SEARCH

SEARCH BY CITATION

Keywords:

  • land use change;
  • la Plata basin;
  • Uruguay River

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] The Uruguay River basin experienced extensive land use change during the second half of the twentieth century as a result of expansion of agricultural area, while streamflow and precipitation increased during the same period. This study assesses the impact of land use change between 1960 and 2000 on streamflow using a hydrology model that explicitly accounts for the role of land cover. Although the model suggests that land use change could have large effects on streamflow, changes in mean streamflow are attributable to climatic variations and not to land cover change. On the other hand, a faster runoff response to precipitation was observed toward the end of the period, which does appear to be attributable to land cover change. Overall, however, the positive trend observed in the Uruguay River streamflow during the second half of the past century should be attributed to increased precipitation, rather than land cover change.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The Uruguay River is the second largest tributary (by discharge) of la Plata River. It stretches over parts of Argentina, Brazil, and Uruguay, and has a drainage area of nearly 3.65 × 105 km2 (Figure 1). As in most of southeastern South America, the Uruguay basin experienced a positive trend in precipitation during the late twentieth century [Barros et al., 2000, 2007], and stream discharge increased as well [García and Vargas, 1998; Tucci and Clarke, 1998]. The Uruguay River has also experienced a trend towards increased flood frequency during the last three decades of the century [Camilloni, 2005]; seven of the ten highest daily streamflow peaks of the 1950–2000 period occurred after 1983. Over the same period, land use has changed substantially. These changes occurred mainly after 1970 and were related to agriculture practice variations [Tucci, 2003], suggesting that the main change over the region was related to crop areas expansion (J. Adamoli, personal communication, 2008).

image

Figure 1. The Uruguay River basin with closing point at Concordia. Gauge stations represented by black triangles: El Soberbio (SOB), Paso de los Libres (PDL) and Concordia (CON). The upper, middle and lower parts of the basin are denoted with UB, MB and LB, respectively.

Download figure to PowerPoint

[3] Conversion of forest to (unirrigated) agriculture is generally expected to increase streamflow. However, Berbery and Barros [2002] and Tucci [2003] argued that most of the observed streamflow trends in the larger la Plata basin should be attributed to the rainfall trends, although they did not discard a secondary contribution from land use change. We focus here on the Uruguay River, where the role of land cover change is better documented than elsewhere in la Plata basin, and where climate networks are generally more dense. In particular, we evaluate the extent to which observed streamflow changes can be attributed to land use change, as opposed to climate.

2. Methodology and Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[4] The Variable Infiltration Capacity (VIC) macroscale hydrology model [Liang et al., 1994, 1996; see also Nijssen et al., 1997] was used to quantify the sensitivity of the Uruguay River streamflow on atmospheric (temperature and precipitation) and land use changes and to assess its land use change impact by simulating two periods: the 1960–1969 decade (60s) and the 1990–1999 decade (90s). The model was implemented over the basin at a resolution of 0.125 degree.

[5] The meteorological input data were daily precipitation, maximum and minimum temperature and wind speed. Downward solar and longwave radiation, and surface humidity, were estimated from the daily temperature and temperature range following methods described by Maurer et al. [2002]. Station precipitation data were taken from the Argentine National Weather Service, the National Meteorological Direction of Uruguay and the Brazilian National Water Agency. The combined data set includes 368 precipitation series for the 90s decade (Figure 2a) and 130 for the 60s, both with a fairly homogeneous spatial resolution. Daily minimum and maximum temperatures included 51 stations during the 90s (Figure 2b) and 16 in the 60s and were obtained from the U.S. National Climatic Data Center (NCDC/NOAA). The number of stations was different for each year, but they generally followed a similar spatial distribution. Precipitation and temperature data were gridded to 1/8 degree spatial resolution using methods outlined by Maurer et al. [2002]. Daily wind speeds were taken from the NCEP-NCAR Reanalysis [Kalnay et al., 1996] and were interpolated to 1/8 degree spatial resolution. Observed streamflows were provided by the Argentine Water Resources Secretary. Three stations were used: El Soberbio in the upper basin, Paso de los Libres in the middle basin and Concordia in the lower basin (see Figure 1). Vegetation information was obtained from the University of Maryland's 1 km Global Land Cover dataset [Hansen et al., 2000], which considers 14 land cover classes of vegetation and, since it was developed during the 90s, it was used as representative of the land use for that period.

image

Figure 2. Spatial distribution of the daily (a) precipitation and (b) minimum and maximum temperature stations available for the period 1990–1999.

