Correct estimation of spatially distributed CO2 flux is of utmost importance for regional and global carbon balance studies. Tower-based instruments provide flux data from a small footprint area and may not be suitable for spatial extrapolation over areas not represented by the towers. In this study we developed a method of combining optical indices from remotely sensed hyperspectral images with flux data from towers covering different vegetation types to make spatially continuous maps of gross CO2 fluxes. Using a simple light-use efficiency model, we tested the ability of spectral indices derived from Airborne Visible Infrared Imaging Spectrometer (AVIRIS) imagery to estimate photosynthetic fluxes of several boreal forest stands. Because CO2 flux from terrestrial ecosystems is dependent on both vegetation cover and physiological state, we hypothesized that measures of both forest structure and physiology were important for flux estimation. Consequently, the modeled fluxes considered both the normalized difference vegetation index (NDVI) and a scaled value of the photochemical reflectance index (PRI), both derived from narrowband reflectance. NDVI alone was of limited use in describing the variation in ecosystem fluxes (R2=0.26). Addition of the PRI, which is related to xanthophyll cycle pigment activity and unrelated to NDVI, improved the agreement between modeled and measured fluxes (R2=0.82). Our results also indicated that simple extrapolation of point-based flux tower data to represent the large-area fluxes of boreal forest may lead to an underestimation of the spatially distributed fluxes, at least for the vegetation types studied in this analysis. ¿ 2001 American Geophysical Union