Impact of Initial Soil Wetness on Seasonal Atmospheric Prediction

This study investigates the importance of initial soil wetness in seasonal predictions with dynamical models. Two experiments are performed, each consisting of two ensembles of global climate model integrations initialized from early June observed atmospheric states. In each experiment the only difference between the two ensembles is that they are initialized with a different soil wetness. In the first experiment both ensembles are initialized from 1988 observed atmospheric states and use observed 1988 SST; one ensemble is initialized with seasonally varying climatological soil wetness, and the other is initialized with proxy 1988 soil wetness derived from the European Centre for Medium-Range Weather Forecasts analysis–forecast system. In the second experiment the two ensembles are initialized from observed atmospheric states and use observed SST for five different years, and each ensemble is initialized with a different climatological soil wetness. After initialization, a coupled atmosphere–biosphere model determines the evolution of the soil wetness fields in all the integrations. The experiments are analyzed to determine the impact of the initial soil wetness differences. In contrast to several previous studies in which initial soil wetness was prescribed arbitrarily, a somewhat more realistic soil wetness impact is analyzed by comparing integrations initialized with climatological soil wetness to integrations initialized with soil wetness derived from the output of an operational analysis–forecast model. The initial soil wetness impact is found to be largely local and is largest on near-surface fields, in agreement with previous results. Significant impacts were found in several tropical and extratropical regions in both experiments. Almost all the regions that had significant increases (decreases) in initial soil wetness had significant increases (decreases) in seasonal mean evaporation and significant decreases (increases) in seasonal mean surface air temperature. Half of the regions had significant increases (decreases) in seasonal mean precipitation in response to increased (decreased) initial soil wetness, though the response of the precipitation was more variable and was highly dependent on the response of the moisture flux convergence to the initial soil wetness anomaly. In order for an initial soil wetness difference to force a significant seasonal mean precipitation difference in a region, it must effectively alter the mean convective stability of the region and thereby the mean convective precipitation. The strength of the impact of initial soil wetness differences, as well as the nature of the impact on precipitation and other atmospheric fields, depends on several factors. These factors include the areal extent and magnitude of the initial soil wetness difference, the persistence of the soil wetness difference, the strength of the solar forcing, the availability of nearby moisture sources, and the strength of the regional dynamical circulation. The results suggest that seasonal atmospheric prediction could be enhanced by using a realistic initial state of soil wetness.