Bias and saturation in estimated LAD and LAI using a rangefinder ground LiDAR

Leaf area index (LAI) and canopy openness were estimated at 2m intervals along 900m of transect in each of four forest types comprising different parts of a topographic sequence in a Central Amazon terra firme forest. Both metrics were obtained with a 1000 Hz rangefinder type LiDAR, using the distance to last return for each pulse. Pulses were directed vertically upward from the forest floor by an operator walking at constant speed. We avoided hours near noon, when direct sunlight could enter the detector. We estimated the leaf area density (LAD) of each voxel by the MacArthur-Horn method. We obtained LAI from the summed LADs of each voxel stack. Three causes of bias in LAD estimation are discussed. Histograms of LAI showed saturation, i.e. an upper limit of detectability, for all four forest types. Consequently, mean LAI did not differ between forest types, even if all 2m sections of transect were assumed independent (n=450). Corrections to these problems are being prepared, led by authors SCS and DRA. Canopy openness did differ between the four forest types. This suggests that LAI does differ but our unadjusted methods did not permit detection of this difference.