A two-dimensional variational analysis system is used over a limited domain (400 km x 400 km) to examine the sensitivity of mesoscale surface analyses to the types of mesonet observations assimilated. Analysis errors assessed by data denial experiments are determined for analyses that rely on surface aviation reports only vs. those that use mesonet observations in urban areas. The sensitivity of the analysis error is evaluated as a function of: (1) the magnitude of the observational error variance of the surface observations relative to the background error variance and (2) the decorrelation length scales of the background error field. Implications for the use of mesonet observations from disparate sources for applications in urban environments are discussed.