1 Introduction

The La-Plata basin is South America’s second largest basin, and it is formed by two main rivers: the Paraná and the Paraguay. The Paraguay River has a drainage area of about 1,095,000 km2 at its confluence with the Paraná River, at the border between Argentina and Paraguay. The upper part of this basin, the Upper Paraguay River Basin (UPRB), drains 600,000 km2 and includes parts of Brazil, Bolivia, Argentina and Paraguay. One of world’s largest (140,000 km2) wetlands is located in the central portion of the UPRB, which is known as Pantanal. The flat topographic relief and the seasonal flood pulse of Pantanal strongly govern the flow regime of Paraguay River and of its upper course tributaries. Floods in rivers flowing through the Pantanal are dampened and delayed. For instance, peak flows along the 1,250-km Paraguay River reach between Cáceres (immediately upstream of Pantanal) and P. Murtinho (downstream of Pantanal) are delayed by 3–4 months.

Human activities in the region are strongly regulated by this hydrologic regime, such as navigation, cattle raising and farming. For instance, the available land for cattle raising and farming is dependent on inundation extent in each wet season. As the Paraguay River flows along the heartland of the South American continent, it forms a natural corridor for the development of the agricultural and mineral resources of the region (Collischonn et al. 2001). On the other hand, floodplain inundation plays a key role for several ecological processes and phenomena (Junk et al. 1989), and thus the river flow regime and floodplain inundation patterns of Pantanal strongly regulate ecosystem functioning, integrity and conservation and make it very vulnerable to human induced changes (Junk et al. 2006; Hamilton 2002). In fact, several anthropogenic activities, such as agriculture, cattle raising, dam building and other hydraulic condition changes, are threatening the Pantanal ecological balance (Tucci and Clarke 1998; Hamilton 1999, 2002; Da Silva and Girard 2004).

Furthermore, changes in climate conditions of the UPRB may also cause significant disturbances on ecosystem functioning, mostly by altering precipitation and evapotranspiration rates, which in turn may affect river flow regime and floodplain inundation dynamics. The impacts of climate change may even amplify and worsen undesirable consequences of some human interventions on hydrologic conditions of the basin.

However, the prediction of certain anomalies on future precipitation and temperature conditions over a watershed are not straightforward translated into prediction of anomalies of the hydrologic regime along the river network. There are several factors influencing the rainfall-runoff transformation, which is a non-linear process, making the response of the basin in terms of river flow regime for the prescribed climate change not directly deductible, unless a specific methodology is employed. For large-scale basins, the distributed hydrologic modeling is the most appropriate tool, allowing taking into account the spatial variability of the predicted climate conditions.

This paper focuses on estimating the impacts of currently prescribed climate change scenarios over river flow regime along the UPRB. The estimates of future precipitation and temperature generated by Atmospheric/Ocean General Circulation Models (AOGCMs) based on two emission scenarios (IPCC 2000) are used as input for running a coupled hydrologic-hydrodynamic model of the UPRB.

In coupled atmospheric-hydrologic modeling for assessing climate change impacts, several sources of uncertainty exist, such as future greenhouse gas emissions, AOGCM structure, downscaling procedure from AOGCM, hydrological model structure, hydrological model parameters (Kay et al. 2009). As it is not the focus of this paper to evaluate and quantify all these uncertainties, we considered the outputs generated by a multi-model ensemble of 20 AOGCMs for running the hydrological model and obtaining river flow projections. This procedure can be viewed as a simplified approach of taking into account part of the uncertainties associated to atmospheric modeling for purposes of flow projections. Estimates of impacts of those climate change scenarios over the UPRB hydrologic regime are assessed through analysis of monthly hydrographs and by comparison of the results corresponding to the set of AOGCM climate predictions relative to the current flow regime.

