It may be possible to estimate surface fluxes of scalar quantities from measurement of their variance and mean wind speed. The flux-variance relation for temperature and humidity was investigated over prairie and desert-shrub plant communities. Fluxes were measured by one-dimensional eddy correlation, humidity by fast-response wet-bulb psychrometers and Krypton open-path hygrometers, temperature by fine-wire thermocouples, and mean windspeed by a cup anemometer. The quality of the flux-variance relation proved to be good enough for application to flux measurement. Regressions of flux estimated by the variance technique versus measured flux usually had r2 values greater than 0.97 for sensible heat flux and greater than 0.88 for water vapor flux. More uniform surfaces tended to yield the same flux-variance relations except when fluxes were small. This exception supported the hypothesis that sparse sources of flux may increase variance downwind. Nonuniform surfaces yielded flux-variance relations that were less predictable, although reasonably accurate once determined. The flux-variance relation for humidity was quite variable over dry surfaces with senescent vegetation.