| Continuous forest cover maps for an area located in the Luce District of the Upper Peninsula of Michigan were generated from two different forest point datasets using different spatial and geostatistical interpolation methods. The two forest datasets interpolated were current Forest Inventory and Analysis (FIA) data collected in 1993 and pre-European settlement General Land Office survey (GLO) data. Given the relationship that exists between soil and site factors, like drainage and slope, and predominant forest types, ancillary soils information were utilized to provide neighborhoods of influence for each of the interpolation methods. Three interpolation methods were applied to the FIA data: Thiessen Polygons (TP), Inverse Distance Weighting (IDW), and Indicator Kriging (IK). The results of the three interpolation methods were compared using the FIA data to assess the differences in outputs and final maps. Final interpolated forest maps were then compared to a landcover map derived from Landsat TM imagery collected in 1991 and calculations of forest type areas generated by a FIAMODEL (Reed et al., 2001) model to judge the appropriateness and success of each procedure. Based on the success of two of these methods, IDW and IK were applied to the GLO data. The interpolated output and final forest maps from the GLO data were also compared to assess any differences in methods and their outputs. Final forest maps were compared to manually generated MNFI pre-European settlement maps to assess the differences. For both datasets, IDW performed the best in delineating all forest groups occurring in the study area while IK performed well for mapping only those forest groups occurring most frequently. IDW maintained the local spatial information and IK mapped more regional patterns of forest occurrence. |