2 Basic features of the UPRB

The Paraguay River Basin (PRB) is located in the centre of the South-American continent, including territories from Brazil (34 % of the basin), Paraguay (32 %), Bolivia (19 %) and Argentina (15 %). The Paraguay River is the second largest tributary of the La Plata River, with 2,612 km of length, and receives contributions of several tributaries (Fig. 1). This study focuses in the Upper Paraguay River Basin, defined as the contributing area of the Paraguay River upstream of the affluence of Apa River, in the border of Paraguay and Brazil (Fig. 1). This basin has three sub-regions: the Planalto (260,000 km2), the Pantanal (140,000 km2), and the Chaco (200,000 km2). The Planalto is a relatively high relief region, presenting a well-defined and convergent drainage network, and where the climate is relatively wet. Located in the central portion of the UPRB, the Pantanal presents very low and flat relief, characterized by a complex drainage system that is seasonally flooded, with a relatively drier climate. Rivers flowing from the Planalto enter the Pantanal and inundate the floodplains, creating an intricate drainage system, with vast lakes, divergent and endorreic drainage networks. Annual rainfall is less than the potential evaporation and drainage is slow due to shallow gradients and low margins of the main channels (Tucci et al. 1999; Bordas 1996). The third part, the Chaco, is located west of the border of Brazil, mostly in Bolivian territory, being characterized by low rainfall (typically less than 1,000 mm per year) and an endorreic and ill-defined river drainage network. This part has a very limited contribution to total Paraguay River discharge.

Fig. 1
figure 1

The Upper Paraguay River Basin (shaded region) and its location within the Paraná–La Plata river system

Full size image

According to the Köppen climate classification, the climate in the region is of the type Tropical Savanna, with rainfall concentrated in summer. Online Resource 1 shows the inter- (part a) and the intra-annual (part b) observed precipitation variability, in terms of areal average values. The rainy season begins in October and ends in April, presenting monthly precipitation ranging approximately from 100 to 300 mm. In the dry period, monthly precipitation ranges between 0 and 100 mm, with lower variability among years than in the rainy period. Along the period 1968 to 2000, annual average precipitation ranged between 920 and 1,540 mm, with a mean value of 1,320 mm. In most of the basin, average temperatures range from 18 °C to 22 °C. July is the coldest month, with mean temperatures ranging from 16 °C to 18 °C. September and October are the hottest months with temperatures averaging 24 °C to 26 °C.

3 Methodology

3.1 Climate change scenarios and anomalies projections

The two IPCC emission scenarios on which the climate projections used in this study were based (A2 and B2) are two of four qualitative scenario storylines characterized by different projections and assumptions regarding socio-economic development, conservation of natural resources, human population growth, technology improvement, and adoption or not of mitigation actions.

We selected the marker scenario of each family. Thus, results from the A2-ASF (where ASF means Atmospheric Stabilization Framework Model, e.g. Pepper et al. 1998) and B2-MES (where MES come from MESSAGE: Model for Energy Supply Strategy Alternatives and their General Environmental Impact, e.g. Riahi and Roehrl 2000) emission scenarios are used. These emission scenarios thus set the forcing conditions for the AOGCMs to estimate future climate conditions.

The AOGCMs, however, were not designed to provide the actual representation of future climate conditions, but to provide reasonable representations of the system in the future based on a limited set of observations (Allen and Ingram 2002). Furthermore, comparison of temperature and precipitation outputs of AOGCMs with observed data for current and past periods indicate poor representation of local climate conditions (Koutsoyiannis et al. 2008). Owing to this, we adopted the approach of considering the climate conditions prescribed by the AOGCM in terms of anomalies of the variables of interest relative to current climate rather than taking its absolute values.

We used the MAGICC/SCENGEN, version 5.3v2 (Wigley 2008; Hulme et al. 1995), to obtain the projected monthly anomalies of mean temperature and precipitation over the UPRB corresponding to the scenarios A2-ASF and B2-MES and simulated by 20 AOGCMs, listed in Online Resource 2. For each scenario and for this set of AOGCMs, average projected anomalies for each month of the year were produced for two 30-years time intervals centered in 2030 and 2070. Thus, a total of 1920 simulations were performed with the MAGICC/SCENGEN (20 AOGCMs × 12 months × 2 scenarios × 2 horizons × 2 variables: precipitation and temperature anomalies).

As presented by other studies (e.g. Serrat-Capdevila et al. 2007), within the set of AOGCMs, some models may be considered more or less adequate for the study region, as deficiencies or simplification in their formulation may not account for all the local meteorological phenomena. However, which model to select may be variable according with which variable is considered for comparison, e.g. mean annual precipitation, seasonal precipitation fluctuation, monthly mean temperatures, etc. Additionally, Reichler and Kim (2008), Hagedorn et al. (2005) and Serrat-Capdevila et al. (2007) have shown as multi-model forecasts outperform using “best” single models in the long run, as those typically performed in climate change analyses. For these reasons, and also taking into account that recent studies suggest that the choice of the AOGCM is the largest quantified source of uncertainty in projected impacts of climate change on river flow (Nóbrega et al. 2011; Bates et al. 2008; Kay et al. 2009; Blöschl and Montanari 2010), this study considered all the 20 models aiming to provide a representative uncertainty characterization in the results.

3.2 Estimates of daily precipitation and temperature projections

The approach adopted was to apply the projected anomalies of each AOGCM onto observed historical data in order to generate the projected daily values. The projected anomalies of mean monthly temperature and precipitation over a 30-year time interval were downscaled and converted to daily estimates in a gridded format, considering the spatial resolution of the distributed hydrological model.

The procedure performed was as following:

  1. 1.

    According to data availability the selected period ranges from 1971 to 2000. For precipitation, observed daily data available from 92 rainfall gauges were interpolated to the center of each grid cell of the hydrologic model, using the inverse-distance-squared method. For temperature, the closest meteorological station was identified for each hydrologic model grid cell among the 18 meteorological station with observed temperature data available.

  2. 2.

    Among the 36 AOGCM grid points covering the study region is identified the one located closest to each hydrologic model grid cell.

  3. 3.

    To produce the daily precipitation/temperature projections for each horizon, a loop in time for the period 01/01/1971 to 12/31/2000 in a daily time step was performed.

  4. 4.

    For each date (day d, month m, and year y), projected precipitation/temperature at the hydrologic model grid cell i (Pproj(d,m,y,i); Tproj(d,m,y,i)) was obtained as:

$$ \mathrm{Pproj}\left(\mathrm{d},\mathrm{m},\mathrm{y},\mathrm{i}\right)=\mathrm{Pobs}\left(\mathrm{d},\mathrm{m},\mathrm{y},\mathrm{i}\right)*\left[1+\varDelta \mathrm{Pproj}\left(\mathrm{m},\mathrm{j}\right)\right], $$
(1)
$$ \mathrm{Tproj}\left(\mathrm{d},\mathrm{m},\mathrm{y},\mathrm{i}\right)=\mathrm{Tobs}\left(\mathrm{d},\mathrm{m},\mathrm{y},\mathrm{i}\right)+\varDelta \mathrm{Tproj}\left(\mathrm{m},\mathrm{j}\right), $$
(2)

where Pobs and Tobs are the daily observed precipitation and temperature in the corresponding date at cell i, respectively; ∆Pproj and ∆Tproj are the projected precipitation (%) and temperature anomalies (°C), respectively, of month m at AOGCM grid point j.

  1. 5.

    Carrying out step 4 for each hydrologic model grid cell, two gridded precipitation data and two gridded temperature data are generated, each with 10,958 time steps: one refers to the period from 01/01/2015 to 12/31/2044 (short horizon, 30-year time interval centered in 2030), and the second corresponds to the period 01/01/2055 to 12/31/2084 (long horizon, 30-year time interval centered in 2070).

3.3 Coupled hydrologic-hydraulic modeling

The hydrologic-hydraulic modeling framework used in this study is the one previously described in Paz et al. (2010) and Bravo et al. (2012). This is a conceptual model composed by two components: (a) a distributed hydrologic model for simulating the rainfall-runoff processes along the UPRB; (b) a full 1D hydrodynamic model for flow routing along the main drainage network.

The first component is the large-scale, distributed hydrologic model MGB-IPH, which calculates soil water budget, evapotranspiration, flow propagation inside a cell, and flow routing through the drainage network, as fully described in Collischonn et al. (2007). This model was developed and has been applied for modeling large-scale South American basins, and acceptable results have been obtained (e.g. Tucci et al. 2003; Allasia et al. 2006; Collischonn et al. 2008; Bravo et al. 2009; Paz et al. 2011). The model was also used more recently to analyze impacts of climate change on water resources of the Rio Grande basin, which is another tributary of the Parana River basin (Nóbrega et al. 2011).

