Successful strategic and tactical decisions that support water resource management, agriculture, and hydroelectric power generation in the Americas, especially in monsoon areas, are only possible if precipitation forecasts on subseasonal to seasonal time scales are skillful. Subseasonal variability of precipitation is evident across the globe, including in North, South and Central America. During the boreal summer, large intraseasonal variability extends eastward across the Pacific Ocean over the equatorial and subtropical regions of North America. During the boreal winter, similar patterns of variance exist but with maximum variance in the southern hemisphere with maxima over Brazil and the South Atlantic Convergence Zone. Previous studies have suggested the existence of a link between North and South America intraseasonal variability with that in the Indo-West pacific basin. Despite its importance modulating climate and weather across the globe, numerical prediction models do not forecast skillfully and robustly intraseasonal variability. The skill of empirical predictions scheme is higher, but also have limitations predicting extreme events. In addition, empirical schemes are often deterministic and do not allow for a direct assessment of error growth associated with sensitivity to initial conditions. This proposal provides a hybrid methodology to improve the forecasting skill of extended rainfall predictions in the Americas by combining numerical weather predictions and empirical schemes into a single system capable of providing probabilistic forecasts using the different ensemble members available in numerical weather prediction models. To design an optimal operational hybrid scheme we will first assess the prediction and simulation skill of intraseasonal variability in the Americas from numerical weather prediction and climate models. We will also assess the interannual variability of intraseasonal variability in the Americas, and its potential impact on intraseasonal prediction. The proposed hybrid scheme uses the forecast circulation structure from each ensemble member and separates it into different temporal scales by projecting the numerical forecasts onto the most important multivariate spatio-temporal modes of variability and then uses the banded numerical forecasts as predictors in the hybrid empirical system.