Combining high-resolution aerial photography with gradient-directed transects to guide field sampling and forest mapping in mountainous terrain

Forest inventory and management relies on accurate maps of vegetation structure and composition. Creating such maps typically proceeds in two main stages: delineating stand boundaries on midscale aerial photography (approximately equal to 1:15,000) followed by collection of reference data in field plots. Technical and logistical limitations arise with respect to the resolution of photography, the allocation of field plots, and the sequence of the stages. Here we present a novel approach for forest mapping that (1) places the mapping process as the last rather than the first stage; (2) concentrates sampling in areas with strong environmental gradients; and (3) uses large-scale, high-resolution aerial photography (approximately equal to 1:2,000) to supplement field data and midscale airphotos. Our design builds on observations that plant communities in mountainous areas often exhibit strong correlations with environmental variables and that digital or hardcopy maps of these variables are becoming more widely available. In the first stage, 1:2,000 aerial photography is obtained via helicopter-mounted small-format cameras along flight lines that are located subjectively to follow significant environmental gradients. In the second stage, field plots are placed within airphoto transects to provide reference data of forest conditions. An integrated analysis of plot data and high-resolution photography provides the empirical basis for the development and calibration of an interpretation key. In the third stage, this key is applied to the delineation and classification of forested area on stereoscopic 1:15,000 aerial photography across the landscape of interest. We demonstrate this approach using a watershed in southwestern British Columbia and compare four probability-sampling designs with the gradient-directed approach proposed here using computer simulations. The results indicate that sampling along topographic gradients leads to a loss of accuracy with respect to estimation of distributional parameters. However, gradient-directed sampling is as likely as probability sampling to capture the full range of variability in forest conditions, with greatly improved logistics and cost-effectiveness. We conclude that our sampling design is a practical alternative for mapping projects in topographically complex landscapes.