General Land Office (GLO) survey records have been used to determine historic vegetation through the application of the distance methods to the witness tree data. We test the suitability of applying modifications of the random pairs method or the quarter method to GLO records from the Shawnee National Forest Purchase Area in southern Illinois. The witness tree data was evaluated with criteria based on the distribution of the witness trees around the section or quarter section points. As a whole, the data set did not meet all criteria for either method. Data were subdivided into groups meeting specific criteria. For interior points with two witness trees, the random pair method with an exclusion angle of 202.93 degrees provided the best modification of the distance methods to convert witness tree distances to tree density. A new correction factor of 0.74 to convert the mean witness tree distance to the square root of the mean area was determined by regressing exclusion angle against correction factor. Wilcoxon sign-rank tests indicate that there are significant differences in the calculated square root of the mean area based on distance method used (quarter method v. random pairs S=3665417, p<0.001; random pairs v. new 0.74 correction factor S=3893716, p<0.001) and depending if tree diameter is included in witness tree distance measurements (with diameter v. without diameter S= -1349.5, p<0.001).

Key words: Distance methods modifications, GLO survey records, Historic vegetation