Calibrations of indirect methods for estimating leaf area are usually based on small data sets and are often species-specific and may even be stand-specific. We used the Licor LAI-2000 plant canopy analyser (PCA) as a reference to calibrate leaf area measurements based on hemispherical photography. Ten stands of 6-8 year-old, plantation grown Tasmanian bluegum (Eucalyptus globulus Labill.) in Western Australia were used to investigate the effects of variations in sampling position, photographic exposure and image processing on leaf area (or leaf area index, L) estimated using hemispherical photography. We also compared our photographic estimates of L with those obtained via destructive sampling (allometry) in both evenly spaced and highly clumped stands of E. globulus. Varying exposure by one stop affected estimated L by approximately 13% but we confirmed that correct exposure can be approximately predicted by metering exposure outside the canopy. In situations where metering exposure outside the canopy is impractical, we recommend the use of empirically derived relationships between L estimated from photographic images made at a constant exposure and actual L derived from other means. Mean tilt angle obtained for E. globulus from the photographic method (68.7 degrees +/- 2.5 s.e.) agreed well with estimates for Eucalyptus species from other studies. Owing to the non-random arrangement of crowns within plantations, sampling position significantly affected mean tilt angle. In highly clumped stands with closely spaced double rows and wide inter-double row gaps, underestimation of L by 16-30% with the photographic method was probably the result of greater foliage clumping at the crown level. We concluded that, in stands of E. globulus and probably other broadleaf species with evenly distributed crowns, foliage clumping at the shoot or branch level is unlikely to be a significant source of error in indirect estimates of L. Scattering of blue light may result in large underestimates of L when using the PCA. In stands with 'extreme' architecture, indirect, light interception-based methods are likely to greatly underestimate L, although, positioning the sensor so as remove large gaps from view may allow accurate estimates of L even in these stands. (C) 2000 Published by Elsevier Science B.V All rights reserved.