Improving accuracy of canopy hemispherical photography by a constant threshold value derived from an unobscured overcast sky

High image discrimination threshold values tend to be given to canopy hemispherical photographs (CHPs) with high exposure (resulting in bright images), but the effects of exposure on image threshold have been overlooked. A model canopy system was developed to precisely manipulate exposure (in relation to reference exposure measured from an unobscured overcast model sky), canopy openness, gap fragmentation and sky illumination of CHPs. We showed that there was a numerical trade-off relationship between exposure and image threshold of CHPs, whereas the last three factors had negligible effects on image threshold. A new thresholding method based on this relationship was shown to be applicable in the field and to allow users of hemispherical photography (HP) to compensate for the effects of higher exposure on CHPs with higher empirical threshold values. The advantages of this method are as follows: (i) it is an objective method; (ii) its overall performance is better than that of automatic thresholding methods throughout a wide range of canopy openness and exposure; and (iii) its efficiency of image analysis is higher than that of conventional manual and automatic thresholding methods, because all CHPs can be analysed with a constant empirical threshold value. Our results highlight the necessity of integrating both exposure and thresholding for optimal CHP results and the contribution that model canopies can make to methodological studies of HP.