In the present study, the MGB-IPH model was applied to the entire UPRB with a 0.1° × 0.1° square-grid discretization (a total of 5,195 grid cells), and considering a daily time step. Different time periods were considered for model calibration at several control points, ranging from 5 to 10 years according to data availability, while the period from 1995 to 2001 was used for model validation. Several performance measures were assessed with acceptable results (Bravo et al. 2012).

The second component of the modeling framework is the well-known HEC-RAS hydraulic model (USACE, US Army Corps of Engineers 2004). About 4,800-km of river drainage system along the UPRB was modeled with HEC-RAS, being discretized into 24 river reaches, 12 junctions, 1124 cross-sections and 11 storage areas. A special effort was conducted for combining detailed cross section profiles of the main channel with elevation values extracted from SRTM-90 m Digital Elevation Model (DEM) for characterizing the large floodplains, and thus properly representing river hydraulics (Paz et al. 2010).

The hydrologic model generates daily streamflow data used as upstream boundary conditions and as lateral incremental contribution for running the hydraulic-hydrodynamic model. Hydrologic and hydrodynamic models are off-line coupled, i.e., the former provides input data to the later, but running the hydrodynamic model has no effect on running the hydrologic model. Following this approach, the observed flow regime along the Paraguay River as well as along its tributaries were satisfactorily reproduced (Bravo et al. 2012).

3.4 Flow regime projections

The projections of temperature and precipitation in the gridded format were used as input for running the MGB-IPH hydrologic model for the two 30-years future time intervals, considering the historical observed values for the other required meteorological data (atmospheric pressure, solar radiation, wind velocity and humidity). Land use representation into the hydrologic model was kept constant through time, i.e. no land use change scenarios were simulated. Projected streamflow produced by the MGB-IPH model was used as upstream boundary conditions for running the HEC-RAS hydraulic model and thus obtaining the projected flow regimes. As each AOGCM climate projection for each scenario and horizon was used for generating the corresponding projected streamflow, the coupled hydrologic-hydraulic model was run 80 times (2 scenarios × 2 horizons × 20 AOGCMs).

3.5 Flow regime comparison statistics

Regarding to flow regime comparison, three sets of model results were obtained. The first one is the current hydrologic regime of the basin and was obtained using a single run of the coupled hydrologic-hydraulic model, considering as input the historical observed values of precipitation and meteorological data during the period 1971–2000. The second and third set of models results were estimated by means of the coupled model using as input the predicted temperature and precipitation time series from each AOGCM, in each scenario and future time interval, as previously presented.

Mean monthly flows were calculated based on daily results of the hydrological model, at each control point for each model run. To avoid the influence of initial conditions of the coupled model, the first 4 years were disregarded and statistics were estimated for a 26-year time interval. For the predicted streamflow time series, three statistics were calculated from the AOGCM model ensemble: the median, as a measure of central tendency, and the 10th and 90th percentiles, for representing a measure of dispersion or uncertainty.

4 Results

4.1 Projected temperature anomalies

In terms of areal average values over the basin and considering the mean values among all the AOGCMs, the monthly temperature anomalies projected for the scenario A2-ASF and 30-years time interval centered in 2030 (named time horizon A) ranged from +0.88 °C on February up to +1.48 °C on October (Fig. 2a). A pattern of positive anomalies was very clear for the projections along the entire year, despite the dispersion among results of the AOGCMs. At each month, the difference among results of the 20 AOGCMs was relatively large, with more discrepancies at February, March and April, when both temperature increases and decreases are projected, and at September and October, where occurred the largest range of projected temperature increases (from +0.29 to +3.3 °C at September, and from +0.66 to +6.07 °C at October).

