Results are presented from the second phase of the multi-institution partnership North American Land Data Assimilation System (NLDAS-2). The NLDAS runs multiple land surface models (LSMs) over the conterminous U.S. (CONUS) retrospectively and in realtime in support of the drought analysis, monitoring and forecasting activities of the National Integrated Drought Information System (NIDIS). The NLDAS-2 builds upon the first phase NLDAS-1 framework through improvements to the accuracy and consistency of the surface atmospheric forcing data, upgrades to the land surface models, and extension from a 3 year (1997-99) retrospective simulation to 30 years (1979-2008). As the first of two parts, this paper describes the NLDAS configuration, upgrades to the four LSMs, and overall comparisons of the land surface water and energy flux and state variables for the CONUS. The second part focuses on validation of model-simulated streamflow against measurements.
Analysis of the mean annual and monthly water and energy budgets shows that all models conserve energy and water balance well. The results show a higher level of agreement among the four models when compared with those from the NLDAS-1 for most regions of the CONUS. This is due, in part, to recent improvements and developments of the four LSMs that were initiated following NLDAS-1. However, there are still large inter-model differences in the northeast, Lake Superior region and western mountainous regions, which are associated with cold season processes. In addition, the evaporation and subsurface runoff exhibit large inter-model differences due to differences in the depiction of sub-surface hydrology in the four LSMs. These issues are being to be addressed by the model development community including NLDAS multi-institution partners in the future.