The North American Land Data Assimilation System (N-LDAS) project is a multi-institutional effort to produce better land-surface hydrology simulations over the continental US for use in weather and climate models. Several land-surface models (NOAH, Mosaic, VIC, and Sacramento) participate in the project. Simulations from these models differ even when they are driven by exactly the same forcing. Understanding the reasons for these differences will allow us to improve land surface models. The goals of N-LDAS are 1) to test state-of-the-art land surface models for use in data assimilation, and 2) once we have a good model, develop a real-time land surface data assimilation system that uses in situ and remotely-sensed soil moisture, skin temperature, and snow to produce (in real time and later in a reanalysis) an accurate soil moisture data set that can be used for a) retrospective land-memory predictability studies, and b) to initialize real-time coupled model predictions of weather and seasonal climate. We are still in phase 1 of the project. We take advantage of actual observations of soil and vegetation properties, soil moisture, soil temperature, and surface fluxes at more than 40 stations from around the US for the period of 1997-1999 to evaluate the results from the standard runs. The N-LDAS models show reasonable simulations of soil moisture and temperature and fluxes compared to observations from the Oklahoma Mesonet and ARM/CART facility. We also conduct special runs with the local observed forcing and local observed soil parameters. Differences between model output and observations are not due to differences between actual and N-LDAS-specified forcing or random observational errors, but are likely due to soil or vegetation differences and model assumptions. We also evaluate the reasons for the differences between the models in their soil moisture, soil temperature, and surface flux simulations.