|Roberts, Elizabeth - MONTANA STATE UNIV|
|Lawrence, Rick - MONTANA STATE UNIV|
Submitted to: Western North American Naturalist
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: October 2, 2003
Publication Date: August 1, 2004
Citation: Roberts, E.A., Sheley, R.L., Lawrence, R.L. 2004. Using sampling and inverse distance weighted modeling for mapping invasive plants. Western North American Naturalist. 64(3):312-323. Interpretive Summary: Currently used hand- and GPS-mapping is too costly and time consuming. It may be possible to create relatively accurate maps by sampling and using spatial statistics, rather than fully delineating all the weeds. We tested the accuracy of using various sampling methods and sampling numbers in combination with spatial modeling. Our study showed that systematic sampling only 1/1000 of the land area provided maps that are at least 80% accurate for mapping leafy spurge and spotted knapweed. Sample mapping may be much more efficient method than hand- or GPS-mapping for mapping invasive plant. This would allow more resources to be used for management.
Technical Abstract: Accurate time- and cost-efficient mapping is central to successful rangeland invasive plant management. In this study, sampling and Inverse Distance Weighted (IDW) interpolation modeling was tested as a mapping alternative to expensive full-coverage delineation survey mapping methods. Our objective was to examine accuracies of presence/absence maps generated from 18 sampling strategies (3 sampling methods x 6 sample densities) using IDW. Invasive plant field survey maps with known accuracies were used for generating samples and to test interpolation results at two sites. Site 1 was approximately 6.0 km2 in size, dominated by Russian knapweed (Acroptilon repens L.). Site 2 was an approximately 13.5 km2 upland area dominated by spotted knapweed (Centaurea maculosa Lam). Sampling method x sample density combinations were gathered from the field survey infestation maps using repeated computer-based sampling methods. IDW modeling was applied to each of the sample data sets. Accuracies of the IDW interpolation results were calculated by re-referencing field survey maps. Sampling at density of 0.25 percent (about 1 point per ha) using a systematic sampling method was determined to be the preferred sampling strategy for both sites. This combination of sampling density and method yielded 85% accurate presence/absence maps. We conclude sampling, when combined with ID interpolation modeling, can generate accurate invasive plant maps and is a potential alternative to current delineation survey methods.