Fig. 2
figure 2

Monthly projected temperature anomalies over the Upper Paraguay River Basin for scenarios A2-ASF and B2-MES and 30-years time interval A (centered at 2030) and B (centered at 2070): mean, minimum and maximum values from the range of the 20 AOGCMs results. The box represents the mean +/− 1 standard deviation

Full size image

For this same short time horizon, similar temperature anomalies were projected for the scenario B2-MES relative to scenario A2-ASF (Fig. 2c). Overall, the results of the scenario B2-MES present a pattern of slightly increasing the positive anomalies observed for the scenario A2-ASF, but also of slightly enlarging the dispersion of results among the set of AOGCMs. Areal average temperature anomalies projected for scenario B2-MES ranged between +0.93 °C at February and +1.55 °C at October, considering mean values of all AOGCMs.

For the long time horizon B (30-years time interval centered in 2070), the mean monthly temperature anomalies projected for the UPRB range from +2.57 to +4.29 °C for the scenario A2-ASF (Fig. 2b), and from +2.01 to +3.37 °C for the scenario B2-MES (Fig. 2d). This slightly smaller increases in temperature projected for scenario B2-MES relative to scenario A2-ASF is consistent to their definitions, as B2 storyline regards a more ecologically friendly world than A2, with emphasis on local solutions to achieve economic, social and environmental sustainability. For both scenarios, the variability of results among the 20 AOGCMs follows the pattern observed for the short time horizon, but with the dispersion among models quite amplified, as observed through comparison among graphs (a)–(c) and (b)–(d) of Fig. 2 (note that temperature scale is distinct for each time horizon). This larger dispersion among AOGCMs’ projections for the longer time horizon is expected, due to the high non-linearity of climate and propagation of uncertainties related to models (structure, parameters) and initial conditions.

4.2 Projected precipitation anomalies

Projected precipitation anomalies considering areal average values over the basin are presented in Fig. 3 for both scenarios A2-ASF (Fig. 3a and b) and B2-MES (Fig. 3c and d), and for both time horizons A and B. These anomalies are presented in terms of percentage changes relative to observed precipitation. Overall, there are large anomalies, predominating the projection of positive increases in precipitation, but with a quite large dispersion among the results of the 20 AOGCMs. However, as anomalies in terms of percentage of change is a relatively concept, it is quite dependent on the mean values used to obtain those anomalies. As a result, the largest anomalies projected for the months of the dry period are due to the relatively low precipitation rates during these months, do not representing the largest projections of monthly precipitation. For instance, July presents anomalies reaching up to +570 % (Fig. 3b), but this is calculated over a mean precipitation of around 20 mm and represents a projection of a monthly rainfall of 116 mm.

Fig. 3
figure 3

Monthly projected precipitation anomalies over the Upper Paraguay River Basin for scenarios A2-ASF and B2-MES and 30-years time interval A (centered at 2030) and B (centered at 2070): mean, minimum and maximum values from the range of the 20 AOGCMs results. The box represents the mean +/− 1 standard deviation

Full size image

Also, the dispersion among the AOGCMs results is quite larger during the low rainfall period, albeit this dispersion is high during the rainy period too. For instance, projected anomalies of precipitation for January (scenario A2-ASF and time horizon A), the wettest month, present a mean value of +3.5 % and the interval defined by mean ± 1sd is from −8.2 % to +15.3 %, occurring a maximum projected anomaly of +44.6 % and a minimum of −11.4 % (Fig. 3a). For the scenario A2-ASF and short time horizon, monthly projected precipitation anomalies in terms of mean values of the set of 20 AOGCMs ranged between −16.2 % and +6.3 %. This range is very similar to the one obtained for the scenario B2-MES (from −16.6 % to +6.6 %), as showed in Fig. 3c. In fact, projections did not differ substantially between these two scenarios for the time horizon A, but mean projections for the scenario B2-MES are slightly higher and present slightly higher dispersion among the models results than for scenario A2-ASF. For the wettest months, from December to March, mean projected anomalies range between −0.3 % on December and around +6.4 % on March for both scenarios.