Download figure to PowerPoint

[6] Accuracy of the model simulations was assessed using the coefficient of determination R2 and the Nash-Sutcliffe Coefficient of Efficiency (NSE). An NSE value of 1 indicates a perfect match between observations and simulations, and earlier studies suggest values greater than 0.5 are indicative of a plausible model performance [Santhi et al., 2001]. VIC model parameters were estimated using data from 1995–1999 and validated on the 1990–1994 period.

[7] Sensitivity tests were performed to assess the impacts of temperature, precipitation and land cover changes on the Uruguay River streamflows. The tests consisted of: a) varying daily minimum and maximum temperatures (increasing and decreasing 1°C and 2°C on each grid point with respect to the daily 90s temperature data), b) modifying precipitation (increasing annual precipitation at each grid point by 100 and 200 millimeters which, given that the basin annual mean precipitation is about 1800 millimeters, roughly represents 5 and 10% increases, respectively), and c) altering land cover. Variations in annual precipitation were performed by distributing the increases proportionately on each rainy day, thus increasing precipitation in each event without modifying precipitation sequencing.

[8] Land use sensitivity tests considered two extreme cases: the entire basin covered by grassland (vegetation type 10 in the University of Maryland data set) and the basin totally covered by wooded grasslands (vegetation type 7). The absence of trees and shrubs in the first vegetation type implies a minimum in surface evapotranspiration and, consequently, a maximum in runoff, with the opposite occurring in the later vegetation type scenario.

[9] To quantify the impacts of land use change throughout the second half of the twentieth century, a run for the 60s decade was performed using model forcing data for that decade but with the 90s land cover. For this comparison, only meteorological stations that were common to both decades were used to produce the gridded data products to avoid the possibility of confounding the results with differences in station coverage. Therefore, these new runs used a smaller number of daily precipitation (101) and temperature (12) stations.

[10] Cross-correlations between the daily simulated and observed streamflows were calculated using the 90s simulations to look for possible lagging errors in the streamflow simulations. The same analysis was then repeated for the 60s decade to determine whether changes in the lags existed between decades (which arguable could be attributed to land use change). These cross-correlations were calculated for the Concordia station only, which was the only station with daily streamflow information for both the 60s and 90s.

3. Evaluation of the Hydrological Model Performance

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[11] Simulated and observed streamflows were compared for annual, monthly, and daily aggregations. Annual simulated flows for the calibration period were in good agreement with observations, with mean differences during the calibration period ranging from 6.1% in Concordia to 1.3% in El Soberbio. Mean differences in the validation period were less than 3% for all three stations. During the calibration period, NSE for monthly aggregations varied from 0.91 at El Soberbio to 0.94 at Concordia and 0.95 at Paso de los Libres, Table 1. In the context of other basins to which VIC and similar models have been applied, these NSE values are quite high [Su et al., 2008; Feyereisen et al., 2007]. The NSE values during the validation period were quite similar. Monthly R2 values were also high, varying between 0.93 and 0.96 for the calibration period and between 0.93 and 0.94 for the validation period (see Table 1). Daily simulations were evaluated only at Concordia, as only monthly streamflows were available at the other two stations. NSE values were 0.76 and 0.66 for the calibration and validation periods, respectively, whereas R2 varied between 0.82 and 0.73 for the same time periods. Cross-correlation between the daily simulated and observed streamflow time series were calculated for 1990–1999 period, Table 2. On average, there was a 1-day lag between observed and simulated streamflow peaks, but in general the streamflow peak magnitudes were well represented. Indeed, the model represented very satisfactorily the June 1992 daily streamflow peak, which was the highest observed since daily streamflow observations began and was associated with extensive flooding (observed 37,714 m3 s−1, simulated 38,133 m3 s−1).

