Broadscale variability in tree data of the historical Public Land Survey and its consequences for ecological studies

Abstract

Historical records provide valuable information on the prior conditions of ecological systems and species distribution, especially in the context of growing environmental change. However, historical records may have associated bias and error because their original purpose may not have been for scientific use. The Public Land Survey (PLS) of the U.S. General Land Office (GLO) conducted from the late 1700s to the early 1900s has been widely used to characterize historical vegetation in the United States prior to major Euro-American settlements. Studies have shown that variability and bias exist in the data. However, these studies have not typically encompassed a region large enough to adequately assess this variability across diverse landscapes, nor attempted to distinguish potential ecological significance from statistical differences. Here we do this by analyzing variability in PLS data across all of northern Wisconsin, USA, a 75 000-km 2 landscape. We found ecologically significant differences among survey point types for tree species, size, and the distance to survey points. Both corner and line trees show some level of bias for species and size, but corner trees are likely the best sample. Although statistical tests show significant differences in species composition, tree size, and distance by tree sequence and location, the differences in species composition and tree size are not ecologically significant. The species differences are probably caused by fine-scale variability in the forest communities. The value of the PLS data remains high; choice of spatial extent, methods of analyses, and bias significance need to be evaluated according to variables of interest and project purpose.