The leaf area density (LAD) of a forest is an important indicator of forest biomass and is therefore pertinent to monitoring carbon sequestration and change. Quantitative physical models were used to estimate forest LAD from radar and hyperspectral airborne remote sensing observations. A parameter-estimation technique based on physical models minimizes the need for in situ observations and thereby facilitates global remote sensing of forest structure. Using data from the NASA Airborne Synthetic Aperture Radar (AIRSAR) and the NASA Airborne Visible and Infrared Imaging Spectrometer (AVIRIS) over three forest plots in Central Oregon, parameters were estimated separately from the radar and hyperspectral data and then combined to form LAD. Gaussian relative LAD profiles were estimated from multialtitude interferometric and polarimetric AIRSAR data. Leaf area indices (LAI) were estimated from AVIRIS data and used to normalize the relative density profiles to produce LAD as a function of height. LAD was also determined from field measurements of geometric tree properties and LAI. LADs in the three forest plots were in the 0.02--0.18 m2 m-3 range, with LAIs in the range 0.8--2.4 m2 m-2. The agreement between the remotely sensed and field-measured LAD was typically 0.02 m2 m-3 but occasionally as high as 0.06 m2 m-3, which was within a 1--2 standard error range. More complex models for the remotely sensed relative density, along with more diverse radar observation strategies, will improve LAD accuracy in the future. LAD estimation will also improve when radar, hyperspectral, and other relevant remote sensing data sets are combined in a single parameter-estimation process, as opposed to the separate estimations performed in this first LAD demonstration.