Isoprene is usually the dominant natural volatile organic compound emission from forest ecosystems, especially those with a major broadleaf deciduous component. Here we report isoprene emission model performance versus leaf and canopy level isoprene emission measurements made at the Duke University Research Forest near Chapel Hill, North Carolina. Emission factors, light and temperature response, canopy environment models, foliar mass, leaf area, and canopy level isoprene emission were evaluated in the field and compared with model estimates. Model components performed reasonably well and generally yielded estimates within 20% of values measured at the site. However, measured emission factors were much higher in early summer following an unusually dry spring. These decreased later in the summer but remained higher than values currently used in emission models. There was also a pronounced decline in basal emission rates in lower portions of the canopy which could not be entirely explained by decreasing specific leaf weight. Foliar biomass estimates by genera using basal area ratios adjusted for crown form were in excellent agreement with values measured by litterfall. Overall, the stand level isoprene emissions determined by relaxed eddy accumulation techniques agreed reasonably well with those predicted by the model, although there is some evidence for underprediction at ambient temperatures approaching 30 ¿C, and overprediction during October as the canopy foliage senesced. A Big Leaf model considers the canopy as a single multispecies layer and expresses isoprene emission as a function of leaf area rather than mass. This simple model performs nearly as well as the other biomass-based models. We speculate that seasonal water balance may impact isoprene emission. Possible improvements to the canopy environment model and other components are discussed.¿ 1997 American Geophysical Union