The projected anomalies of precipitation for the long time horizon present even more dispersion among the AOGCMs. For instance, for January scenario A2-ASF, mean projected anomaly is +10.7 % and the mean ± 1sd interval is between −23.9 % and +44.5 %, while maximum and minimum anomalies are −33.1 % and +130.3 %, respectively (Fig. 3b). The projections for this horizon also present a larger distinction between scenarios A2-ASF and B2-MES (Fig. 3c) than that observed for the short horizon. Overall, there is a projection of more rainfall for all months for scenario B2-MES relative to scenario A2-ASF, this difference increasing for the months with more intense anomalies projected in scenario A2-ASF. An additional analysis of the spatial variability over the UPRB of the projected precipitation anomalies is presented in Online Resource 3.

The individual analysis of AOGCM results shows that, for almost every month, half of the models predicted positive average precipitation anomalies over the UPRB, while the other half predicted negative anomalies. Intermodel variability is larger during the dry season months (May to September) than during the wet season. Meanwhile, for almost all models, positive average temperature anomalies were predicted for the entire year.

4.3 Projected flow regime

Comparison between the current streamflow and models ensemble statistics of projected streamflows for several control points at the UPRB are presented in Figs. 4, 5 and 6, for scenario A2-ASF and for both time horizons A and B. Similar representation of the results from B2-MES scenario are presented in Online Resource 4.

Fig. 4
figure 4

Monthly flow at the P. Murtinho control point (UPRB outlet) for the scenario A2-ASF and 30-years time interval B (centered at 2070) for the set of AOGCMs

Full size image
Fig. 5
figure 5

Currently and projected monthly flow statistic at eight control points along several rivers of the UPRB for the emission scenario A2-ASF and 30-years time interval A (centered at 2030): Models ensemble Median, percentile 10 % and percentile 90 % are presented

Full size image
Fig. 6
figure 6

Currently and projected monthly flow statistic at eight control points along several rivers of the UPRB for the emission scenario A2-ASF and 30-years time interval B (centered at 2070): Models ensemble Median, percentile 10 % and percentile 90 % are presented

Full size image

As presented in a previous section, patterns of temperature increase persist over the entire year for almost all AOGCMs in both time horizons. When this information is used as input to the hydrological model, the evapotranspiration rate results higher than the current rate. The anomalies of precipitation show larger dispersion than the temperature anomalies, with projection of increase or decrease in the future. As the hydrologic response of the basin (i.e. the streamflow) is a combination of multiple factors, future streamflow presents positive anomalies if an increase in precipitation counterbalances the increase in evapotranspiration. On the other hand, future streamflow present negative anomalies if the increase of evapotranspiration could not be overcome by an increase in precipitation, or when the AOGCM also projects a decrease in precipitation. An example of individuals AOGCM results in terms of flow at the P. Murtinho control point (UPRB outlet) for the scenario A2-ASF and 30-years time interval B (centered at 2070) is presented in the Fig. 4.

Overall, the Median of models ensemble is very close to the current streamflow in all months of the year, in both scenarios and both time horizons. The Median of models ensemble projections is slightly lower (higher) than current streamflow during almost all months of the year in the A2-ASF scenario for time horizon A (time horizon B) (Figs. 5 and 6). For the B2-MES scenario, the Median of models ensemble are almost equal between the two time horizons.

The dispersion of models results is very similar between the scenarios A2-ASF and B2-MES in the short time horizon A, as presented in Online Resource 5 in terms of the 10th and 90th percentiles of projected anomalies relative to the current mean annual flow. The mean range of these percentiles is from −12 % to about +8 % considering all control points. The dispersion of models results are enlarged in both scenarios during the long time horizon B, with higher increments of the uncertainty range in the A2-ASF scenario. For instance, the mean range of percentiles is from −21.5 % to +31 % considering all control points in the long time horizon B and A2-ASF scenario, while slightly lower mean range of percentiles are obtained in the B2-MES scenario (from −19 % to +22.5 %).

