Historical GIS involves applying GIS to historical research. Using a unique method, I recovered historical tree survey information stored in bar chart figures of a 1956 publication. I converted PDF files to TIF files, which is a format for a GIS layer. I then employed GIS tools to measure lengths of each bar in the TIF file and used a regression (R2 = 97%) to convert bar lengths to numerical values of tree composition. I joined this information to a spatial GIS layer of Indiana, USA. To validate results, I compared predictions against an independent dataset and written summaries. I determined that historically (circa 1799 to 1846) in Indiana, oaks were 27% of all trees, beech was 25%, hickories and sugar maple were 7% each, and ash was 4.5%. Beech forests dominated (i.e.,>24% of all trees) 44% of 8.9 million ha (i.e., where data were available in Indiana), oak forests dominated 29%, beech and oak forests dominated 4.5%, and oak savannas were in 6% of Indiana, resulting in beech and/or oak dominance in 84% of the state. This method may be valuable to reclaim information available in published figures, when associated raw data are not available.