Initializing Numerical Weather Prediction Models with Model-Derived and Satellite-Based Soil Moisture Data

Modeling studies show that soil moisture plays a pivotal role in planetary boundary layer (PBL) evolution through positive and negative feedback loops in the local land-atmosphere coupling (LoCo) process chain (Santanello et al, 2011). Variability in the land surface energy balance (i.e. surface sensible and latent heat flux) corresponds directly with that of soil moisture heterogeneity, and is a crucial component in determining the state of the PBL. To accurately account for antecedent soil moisture variability, this study uses NASA’s Land Information System (LIS) to produce offline “spin-ups” with varying meteorological forcing quality and varying greenness vegetation fraction (GVF). The LIS simulations are then used to initialize short-term NASA Unified Weather Research and Forecasting (NU-WRF) forecasts through the coupled LIS-WRF framework. A comparison is made between the LIS spin-ups, Soil Moisture Active Passive (SMAP) satellite data, and in situ soil moisture probes to define any biases that may be present in the seasonality of the LIS runs, and to address the ability of SMAP to convey soil moisture data that follow proximal observations. The coupled LIS-WRF simulations followed observations more closely when using the LIS spin-up with the best-available meteorological forcing and real-time GVF when compared to the control run that was initialized by operational standards. This serves as an early step in using offline spin-ups to initialize forecast models, and it suggests that there is significant impact of the land surface initialization approach on operational NWP forecasts.