Thematic Accuracy Assessment Procedures. Version 2 - USGS
Thematic Accuracy Assessment Procedures. Version 2 - USGS
Thematic Accuracy Assessment Procedures. Version 2 - USGS
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
2.0 Sampling Design<br />
2.1 Sampling Design and Data Collection Objectives<br />
The sampling design (Stehman and Czaplewski 1998) for accuracy assessment within the NPS<br />
Vegetation Inventory adheres to the scientific principles that govern sampling and statistical<br />
analysis, while also striving to be practical and cost effective. Specifically, the methodology<br />
should satisfy the following objectives (Stoms et al. 1994):<br />
1. The methodology should be scientifically sound. In order to accomplish this need, the method<br />
should be repeatable, and the sampling design should permit the adequate representation of the<br />
population about which statistical inferences are to be drawn.<br />
2. The methodology should be economically feasible in view of both time and cost constraints.<br />
3. The methodology should be applicable to all areas that are part of the project. Although there<br />
may be some regional variation in the implementation of the accuracy assessment, these<br />
variations should be based on the same theoretical foundation, so that the results of individual<br />
project assessments are comparable.<br />
The objective of collecting observations for the thematic accuracy assessment is to draw<br />
inferences about the magnitude of discrepancies between the true attributes of a point or area and<br />
its representation on the map. Thus, the accuracy assessment objective contrasts with the<br />
objective of classification plots (releves), which is to provide raw data for the thorough<br />
description of vegetation types as efficiently as possible and is often accomplished by subjective<br />
sampling (Mueller-Dombois and Ellenberg 1974).<br />
With these types of objectives, the randomness of the sample sites should be emphasized, and the<br />
number of samples sites will be heavily influenced by statistical constraints.<br />
Appendix C describes an example of a sampling design in steps for Shenandoah National Park<br />
(Young et al. 2009).<br />
2.2 Observation Area (Minimum Mapping Unit) Size<br />
While the issue of observation area size is of concern for field methods (Chapter 3), it also places<br />
some constraints on the sampling design and is treated here. As stipulated in the previous<br />
chapter, the thematic accuracy assessment observation area will be equivalent in size to the size<br />
of the minimum mapping unit designated for that map class.<br />
The default (standard) minimum mapping unit (MMU) size for the NPS Vegetation Inventory is<br />
0.5 hectare. While this size suffices for many vegetation types, mappers are often able to map<br />
vegetation that occurs in smaller stands because these stands contrast well with surrounding<br />
stands in imagery. An example of this is a small herbaceous wetland that is surrounded by an<br />
upland evergreen woodland. It is often advantageous to delineate these smaller stands when they<br />
are easily recognized and reasonably few in number because: (1) they represent vegetation types<br />
important to management and (2) designating them as separate map classes can increase the<br />
accuracy of surrounding or adjoining classes. This is because these small stands then will not<br />
occur as unmapped inclusions that the field observer may fail to recognize as different and that<br />
15