Long-range empirical forecasts of North Atlantic anomalous conditions are issued, using sea ice concentration anomalies in the same region as predictors. Conditions in the North Atlantic are characterized by anomalies of sea surface temperature, of 850 hPa air temperature and of sea level pressure. Using the Singular Value Decomposition of the cross-covariance matrix between the sea ice field (the predictor) and each of the predict and variables, empirical models are built, and forecasts at lead times from 3 to 18 months are presented. The forecasts of the air temperature anomalies score the highest levels of the skill, while forecasts of the sea level pressure anomalies are the less sucessful ones. To investigate the sources of the forecast skill, we analyze their spatial patterns. In addition, we investigate the influence of major climatic signals on the forecast skill. In the case of the air temperature anomalies, the spatial pattern of the skill may be connected to El Nino Southern Oscillation (ENSO) influences. The ENSO signature is present in the predictor field, as shown in the composite analysis. The composite pattern indicates a higher (lower) sea ice concentration in the Labrador Sea and the opposite situation in the Greenland-Barents Seas during the warm (cold) phase of ENSO. The forecasts issued under the El Nino conditions show improved skill in the Labrador region, the Iberian Peninsula and south of Greenland for the lead times considered in this paper. For the Great Lakes region the skill increases when the predictor is under the influence of a cold phase. Some features in the spatial structure of the skill of the forecasts issued in the period of the Great Salinity Anomaly present similarities with those found for forecasts made during the cold phase of ENSO. The strength of the dependence on the Great Salinity Anomaly makes it very difficult to determine the influence of the North Atlantic Oscillation.