<p>The second phase of the Global Land-Atmosphere Coupling Experiment (GLACE-2) is a<br /> multi-institutional numerical modeling experiment focused on quantifying the subseasonal (out<br /> to two months) forecast skill for precipitation and air temperature derived from the realistic<br /> initialization of land surface states, notably soil moisture. An overview of the experiment and<br /> consensus model behavior at the global scale is described here. The models show modest but<br /> significant skill in predicting air temperatures, especially where the rain gauge network is dense;<br /> assuming rain gauge density is a reasonable proxy for the adequacy of the observational network<br /> contributing to soil moisture initialization, this result indeed highlights the potential contribution<br /> of enhanced observation to prediction. Precipitation forecast skill is much weaker than that for<br /> air temperature. The skill for predicting air temperature, and to some extent precipitation,<br /> increases with the magnitude of the initial soil moisture anomaly. GLACE-2 results are<br /> examined further to provide insight into the asymmetric impacts of wet and dry soil moisture<br /> initialization.</p>