Tropical origins of Weeks 2-4 forecasts errors during Northern Hemisphere cool season

Abstract

AbstractA set of 30-day reforecast experiments, focused on the Northern Hemisphere (NH) cool season (November–March), is performed to quantify the remote impacts of tropical forecast errors on the National Centers for Environmental Prediction (NCEP) global forecast system (GFS). The approach is to nudge the model towards reanalyses in the tropics and then measure the change in skill at higher latitudes as function of lead time. In agreement with previous analogous studies, results show that midlatitude predictions tend to be improved in association with reducing tropical forecast errors during weeks 2-4, particularly over the North Pacific and western North America, where gains in subseasonal precipitation anomaly pattern correlations are substantial. It is also found that tropical nudging is more effective at improving NH subseasonal predictions in cases where skill is relatively low in the control reforecast, whereas it tends to improve less cases that are already relatively skillful. By testing various tropical nudging configurations, it is shown that tropical circulation errors play a primary role in the remote modulation of predictive skill. A time dependent analysis suggests a roughly one week lag between a decrease in tropical errors and an increase in NH predictive skill. A combined tropical nudging and conditional skill analysis indicates that improved Madden Julian Oscillation (MJO) predictions throughout its lifecycle could improve weeks 3-4 NH precipitation predictions.