Understanding Uncertainty in Broad-Scale Mapping of Historical Vegetation in the Great Lakes Region

In the Great Lakes Region, minor differences in soils and location (e.g., proximity to the Great Lakes) can lead to strong differences in vegetation; thus, the utility of broad-scale mapping often depends on capturing subtle landscape features and local processes. Similarly, vegetation patterns are in part a result of disturbances that have changed drastically over time, therefore mapping efforts must take into account vegetation-fire relationships to various biophysical settings (e.g., landtype associations, climate, and soils). Despite this, too little attention has been given to potential sources of mapping error, which include data limitations, ecological similarity, community classifications, locational error, sample quality, and lack of knowledge of systems-specifically natural disturbance regimes. We used similar to 23,500 plots with detailed vegetation, soils, and classification information to (1) evaluate LANDFIRE (Landscape Fire and Resource Management Planning Tools) historical vegetation (Biophysical Settings or BpS) classifications, (2) refine these classifications based on detailed soil regime and plant associations, and (3) draft fuzzy set soil-classification gradient maps to evaluate uncertainty in mapping and sources of mapping errors. Locally derived reference plot data often did not agree with LANDFIRE BpS mapping even for classifications generalized broadly by Fire Regime Groups. Our fuzzy methodological approach improves decision-making processes by assessing mapping confidence and highlighting potential sources for errors including classifications themselves. Our mapping efforts suggest that soil drainage and productivity data helped to delineate BpS classifications, which may in turn help stratify Existing Vegetation Types into feasible options.