki related extremely low-streamflow events (figure not shown), we find here that the tropospheric subsidence associated with La Niña condition is more confined to Amazon basin.
We also notice anomalously strong winds blowing from tropical Atlantic to most parts of Northeast Brazil including the Paranaíba catchment thereby introducing more surface moistures over that region. This also explains the negative OLR anomalies seen above that region and associated extremely high streamflows. Further velocity potential at 200 hPa shows significant convergence over the Paranaíba catchment (Figure 9). If we take the probability of occurrences because of La Niña, La Niña influences around 80% of the extremely high discharge events.
If we compare these analyses with the multivelocity TOPMODEL output, we may conclude that climate variability such as La Niña influences the extremely high discharges events more than any other factor in the Paranaiba catchment, as it is a general acceptance that land-use influenced more to the high discharge events due to soil erosion, sediment deposits, and other anthropogenic land-use changes. Here we recognize that climate modes could cause equal or more amounts of damages to the streamflows.
6. Conclusions
In this study we analyzed the daily streamflow of the Paranaíba River at the Fazenda Santa Maria gauge station on investigate the impact of climate variations. Also, we examine the land-use influences to the streamflow by applying the multivelocity TOPMODEL approach by the residual analysis. During DJF or austral summer season we found that 80% of the extremely high discharge events occurred when eastern Pacific represents a La Niña-like situation.
The La Niña has significantly influenced the extremely high-streamflow characteristic of the Paranaíba River Upper catchment. However, the model residual trend analysis of the TOPMODEL approach cannot quantify the extent of land-use impact, which implies that rainy season’s extremely high discharge events of the Paranaíba River catchment at the Fazenda Santa Maria gauge stations are influenced mostly by the La Niña phases of the Pacific. Hence, for the societal benefits of this densely populated region climate factors should be investigated properly with special references to the La Niña phase of the Pacific.
Acknowledgments
NCEP/NCARreanalysis and ANA (Brazilian National Agency of Water Resources) and OISST analysis version 2 AVHRR-AMSR (Advanced Very High Resolution Radiometer-Advanced Microwave Scanning Radiometer) products from NCDC (National Climate Data Center) are provided by NOAA (available online), USA.