Response of short term precipitation to initial soil states in WRF-ARW model

Despite agreement at large-scale and synoptic scales in numerical weather prediction (NWP) models, parametric uncertainties lead to significant scatter in predictions of planetary boundary layer structure and mesoscale sensible weather phenomena. Of particular importance for the disagreement are errors induced through land-atmosphere interactions. Three main land surface properties govern the interactions: albedo (radiative transfer), surface roughness (momentum transfer), and evapotranspiration (heat transfer). evapotranspiration, as the integrator in soil-precipitation feedback, is particular important for short-term (a few days) numerical weather predictions. Errors in land surface forcing and parameterizations tend to accumulate in soil properties (moisture and temperature), which leads to incorrect surface water and energy partitioning. The biases induced by land surface forcing and parameterizations can be constrained through a high-resolution land surface data assimilation (HRLDAS) system, which has the ability to maximize the utility of available land surface observations. However, uncertainties in land surface parameters (e.g., surface roughness length, vegetation fraction, albedo and stomatal resistance) that usually prescribed in land surface models still present challenging in predicting short-term weather.

In this study, with aid of available high-resolution satellite data (AVHRR and MODIS), physics-based land surface parameter perturbation is used to construct ensemble members for a Noha-based HRLDAS system. By perturbing sensitive land surface parameters like fractional vegetation cover, albedo, roughness and stomatal resistance, 20 realizations are achieved by running HRLDAS for two regions with quite different land cover: the mountainous desert area in Utah and the Northeastern region of the United States. The HIRLDAS ensemble simulations are first examined against available observations. Then an ensemble of WRF simulations forced with individual HIRLDAS member is conducted to investigate the responses of the planetary boundary layer structure and near-surface meteorological variables. The study particularly focuses on understanding soil-precipitation feedback and sensitivity of evapotranspiration to perturbed land surface parameters. Statistical analysis shows the ensemble simulations well capture the uncertainty spread in land surface model forcing and parameterizations. The inclusion of land surface forcing (soil properties, sensible and latent heats) diversity improved forecast probability skill. This indicates a general need in mesoscale ensemble prediction models to considering uncertainties in land surface states.