Skill in streamflow forecasts derived from large-scale estimates of soil moisture and snow

Seasonal predictions of streamflow can benefit from knowledge of the amounts of snow and other water present in a basin when the forecast is issued(1-5). In the American west, operational forecasts for spring-summer streamflow rely heavily on snow-water storage and are often issued at the time of maximum snow accumulation. However, forecasts issued earlier can also show skill, particularly if proxy information for soil moisture, such as antecedent rainfall, is also used as a predictor(1,4). Studies using multiple regression approaches and/or model-produced streamflows(6-9) indeed suggest that information on soil moisture-a relatively underappreciated predictor-can improve streamflow predictions. Here, we quantify the relative contributions of early-season snow and soil moisture information to the skill of streamflow forecasts more directly and comprehensively: in a suite of land-modelling systems, we use the snow and soil moisture information both together and separately to derive seasonal forecasts. Our skill analysis reveals that early-season snow water storage generally contributes most to skill, but the contribution of early-season soil moisture is often significant. In addition, we conclude that present-generation macroscale land-surface models forced with large-scale meteorological data can produce estimates of water storage in soils and as snow that are useful for basin-scale prediction.