The individual analysis of AOGCM results was extended to the streamflow anomalies, as shown in Fig. 7, in which six scatterplots of projected annual anomalies are presented (scenario A2-ASF and both time horizons): UPRB mean areal temperature (∆T, °C) versus UPRB mean areal precipitation (∆P, %) (Fig. 7a and b); UPRB mean areal precipitation versus mean streamflow (∆Q, %) at the basin outlet (Pto. Murtinho control point) (Fig. 7c and d); UPRB mean areal temperature (∆T, °C) versus mean streamflow (∆Q, %) at the basin outlet (Fig. 7e and f). Results for time horizon A are presented in Fig. 7a, c and e, and results for time horizon B are shown in Fig. 7b, d and f.

Fig. 7
figure 7

Scatter plot of projected UPRB annual mean areal temperatures anomalies (∆T[°C]) vs. projected annual mean areal precipitation anomalies (∆P[%]); projected annual mean areal precipitation anomalies vs. projected annual mean streamflow anomalies (∆Q[%]) at the basin outlet (Pto. Murtinho control point) and projected annual mean streamflow anomalies at the basin outlet vs. annual mean areal temperatures anomalies, for scenario A2-ASF and 30-years time interval A (centered at 2030) and B (centered at 2070) by 20 AOGCMs. *Results from GIS—ER model are outside of the scatterplot (∆P = 116 %; ∆T = 3,03 °C; ∆Q = 275 %)

Full size image

A clear pattern of basin hydrologic behavior emerge from these results, as models with higher positive temperature anomalies, which are also associated to higher negative precipitation anomalies, result in higher negative streamflow anomalies. These streamflow anomalies are, in percentage, higher than the precipitation anomalies (for instance, see results from models GFDLCM21 and CSIRO-30). On the other hand, projected positive precipitation anomalies are not necessarily related to positive streamflow anomalies, as illustrated by the results from the MIROCMED model in time horizon B. This may be due to the projection of high positive temperature anomalies that reduce the water availability to generate runoff, compensating the effect of increasing water input by means of increased precipitation.

Finally, we present statistics summarizing projected temperature, precipitation and streamflow anomalies obtained in this study. Projected temperature and precipitation anomalies are represented by annual mean values over the basin, while projected streamflow anomalies are given as Median annual values at basin outlet. For both scenarios (A2-ASF and B2-MES), in time horizon A there is a projection of decreasing streamflow of −1.7 %, which is due to the increase in temperature (+1.1 °C), despite of the slight increase in precipitation (+0,4 %). In turn, in time horizon B, the results between the scenarios are more distinct, but both project increase in streamflow. This streamflow increase projection is larger for scenario A2-ASF (+5.1 %), due to the larger increase projection in precipitation (+9.5 %), in spite of the also larger increase projection in temperature (+3,2 °C), in comparison to the results of the scenario B2-MES (∆Q = 1.9 %; ∆P = +6.0 %; ∆T = 2.5 °C).

5 Summary and conclusions

We presented an analysis of predicted climate change on streamflow of the Upper Paraguay River basin, in the region of the Brazilian Pantanal, which is recognized as a sensitive environment due to its climate and other geographical features. The analysis was based on temperature and precipitation anomalies projected by 20 AOGCMs and considering two climate change scenarios from the IPCC (scenarios A2-ASF and B2-MES) and two 30-years time intervals centered at 2030 (time horizon A) and 2070 (time horizon B). These anomalies projections were converted to estimates of daily temperature and precipitation projections, which were then used as input for running a coupled hydrologic-hydraulic model and obtaining the flow regime projections.

Results show that the impacts on streamflow are highly dependent on the AOGCM used to obtain the climate predictions. Patterns of temperature increase persist over the entire year for almost all AOGCMs in both time horizons, increasing the evapotranspiration rate in the hydrological model. The precipitation anomalies show major dispersion with projection of increase or decrease in the future. Based on these inputs, results from the coupled hydrologic-hydraulic model show nearly one half of the projections resulting in increasing river discharge, and the other half result in decreasing river discharge. If the mean or median of the predictions is considered, no discernible change in river discharge should be expected, despite the dispersion among results of the AOGCMs that reached +/−10 % in the short horizon and +/− 20 % in the long horizon at several control points.

Despite of the simplifications and limitations regarding the adopted approach, the results obtained in this study may be seen as a reasonable guess of the UPRB flow regime projection regarding the climate change prescribed in scenarios A2-ASF and B2-MES from IPCC.