Calibrating photo-interpreted forest cover types and relative species compositions to their ground expectations

Forest inventories gather data from many sources at different scales and resolutions. A common reference level and scale is needed to assess quality and accuracy of an inventory. This study demonstrates, with data from New Brunswick, a procedure for calibration of photo interpretations of cover types and relative species composition (by volume) to the expected ground values. The species volume composition was calibrated with a synthetic contingency table because of the lack of a one-to-one object linkage in the two sets of estimates. Generally, the agreement between photo estimates and ground estimates was weak to poor (kappa average of 0.3), when adjusted for chance agreement. Different scales, resolutions, and points of observations all contribute to these results. Calibrations resulted in significant shifts in cover types and species compositions and enlarged the associated standard errors by 10-300%.