Environmental conditions at a site can often be inferred from biological observations.  The application of this technique requires that one develop a transfer function relating biological assemblage composition with particular environmental conditions.  Most transfer functions are tested and validated by assessing the accuracy of inferences across different streams.  Here, we examine the accuracy with which biologically-based environmental inferences reproduce temporal trends within single sites.  More specifically, we compare direct measurements and inferences of temporal trends in stream temperature at several sites from a restoration project in northeastern Oregon, USA.  Biological samples and continuous temperature data from these sites were collected regularly for approximately 10 years, providing a rare opportunity to assess within-stream temporal trends.  In general, temporal trends estimated from biologically-based inferences were very similar to trends estimated from direct measurements.  Furthermore, combining direct measurements and inferences yielded more precise estimates of temporal trends at each site.  This analysis further demonstrates the utility of biologically-based environmental inferences for monitoring environmental conditions in aquatic systems.