Table 1. Statistics of the Monthly Simulations
River StationArea (km2)Altitude (m)CalibrationValidation
NSER2NSER2
El Soberbio80 0001500,910,930,920,93
Paso de los Libres189 000550,950,960,920,92
Concordia240 000100,940,950,940,94
Table 2. Correlation Coefficients Between the Daily Simulation and Observationa
LagR1990–1999R1960–1969
  • a

    Positive lags mean the simulated streamflow precedes the observations.

−50.63360.5872
−40.68930.6337
−30.74230.6838
−20.78880.7332
−10.82590.7776
00.85090.8134
+10.86180.8384
+20.85920.8523
+30.84390.8552
+40.81740.8476
+50.78220.8304

4. Sensitivity Tests

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[12] Table 3 summarizes the sensitivity tests results. The model streamflow sensitivity to land cover change is quite high for the two “end point” scenarios considered: for 100% grassland vegetation, the mean streamflow increased by nearly 45%, whereas assuming fully forested land cover reduces the streamflow by 15–20%. Although these are clearly extreme cases, they are indicative of the potential significance of land use change in the basin.

Table 3. Mean Observed Streamflow Qobs (1990–1999) and Percentage Variations in the Sensitivity Testsa
River StationQobs (m3 s−1)T + 1 (%)T + 2 (%)T − 1 (%)T − 2 (%)
  • a

    Shown are increasing 1°C (T + 1), increasing 2°C (T + 2), decreasing 1°C (T − 1) and decreasing 2°C (T − 2) minimum and maximum daily temperatures, increasing 100 mm (pp + 100) and 200 mm (pp + 200) annual precipitation and considering the basin covered by a wooded grassland (Veg 7) and by grasslands (Veg 10).

El Soberbio2805−4−7+3+7
Paso de los Libres4972−3−5+4+8
Concordia5692−1−4+5+9
River StationQobs (m3 s−1)pp + 100 (%)pp + 200 (%)Veg 7 (%)Veg 10 (%)
El Soberbio2805+10+21−15+45
Paso de los Libres4972+12+24−19+45
Concordia5692+14+27−18+46

[13] Modifications in land cover imply changes in the annual streamflow regime. Figure 3 shows the mean annual hydrograph for the 90s decade at Concordia along with the modeled mean annual hydrograph with current land use, and the two endpoint cases. An interesting result is that during the warm season (DJF) the streamflow in the completely wooded grassland scenario is very similar to present conditions. This suggests that during this season, evapotranspiration is highly moisture limited, so differences in potential vegetation between wooded and grassland land cover are relatively unimportant. On the other hand, simulated streamflow for current land cover becomes separated from the wooded condition and becomes closer to the grassland curve in winter and early spring, suggesting that during these seasons, the current (mostly grassland and agriculture) land cover behaves more like the 100% grassland scenario.

image

Figure 3. Observed (Obs) monthly mean streamflow in the 1990–1999 decade and simulated streamflow using the actual land cover (Mod), considering the basin covered by wooded grasslands (Mod type 7) and by grasslands (Mod type 10). Units are m3 s−1.

Download figure to PowerPoint

5. Land Use Change Implications

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[14] The mean 90s streamflow at Concordia was 5,790 m3 s−1 and modeled was 5,670 m3 s−1, while in the 60s period the equivalent values were 4,310 m3 s−1 and 4,310 m3 s−1 respectively. In both cases, differences were well below the possible error of the model simulation or measurements. Therefore, since the streamflow of the 60s was almost exactly reproduced by the model with the vegetation cover of the 90s, we conclude that land use change did not change the mean (monthly) flows (which we would have expected would occur due to reduced evapotranspiration over the basin). Hence, we conclude that the observed increase in the streamflow of the Uruguay River during the second half of the twentieth century has to be attributed to increased precipitation and not to land use change.

