Uncertainty Estimation for Map-Based Analyses

Traditionally, natural resource managers have asked the question, “How much?” and have received sample-based estimates of resource totals or means. Increasingly, however, the same managers are now asking the additional question, “Where?” and are expecting spatially explicit answers in the form of maps. Recent development of natural resource databases, access to satellite imagery, development of image classification techniques, and availability of geographic information systems has facilitated construction and analysis of the required maps. Unfortunately, methods for estimating the uncertainty associated with map-based analyses are generally not known, particularly when the analyses require combining maps (Foody 2006, van Oort and others 2006). A variety of uncertainty methods is illustrated for map-based analyses motivated by the threat of the emerald ash borer to the ash tree resource in southeastern Michigan. The analyses focus on estimating the uncertainty in forest/nonforest maps constructed using forest inventory data and satellite imagery, ash tree distribution maps constructed using forest inventory data, and estimates of the total number of ash trees for a selected region obtained by intersecting the two maps. A crucial conclusion of the study is that spatial correlation, an often ignored component of uncertainty analyses, made the greatest contribution to uncertainty in the estimate of the total number of ash trees.