Forest fire smoke and a test of hemispherical photography for predicting understorey light in Bornean tropical rain forest

Hemispherical canopy photographs are now widely used to estimate the propor tion of above canopy PPFD (photosynthetic photon flux density) that is tran smitted to the forest understorey. Some studies have obtained good results, while others have found that predictions are not always independent of for est type or condition, and that errors are sometimes large beneath very den se canopies. Lack of repeatability can make it difficult to compare sites, studies, or photographs taken at different times. In this study the method for calibrating predictions from photographs was tested in logged and,undis turbed tropical broadleaf forest in Borneo, during both the wet and dry sea sons. During the dry season a regional smoke 'haze' caused by widespread fo rest fires provided very uniform diffuse conditions and thus a unique test of the calibration. Total transmission (the 'global transmission factor', T -t) was modelled as the weighted average of direct and diffuse transmission , with the weighting on diffuse transmission (f(d)') determined as the valu e which maximised the correlation between T-t and measured PPFD. f(d)' is a n indirect measure of f(d), the actual proportion of above canopy PPFD that is diffuse. f(d)' should increase with cloudiness or smoke haze. Accurate determination of f(d)' is important for prediction of transmission over sho rt time-scales, such as days or weeks. fdr was modelled for each day and wa s correlated (r(2) = 0.65) with daily shadow band measurements of the above canopy diffuse component during the wet season. During the dry season, fd' initially decreased with the expected increase in sunshine hours, then the smoke haze halved above canopy PPFD and f(d)' increased from less than 0.5 to 1.0, thus validating this method of calibration. After calibration, pre dictions of T-t from photographs in logged forest were repeatable, accurate (r(2) > 0.96) and linear for sites with PPFDs ranging from less than 5% of above canopy PPFD to sites with more than 50% of above canopy PPFD. fd' mo delled from monthly or longer averages of transmission factors and PPFD was within 0.12 of f(d). Results were more variable (r(2) = 0.68) if calibrati on was based on data from undisturbed forest only. Overall accuracy and rep eatability between sites were attributed to increased digitised image resol ution, the use of a spot meter with a narrow field of view to set photograp h exposure times, and other improvements in methodology. (C) 1999 Elsevier Science B.V. All rights reserved.