[15] Even if the mean streamflow was not affected by land use change, the hydrographs could be affected for shorter time steps. Figure 4 displays the monthly mean simulated and observed streamflow at Concordia for the 1960–1969 and 1990–1999 periods. There is a different behavior in the two decades: during the 60s, the observed hydrograph shows a smoother behavior, with a unique peak in September and much lower values during the December–June time period. In the 90s, on the other hand, the hydrograph has several peaks. Furthermore, the streamflow peak observed in September during the 60s is displaced to October in the 90s. Overall, the observed streamflow is considerably higher in the 90s than in the 60s. As noted, the model fits the observations quite well in the 90s but differences from observations are larger for the 60s observed hydrograph. Indeed, differences between the observations and the simulations are especially apparent in January–February, March–April and September-October-November, when the modeled streamflow has an abrupt change in its monthly trend while the observed streamflow is similar for one more month. Since this behavior was not found during the 90s, it might be attributable to land use change. To examine this point further, daily streamflow data were analyzed in both periods (1960–1969 and 1990–1999) to determine if changing vegetation did have an impact on streamflow.

image

Figure 4. Monthly mean observed (Obs, solid lines) and simulated (Mod, dashed lines) streamflows for the (a) 1960–1969 decade, and (b) 1990–1999 decade. Units are m3 s−1.

Download figure to PowerPoint

[16] As noted above, the 90s R2 and NSE daily statistics for Concordia were 0.72 and 0.84 respectively. For the 60s, the values were 0.66 and 0.57, respectively. This suggests that daily simulations may have been more influenced by changes in land use than monthly simulations, since both NSE and R2 are reduced more between the two decades for daily than for monthly values. The possibility of land cover-related daily streamflow changes was further evaluated through cross-correlations of observed and simulated flows, to determine whether streamflow response characteristics have changed over time. Modeled and observed streamflow during the 90s were lagged by about one day, with the observed streamflows following the modeled ones. This difference was amplified to 3 days in the 1960–1969 decade (Table 2), suggesting a changed behavior in the basin response to runoff originated upstream of Concordia. Stated otherwise, the use of the 90s land cover in the 60s simulation made the model generate more rapid surface discharge than really occurred, when the greater abundance of trees and shrubs arguably were leading to slower runoff. However, the mean modeled and observed streamflows were roughly the same in spite of the documented land cover change, which suggests that lag differences were only because of changes in the timing of the runoff response, and not in evapotranspiration which would affect runoff volumes.

6. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[17] The VIC model fits well the observed mean Uruguay River discharge at monthly and daily streamflows. The mean discharges were simulated with errors less than 3% and the explained variance of the monthly values exceeded 90%. At the daily scale, the model has a mean lag of only one day ahead of the observations.

[18] A sensitivity assessment showed that within the bounds of potential possible changes, precipitation impacts on streamflows were considerably larger than those of temperature. When endpoint scenarios of land use were considered, namely the entire basin covered either with forest or grasslands, the potential impacts on streamflows were quite large, about a factor of two in the latter case. These results imply that potential future changes in precipitation and temperature could be either enhanced or even reversed in the streamflow response depending on concurrent land use changes.

