Characteristics of Background Error Covariance of Soil Moisture and Atmospheric States in a Strongly Coupled Land-Atmosphere Data Assimilation: The Influences on SMAP Data Assimilation

Remotely-sensed soil moisture data has been incorporated into numerical weather models for improving weather forecasts. The most common way is via a weakly-coupled data assimilation (WCDA) framework, with which the soil moisture data assimilation is only seen via the land-surface model integration. A strongly-coupled data assimilation (SCDA) requires the estimation of cross-model background error covariance 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-atmospheric interfaces, SCDA becomes a challenging problem. In this study, we estimate the cross-model spatiotemporally-varying background error covariance between land surface soil moisture and atmospheric temperature and humidity according to the National Meteorological Center method by collecting forecast samples with different lead times but valid at the same time. We conducted the forecast experiments using the Weather Research and Forecasting (WRF) with the Noah land surface model from 2015 to 2017 over the contiguous United States. The results show that the forecast errors in top-10-cm soil moisture and air temperature and humidity are correlated and relatively large during the daytime in the summer. During the summer, the forecast errors in surface soil moisture are correlated with those of atmospheric states up to a sigma pressure level of 0.9 (approximately 900 hPa for a sea-level location) with a domain-mean correlation of -0.15 and 0.1 for temperature and humidity, respectively. Then, we use the estimated background error covariance to assimilate 9-km Soil Moisture Active Passive (SMAP) level-2 enhanced soil moisture data into the WRF-Noah model using a variational approach. Over the Great Plains, the results show that WCDA results in a reduction of bias in temperature and humidity forecasts by 7.8% and 20.5%, respectively, while SCDA can provide additional 2% and 4% reductions.