CONICET | Buscador de Institutos y Recursos Humanos

La Plata basin (LPB) is the fifth largest basin in the world and the second in South America. Much of its economy comes from agriculture and hydroelectric power generation which are both highly vulnerable to precipitation variability and change. Given its socio-economic importance, it is essential to generate information that is relevant to hydrologists and water resources managers. However, one challenge to achieve this goal is that the performance of the classical hydrological distributed models is highly constrained by the complexity of the hydrological system and the limited availability of data to describe it accurately. This constraint is particularly important in large river basins such as LPB where observations are limited. On the contrary, a simple statistical model in which calculations can be undertaken with little computational resources and minimal information on the basin?s state, may be useful to understand more clearly specific aspects of a river basin?s response at a large scale. For instance, an approach using such a low complexity rainfall-runoff statistical model (RRM) may allow to assess the extent to which the climate forcing variability explains runoff variability. The main objective of this study is to apply this approach to properly understand the long-term relationship between three hydroclimatic variables: precipitation, evapotranspiration, and streamflow. For this purpose, a RRM is developed within the unit hydrograph framework under a linear regression setting and using the Turc-Pike empirical estimation to represent evapotranspiration. The model is evaluated over three main rivers of LPB: Paraguay, Parana and Uruguay, considering the period 1931-2010. Overall, river discharges derived by our RRM captured well observed streamflows variability, though significant deficiencies were found in Paraguay River in representing extreme fluctuations. Uruguay sub-basin had shown both the variability and magnitude fairly well represented. The skill of the RRM to represent observed river discharges at the monthly and annual scale is quantified for each sub-basin based on mean square error. Particularly in Parana and Paraguay rivers, RRM captured very accurately some aspects of the annual cycle (at monthly basis, 66% and 47% of the total variance was explained, respectively). Uruguay River simulations were able to explain more than 60% of the total observed variance of monthly runoff and 90% at the annual scale. In addition, RRM performance in Uruguay River basin is compared with the one derived by the state-of-the-art VIC hydrological model. The RRM is found to outperform the VIC model on periods where limited data was available to run the latter, and to perform only slightly worse otherwise, thus highlighting the potential of the RRM for studies of the hydroclimatic variability on large river basins that are poorly observed, and for the generation of future hydrological scenarios under climate change.