Quantifying the accuracy of digital hemispherical photography for LAI estimates on broad-leaved tree species

Digital hemispherical photography (DHP) has been widely used to estimate leaf area index (LAI) in forestry. Despite the advancement in the processing of hemispherical imageswith dedicated tools, several steps are still manual and thus easily affected by user’s experience and sensibility. The purpose of this study was to quantify the impact of user’s subjectivity on DHP LAI estimates for broad-leaved woody canopies using the software Can-Eye. Following the ISO 5725 protocol, we quantified the repeatability and reproducibility of themethod, thus defining its precision for a wide range of broad-leaved canopies markedly differing for their structure. To get a complete evaluation of the method accuracy, we also quantified its trueness using artificial canopy images with known canopy cover. Moreover, the effect of the segmentationmethod was analysed. The best results for precision (restrained limits of repeatability and reproducibility) were obtained for high LAI values (>5) with limits corresponding to a variation of 22% in the estimated LAI values. Poorer results were obtained formediumand low LAI values, with a variation of the estimated LAI values that exceeded the 40%. Regardless of the LAI range explored, satisfactory results were achieved for trees in row-structured plantations (limits almost equal to the 30% of the estimated LAI). Satisfactory resultswere achieved for trueness, regardless of the canopy structure. The paired t-test revealed that the effect of the segmentationmethod on LAI estimates was significant. Despite a non-negligible user effect, the accuracymetrics for DHP are consistent with those determined for other indirectmethods for LAI estimates, confirming the overall reliability of DHP in broad-leaved woody canopies.