Dynamical seasonal prediction including the effects of land surface anomalies

Dynamical seasonal predictions in hindcast mode using global atmospheric general circulation models have shown that "perfect" knowledge of sea surface temperature can contribute to seasonal predictability, particularly in extreme phases of the ENSO cycle during boreal winter. Results from boreal spring and summer, when land surface conditions play a larger role in climate, have been less impressive. We are examining the role that knowledge of land surface conditions (soil moisture and snow cover) can play in improving predictability. We employ a similar multi-year suite of ensemble forecasts using the best possible land surface model initial conditions. These are generated offline by driving the same land surface scheme as coupled to the GCM with observed/analyzed near surface meteorology in a manner similar to that of the GEWEX Global Soil Wetness Project. Thus, in addition to sea surface temperature, realistic interannually-varying land surface conditions are initialized for each forecast. Furthermore, we are examining how systematic errors in simulated precipitation and radiation degrade predictability through their impact on the land surface. Upper-limit tests of predictability are also being performed with specified land surface boundary conditions throughout the integrations.