Predictive Ecosystem Mapping (PEM) for 8.2 Million ha of Forestland, British Columbia, Canada

Operational predictive ecosystem mapping (PEM) at a scale of 1: 20,000 is described for an area of 8.2 million ha in the former Cariboo Forest Region of British Columbia (B.C.), Canada. Mapping was conducted over 5 years by a small team consisting of a knowledge engineer, a local ecological expert, a project technical monitor, a project manager and a number of short-term contractors. The total cost for all project activities was $2.8 million Canadian dollars or 34 cents per ha. The rate of progress was 2 million ha per year for the 2 person modeling team. The predictive map was assessed for accuracy in terms of its ability to provide reliable estimates of the proportions of ecological site types within small areas. Accuracy assessments were made using 345 km of independently classified ecological observations collected along 230 randomly selected, closed linear field traverses of 1.5 km total length. The final PEM maps achieved an average accuracy of 69% across the entire map area. We summarize and generalize our experiences by recasting them in the form of ten principles that we feel are applicable to all efforts to make predictive mapping operational. We hope that these principles will stimulate discussion among practitioners of digital soil mapping and may help others to consider how best to achieve their own success in operational digital soil mapping.