Are some RAWS observations more critical to an analysis than others?

The impact of the present distribution of Remote Automated Weather Station (RAWS) observations was assessed objectively in order to demonstrate both the value of the existing network as well as identify potential redundancies. Control analyses valid at 1800 UTC for selected days from May to November 2008 were created using background fields from the National Centers for Environmental Prediction and all available RAWS and National Weather Surface (NWS) surface observations between 1730-1830 UTC. Then, sets of additional analyses for each time were generated using all RAWS and NWS observations except each observation is withheld once. A two-dimensional variational analysis technique was optimized in order to be able to generate thousands of such data denial experiments for 51 subdomains covering the continental United States for analyses of surface temperature, wind and relative humidity. Objective measures are used to evaluate the extent to which the analyses are degraded by the removal of some observations relative to others.