Inventory and Survey Methods for Nonindigenous Plant Species (PDF)

Chapter 1 • Getting StartedThere are three considerations for locating sampling unitsin a landscape: (1) randomly locating the sampling units toensure the data are not biased (i.e., not influenced by theposition of the sampling unit), (2) positioning sampling unitsto achieve good interspersion across the entire area of interest,and (3) defining sampling units that are independent ofeach other (Elzinga et al. 2001).The sampling units (points, transects, swaths) describedhere are located in the landscape according to either subjective(biased), random (unbiased), systematic, or stratifiedsampling designs. A random design can also be incorporatedinto the systematic or stratified sampling designs if a subsetof the sample units/area is desired. The design selectedis based on the manager’s objectives, constraints, andknowledge of the area to be sampled (Herrick et al. 2005).Subjective sampling is biased because it does not samplefrom all areas in the landscape; for example, only specificareas such as roadsides may be selected for sampling.Random location of sampling units, on the other hand, isan unbiased process of selection in which any area has anequal probability of being chosen. Systematic samplinguses a defined order or plan to locate sample units in thelandscape, but the initial point may be randomly selected. Ina stratified sampling design, the sample area is divided intoone or more subgroups (strata) before sampling units arelocated. For example, the sample area may be stratified onelevation ranges, roads, or habitat types.Crew members may cover several miles in a long day of fieldwork. Photo courtesy of Shana Wood.We recommend using a random sampling design(described below) over a subjective design (i.e., a personaldecision of where to locate the sample; Figure 2a). Whilesubjective sampling can be inexpensive and sensitive tolocal land use, it is biased, difficult to extrapolate, anddependent on individual knowledge that may or may notbe correct or complete (Herrick et al. 2005). For example, ifmanagers sampled only where they knew NIS existed, theywould overlook where, how, and in which environmentsNIS are spreading, all of which are important for effectivemanagement.Simple random sampling involves randomly selectingareas of the landscape to sample (Figure 2b) by using maps,aerial photographs, GIS software, or other means. Forexample, a sample point, transect, or swath location, or atransect/swath starting location, could be randomly locatedat the intersection of a latitude/longitude location. Becausethe sample units are randomly located in a landscape, ifa sufficient number of samples are taken, there is a goodchance of sampling all environmental variables or habitattypes in that landscape at the proportion in which theyoccur. In general, simple random sampling is easy to applyand statistically valid (Herrick et al. 2005). If the landscapeof interest is large and point sample units are used, simplerandom sampling may involve high travel and labor coststo implement (Rew et al. 2006). Also, with the randompoint design, or any point design, there is a relatively highchance that field personnel will encounter NIS patches onthe way to the next survey point, but these NIS should notbe recorded under this sampling strategy. To decrease traveltime and improve sampling efficiency, survey personnelcould use randomly located transects that allow the collectionof continuous data (Rew et al. 2006).Systematic sampling is an easy way to establish samplingunits in a landscape, and it has good interspersion betweensampling units (Elzinga et al. 2001). Systematic randomsampling locations follow a system or grid in the landscapegenerated from a random starting point, from which a numberof data location points are randomly selected (Figure 2c). Thisis an unbiased method. Randomly subsampling from the fullgrid should provide equally accurate results as sampling thefull grid, but will be quicker. An important benefit of systematicrandom sampling is that it is easily repeatable because itis based on a sampling unit (or point) at a specific location.However, depending on the distance between sample points,the field crew could possibly walk past NIS patches or misssampling a habitat or environmental variable. Systematicrandom sampling generally involves a great deal of traveltime and labor, which can be somewhat reduced if transectsare used rather than points. However, even systematictransect sampling could miss some environmental habitat13Inventory and Survey Methods for Nonindigenous Plant Species