Six simulated long-term land surface assimilation data sets, namely three retrospective “offline” datasets based on the Noah land surface model (LSM), VIC LSM and CPC leaky bucket soil model respectively, and three Reanalysis datasets (North American Regional Reanalysis, NCEP/DOE Global Reanalysis and ECMWF ERA40), are used to study the spatial and temporal variability of soil moisture and their representation of the major land surface hydrological events. The offline runs are forced by near-surface observations. The results show that the six land data assimilation systems can fairly well reproduce the observed soil moisture variations, especially the offline runs, at least when compared to observations over Illinois, one of the few places with observations. The study is focused on the role of dominant land surface hydrological processes, the origin of low frequency variations in soil moisture and the magnitude and partitioning of the land surface water balance component (Precipitation, Evaporation and Runoff) variations on seasonal to inter-annual time scales. Inter-comparison of six land surface data sets over the US as a whole on multi-decadal time-scale indicates that the six land data assimilation systems have the ability to capture the large extreme hydrological events, such as the 1988 summer drought and the 1993 summer flood, and can be used to assess the severity and distribution of droughts and floods. Further study reveals that in the warm season large and well organized soil moisture anomalies do have important impacts on the temperature anomalies for the next one to two months.
Supplementary URL: