Land Cycling for Improving Short-Range Weather Prediction

Land states are closely linked to daily weather, and their impacts are not confined to near-surface atmospheric conditions but extend through the atmosphere. Land or soil variables are constantly changing in time and space, yet they are often not well spun up or properly updated in Numerical Weather Prediction (NWP) systems, especially during atmospheric cycling with data assimilation for extended periods (> 1 month). This talk aims to address two pivotal questions: 1) the influence of cycling land states on short-range weather forecasts, such as temperature and precipitation, and 2) which land state is most critical for improving weather forecast skills. Using the limited-area Model for Prediction Across Scales (MPAS) model over the CONUS domain, we compare forecasts from various cycling experiments with different land states, validating them with respect to independent observations. The talk will conclude with the discussion on our strategy of soil data assimilation.