Double sampling to assess the accuracy of ground-based surveys of tree hollows in eucalypt woodlands

Ground‐based surveys of tree hollows often give poor estimates of hollow abundance in forests. Woodlands have shorter trees and a more open structure than forests, which may make hollows easier to detect. Therefore, one would expect ground‐based surveys of tree hollows to be more accurate in woodlands than in forests. We compared hollow counts from ground‐based and climbing surveys (double sampling) for four species of Eucalyptus trees in woodlands of central‐western New South Wales, Australia: E. camaldulensis Dehnh, E. melliodora A. Cunn. ex Schauer, E. microcarpa Maiden and E. populnea F. Muell. ssp. bimbil L.A.S. Johnson & K.D. Hill and E. melliodora A. Cunn. ex Schauer. Overall, 83% of hollow‐bearing trees and 93% of trees without hollows were correctly classified by ground‐based surveys. Mean difference in hollow counts of ground‐based surveys to climbed surveys was 1.7 hollows ± 0.2 SE (all species combined) with 91% of ground‐based hollow counts being within five of the actual number of hollows. The error in ground‐based counts of hollows in E. microcarpa was larger than for the other three species. Errors in all species resulted from both overestimation and underestimation of hollow abundance by ground‐based surveys. A larger error was associated with the detection of hollows located in branches compared with hollows located in the main trunk(s). Total number of hollows in the tree (based on climbing surveys), crown area or maximum trunk diameter were significant predictors of ground‐based survey accuracy. Overall, the accuracy associated with ground surveys was relatively high and generally error rates were lower than those published for forests.