Impact of Land Initialization on Short-Term Forecast Skill During Episodes of Wet and Dry Land-Atmosphere Coupling

Abstract:

We use output from the North American Land Data Assimilation System Phase 2 (NLDAS-2) and Weather Research and Forecasting (WRF)_Version 3.6 - both coupled to the Noah Unified land model - to investigate the short-term forecast error covariance structure of surface and atmospheric variables linked in the local coupling process chain, which spans the soil moisture-precipitation feedback pathway. Quantities of interest include: soil moisture, temperature, humidity, surface turbulent heat fluxes, convective available potential energy, lifting condensation level, planetary boundary layer (PBL) height, total column integrated water vapor, and precipitation. Using the five longest sustained periods of wet and dry coupling in the recent period of record at U.S. DOE's Atmospheric Radiation Measurement (ARM) Program Southern Great Plains (SGP) site, we produce analogue error structures of default WRF. These analogues are then used to point to aspects/vertical model layers where the greatest gains in model forecast skill can be expected during like coupling regime events at the SGP. Finally, the same wet- and dry-coupling periods are run with the initial soil moisture and temperature prescribed from NLDAS-2 Noah output. These runs provide analogues of model improvement magnitudes, which we evaluate in the context of the land-atmosphere coupling process chain.

In this study, two key and highly interdependent issues are the land initialization sequence protocol (especially when the perturbation is of considerable magnitude) and configuration-related systematic biases in WRF tied to PBL, convective parameterization, and (Noah) land schemes. We provide guidance on both issues for simulations at SGP. Insights gained from our study are directly applicable to future fully-coupled land data assimilation runs over the region.