Authors: Carrie R. Levine, Charles V. Cogbill, Brandon M. Collins, Andrew J. Larson, James A. Lutz, Malcolm P. North, Christina M. Restaino, Hugh D. Safford, Scott L. Stephens, John J. Battles
In the Western United States, historical forest conditions are used to inform land management and ecosystem restoration goals (North et al. 2009, Stephens et al. 2016). This interest is based on the premise that historical forests were resilient to ecological disturbances (Keane et al. 2018). Researchers throughout the United States have used the General Land Office (GLO) surveys of the late 19th and early 20th centuries to estimate historical forest conditions (Bourdo 1956, Schulte and Mladenoff 2001, Cogbill et al. 2002, Paciorek et al. 2016). These surveys were conducted throughout the United States and represent a systematic, historical sample of trees across a broad geographic area. A challenge of using GLO survey data is the accurate estimation of tree density from sparse witness tree data. Levine et al. (2017) tested the accuracy and precision of four plotless density estimators that can be applied to GLO survey sample data, including the Cottam (Cottam and Curtis 1956), Pollard (Pollard 1971), Morisita (Morisita 1957), and mean harmonic Voronoi density (MHVD; Williams and Baker 2011) estimators. The Cottam, Pollard, and Morisita are count‐based plotless density estimators (PDE) and have a history of being applied to GLO data (e.g., Kronenfeld and Wang 2007, Rhemtulla et al. 2009, Hanberry et al. 2012, Maxwell et al. 2014, Goring et al. 2016). The MHVD estimator is an area‐based PDE that has been applied by the study's authors to sites in the western United States (Baker 2012, 2014), but had not been independently evaluated. Levine et al. (2017) found that the Morisita estimator was the least biased and most precise estimator for estimating density from GLO survey data, with a relative root mean square error ranging from 0.11 to 0.78 for the six study sites. Levine et al. (2017) also demonstrated the MHVD approach consistently overestimated density from 16% to 258% in all six study areas that were analyzed.