Strongly Coupled Land–Atmosphere Data Assimilation and Its Influence on Weather Forecasting

Accurate prediction of near-surface atmospheric variables, boundary layer conditions, and precipitation remains a challenge in modern numerical weather prediction (NWP). Coupled land-atmosphere data assimilation could help to mitigate forecast errors. In the past, the most common method of coupling was via a weakly-coupled data assimilation framework, with which the land (e.g., soil moisture) data assimilation was carried out via the land-surface model integration. Meanwhile, strongly-coupled data assimilation requires the estimation of cross-model background error covariances and the simultaneous correction of atmospheric and land surface model states during the analysis procedure. Due to uncertainties in numerical models in representing the land-atmosphere interfaces, strongly-coupled data assimilation becomes a challenging problem; thus, it has not been commonly used up to now.

In this study, we made significant attempts to address these challenges from the strongly-coupled data assimilation. We used the Noah land surface model coupled with the mesoscale community Weather Research and Forecasting (WRF) model to examine the effectiveness of strongly-coupled data assimilation. This was done by assimilating satellite-derived soil moisture data on the weather forecasts over the US southern Great Plains (SGP) and compares the results with a weakly-coupled data assimilation framework. First, the spatial and temporal variability of the background error covariance between the land surface soil moisture and atmospheric states in strongly coupled land-atmosphere data assimilation were characterized. Results show that the forecast errors in the top-10-cm soil moisture and near-surface air conditions (e.g., potential temperature and specific humidity) are correlated and relatively large during the daytime in the summer. The magnitude of the error correlation between these three states is comparable, suggesting that assimilation of satellite soil moisture data could provide cross-variable impacts similar to those assimilating conventional near-surface temperature and humidity data. Following this, we assimilated 9-km SMAP level-2 enhanced soil moisture data into the WRF-Noah model. Results prove that the strongly coupled soil moisture data assimilation outperforms the weakly-coupled data assimilation in regards to forecast errors of near-surface atmospheric variables, boundary layer conditions, and precipitation.

Following the above results, we developed a strongly-coupled land-atmosphere data assimilation system based on the NCEP GSI-based ensemble Kalman filter data assimilation system. Soil moisture states at all soil levels in the WRF-Noah model were added as the analysis variables, along with all other atmospheric analysis variables. Preliminary evaluation shows that: 1) Including soil moisture as the analysis variable helps to improve the analysis and short-range forecasts of near-surface atmospheric variables and boundary layer conditions. 2) Simultaneous assimilation of atmospheric observations and soil moisture data in the strongly coupled land-atmosphere data assimilation outperforms either soil moisture data assimilation or atmospheric data assimilation alone in terms of the analysis and forecasts of atmospheric conditions and soil states. Detailed results will be presented at the conference.