Long-term measurements of above canopy isoprene emissions are reported for a mixed hardwood forest in northern lower MI, USA. Eddy covariance techniques were used to obtain fluxes of sensible heat (H), latent heat (LE), momentum, CO2, and isoprene. Results presented here include years 1999-2002. Measurements were made in collaboration with the AmeriFlux site located at the University of Michigan Biological Station (UMBS) and the Program for Research on Oxidants: PHotochemistry, Emissions and Transport (PROPHET). This work provides a unique long-term dataset useful for verifying canopy scale models and to help us better understand the dynamics of the biosphere atmosphere exchange of isoprene. In general, isoprene emissions increased throughout the day with increasing temperature and light levels, peaked at mid-afternoon, and declined to zero by night. Average midday isoprene fluxes were 2.8, 3.2, and 2.9 mg C/m2/h for 2000 through 2002 respectively. Last frost and full leaf out were significantly delayed in 2002 compared to the other years, however, total accumulated isoprene emissions for each year varied by less than 10%. Fully developed isoprene emissions occurred between 400 and 500 heating degree days, roughly half those required at other sites. Using long-term net ecosystem exchnage measurements from the UMBS Flux group, isoprene represents between 1.7 to 3.1% of the net carbon uptake at this site. Seasonally averaged LE fluxes were reduced in 2000 as a result of reduced rainfall, and average isoprene fluxes were 1.5 times greater in 2000 compared to years 2001 and 2002. Daytime fluxes of isoprene and both H and LE flux were linearly correlated on a daily basis, but the slopes of these relationships varied from one day to the next. The strong correlation between isoprene fluxes and associated energy fluxes is an important relationship that should be accurately reflected in canopy models used for estimating biogenic emissions. Observations were compared with the BEIS3 emission model and to a canopy scale biogenic emission model (WSU-BEIS). Estimates of isoprene agree well with observations during the mid-summer period, but BEIS3 overestimates observations during the spring and the fall. Estimates of sensible heat flux with WSU-BEIS were larger than observed, and the estimates of LE were within approximately 50% of observations.