Effect of Realistic Soil Moisture Initialization on the Canadian CanCM3 Seasonal Forecast Model

This paper presents the results of a direct comparison of sub-seasonal (60-day) forecast skill using two different land surface initializations in the Canadian Climate Centre for Modelling and Analysis (CCCma) CanCM3 coupled global climate model. The first land surface initialization uses randomized values of soil moisture whereas in the second case the model is initialized with a "best-estimate" of soil moisture derived from offline land surface model simulations. In this experiment the realistic soil moisture initialization improved temperature forecast skill during the boreal summer. Improvement was particularly evident for the wettest and driest quartiles of soil moisture initial conditions. Certain geographic regions, such as North America, showed the greatest improvement in temperature forecast skill. In contrast to temperature forecasts, there was much less skill improvement in precipitation forecasts between the two different soil moisture initializations, although there are geographic regions, such as North America, that do show increased skill.