Predictability of Ecuadorian seasonal streamflow in a changing climate

Seasonal variability of streamflow (Q) plays a vital role in developing and managing the water resources of most world regions. The South American continent, in particular, is currently showing increased water stress. Additionally, sea surface temperature (SST), particularly in the tropical Pacific region, is one of the most convenient variables used as a predictor for the climate of Ecuador. Moreover, its location on the equatorial coast of the eastern Pacific places it directly in the center of the area where El Niño phenomenon presents the most pronounced intensity. Therefore, numerous efforts have been initiated in recent years to determine the influence of ENSO on the behavior of hydrometeorological variables in Ecuador, such as precipitation, temperature, and streamflow. The main goal of this work is to explore the feasibility of the seasonal prediction for Ecuador’s streamflow and the development of prediction models for it. To this end, the role of the tropical Pacific SST, taking into account the different ENSO patterns, has been investigated using the singular value decomposition (SVD) method. SVD is a multivariate technique that identifies coupled modes of variability between the SST anomalies and the rivers’ streamflow. Furthermore, for forecasting purposes, seasonal delays (from 1 to 4 seasons) between these fields were established (SVD_Lag), and the “leave one out” cross-validation scheme was applied. Results show that the two main variability modes of the tropical Pacific SST, associated with canonical El Niño (EN) and El Niño Modoki (ENM), can be used to predict the February-March-April (FMA) and June-July-August (JJA) streamflows with up to three lag seasons. For FMA streamflow, the mode associated with El Niño Modoki presents the greatest contribution to Q, with this influence up to for one year. Finally, the experiment of combining both modes (canonical and Modoki) for a statistical model allows the Q prediction for FMA with one year of anticipation in localities placed mainly in the Andes region, while the streamflow of JJA can be predicted up to three seasons in advance in localities sited in the coastal region of Ecuador.