[19] However, VIC model simulations suggest that the positive trends observed in the mean streamflow of the Uruguay River during the second half of the last century are attributable to increased precipitation over the basin and not to land use change. This implies that land use changes were not large enough to produce appreciable changes in basin runoff at monthly and annual aggregation levels, probably at least in part because the changes were basically from grassland (pasture) to crops and not from deforestation. However, while changes in mean flows seem to be attributable almost entirely to precipitation, rather than land cover change, there has been a change in basin response – flows at the basin outlet now occur about two days sooner than in the 60s. Because timing of precipitation is accounted for in the model forcing data, this change appears to be solely attributable to land cover change between the 1960s and 1990s.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[20] The research reported in this paper was funded in part by grant PICT 07-14202 from the ANPCyT and grant 01X199 from UBACyT to the University of Buenos Aires and also by grant NNG04GD12G from the National Aeronautics and Space Administration, and grant EAR-0450209 to the University of Washington. The authors would like to thank the two anonymous reviewers for their useful suggestions and commentaries. The assistance of Fengge Su and Ted Bohn in the VIC model implementation is kindly appreciated.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information
  • Barros, V., M. E. Castañeda, and M. Doyle (2000), Recent precipitation trends in southern South America east of the Andes: An indication of climatic variability, in Southern Hemisphere Paleo- and Neoclimates: Key Sites, Methods, Data and Models, edited by P. P. Smolka, and W. Volkheimer, pp. 187206, Springer, Heidelberg, Germany.
  • Barros, V., M. Doyle, and I. Camilloni (2007), Precipitation trends in southeastern South America: Relationship with ENSO phases and the low-level circulation, Theor. Appl. Climatol., in press.
  • Berbery, E. H., and V. R. Barros (2002), The hydrologic cycle of the La Plata basin in South America, J. Hydrometeorol., 3, 630645.
  • Camilloni, I. A. (2005), Extreme flood events in the Uruguay River of South America, in VAMOS! Newsl., number 2, pp. 2325, Clim. Variability and Predict., Southampton, U. K. (Available at www.clivar.org/organization/vamos/Publications/vamos_nl2.pdf).
  • Feyereisen, G. W., T. C. Strickland, D. D. Bosch, and D. G. Sullivan (2007), Evaluation of SWAT manual calibration and input parameter sensitivity in the Little River Watershed, Trans. ASABE, 50, 843855.
  • García, N., and W. M. Vargas (1998), The temporal climatic variability in the ‘Rıacute;o de la Plata’ basin displayed by the river discharges, Clim. Change, 38, 359379.
  • Hansen, M., R. DeFries, J. R. G. Townshend, and P. Sohlberg (2000), Global land cover classification at 1 km resolution using a decision tree classifier, Int. J. Remote Sens., 21, 13311364.
  • Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437471.
  • Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges (1994), A simple hydrologically based model of land surface water and energy fluxes for general circulation models, J. Geophys. Res., 99(D7), 14,41514,428.
  • Liang, X., E. F. Wood, and D. P. Lettenmaier (1996), Surface soil moisture parameterization of the VIC-2L model: Evaluation and modifications, Global Planet. Change, 13, 195206.
  • Maurer, E. P., A. W. Wood, J. C. Adam, D. P. Lettenmaier, and B. Nijssen (2002), A long-term hydrologically based data set of land surface fluxes and states for the conterminous United States, J. Clim., 15, 32373251.
  • Nijssen, B., D. P. Lettenmaier, X. Liang, S. W. Wetzel, and E. F. Wood (1997), Streamflow simulation for continental-scale river basins, Water Resour. Res., 33, 711724.
  • Santhi, C., J. G. Arnold, J. R. Williams, W. A. Dugas, R. Srinivasan, and L. M. Hauck (2001), Validation of the SWAT model on a large river basin with point and nonpoint sources, J. Am. Water Resour. Assoc., 37, 11691188.
  • Su, F., Y. Hong, and D. P. Lettenmaier (2008), Evaluation of TRMM Multi-satellite Precipitation Analysis (TMPA) and its utility in hydrologic prediction in la Plata basin, J. Hydrometeorol., in press.
  • Tucci, C. E. M. (2003), Variabilidade climática e o uso do solo na bacia brasileira do Prata (Climate variability and land use in the Brazilian Plata basin), in Climae Recursos Hídricos no Brasil (Climate and Water Resources in Brazil), Coleção ABRH, vol. 9, edited by C. E. M. Tucci, and B. P. F. Braga, pp. 163242, Assoc. Bras. de Recursos Hidricos, Porto Alegre, Brazil.
  • Tucci, C. E. M., and R. T. Clarke (1998), Environmental issues in the la Plata basin, Int. J. Water Resour. Dev., 14, 157173.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methodology and Data
  5. 3. Evaluation of the Hydrological Model Performance
  6. 4. Sensitivity Tests
  7. 5. Land Use Change Implications
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
grl24525-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
grl24525-sup-0002-t02.txtplain text document0KTab-delimited Table 2.
grl24525-sup-0003-t03.txtplain text document1KTab-delimited Table 3.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.