A fundamental task of the Community Land Model (CLM; the land component of the Community Earth System Model) is the partitioning of precipitation into evapotranspiration (ET), runoff, and water storage. Testing model performance against site-level observations provides important insight, but can be challenging to extrapolate to the larger spatial scales at which Earth System models typically operate. Traditionally, measurements of river discharge have provided the best, and in many cases only, metrics with which to assess the performance of land models at large spatial scales (i.e. regional to continental scale river basins). Because the quantity of discharge measurements has declined globally, and the human modification and management of rivers has increased, new methods of testing land model performance are needed. As global observations of total water storage (TWS) and ET have become available, the potential for direct assessment of the quality of the simulated water budget exists. In this presentation, we use TWS observations from the GRACE satellite project and upscaled flux tower measurements from the FLUXNET-MTE dataset to assess the performance of CLM parameterizations such as canopy interception, storage, and evaporation, soil evaporation, and soil moisture and groundwater dynamics. We then give examples of alternative model parameterizations, and show how these parameterizations improve model performance relative to GRACE and FLUXNET-MTE based metrics.