Surface analyses of weather parameters such as temperature, relative humidity, and wind depend on the quality of the observations used to create those analyses. The adjoint of a two-dimensional variational analysis system of surface weather parameters covering the entire continental United States has been developed to assess objectively the sensitivity of the resulting analyses to the observations used in the analysis. The analysis system uses the 5 km resolution background fields used by the Real Time Mesoscale Analysis (RTMA) developed by the National Centers for Environmental Prediction. Roughly 12000 observations available each hour from many different mesonets are then used to modify the background grids and obtain hourly analyses. The sensitivity of the differences in weather parameters between the resulting analyses and the background fields are examined as a function of the various data assets. Attention is placed on how to identify observations that have a large impact due to poor quality observations as opposed to ones that have a large impact due to actual weather departures from the background fields. Stations that exhibit large representative errors often arising as a result of differences in elevation between the background fields and observations are also examined.