Streamflow prediction constitutes a fundamental tool in water resources decision-making and risk management actions. The possibility of implementing a mathematical model for the prediction of streamflow in an Andean basin was studied, identifying, through Pearson's correlation coefficient, the best correlations between the time series of mean monthly streamflow (Qm) and the sea surface temperature (SST), considering up to 11 lags. The monthly SST data were obtained from the MODIS-NASA sensor processed on the Ocean Color platform, selecting cells of 2° longitude and 4° latitude to cover the strip +180° lon / -180° lon, from -20 °lat / +20° lat, analyzing a total of 3600 cells. The water resource was characterized by the Qm time series of the La Donjuana station on the Pamplonita river (Norte de Santander, Colombia). The time window studied was from July 2002 to December 2015 (162 months). Linear models were built for each month by selecting the lag that produced the maximum correlation and verifying values of the p statistic, which were much lower than 0.001. The models were evaluated using the mean square error and the Nash-Sutcliffe efficiency, differentiating Normal years, El Niño years, and La Niña years. Satisfactory results were found for the prediction in La Niña years (wet) with lags greater than three months. The investigation is expected to be extended to consider a greater range of latitudes and to consider other Andean basins.