Inventory and Survey Methods for Nonindigenous Plant Species (PDF)

Chapter 6 • Stratified Random Sampling Methoda data dictionary that would contain all the data fieldswhich need to be collected in the field. The data dictionarystreamlines data processing in the office because the dictionaryis designed so that it does not allow fields to be skippedor incorrectly formatted data to be entered. In addition,a handheld compass can be used to navigate transectbearings, which is particularly useful in steep and variablecountry where the GPS reception is likely to be intermittent.It is possible to train field crews relatively quickly even ifthey have little prior knowledge of plant identification andGPS. Our field crews have generally been undergraduatestudents with little to some prior plant identification skillsand/or knowledge of GPS and GIS. Training the crew toaccurately identify the target NIS and dominant nativevegetation takes longer than teaching them to masterthe GPS, including the data dictionary—mastering thisgenerally takes only a few hours.Analyzing the field data to create diagrams such asthose in Figure 2, 3, and 4 can be performed in Excel withbasic Excel knowledge. The predictive models require astatistical and GIS package; we use S-Plus (Mathsoft Inc.)and ArcView. Statistical and GIS training is required, but theauthors can provide some guidance if requested.Advantages and Disadvantages of theStratified Random Sampling MethodWe developed a simulation model in a GIS to evaluateseven different survey methods for consistency and reliabilityof (1) intersecting patches of NIS in a large landscape(24,710 acres or 10,000 ha), (2) producing samples whichreflected the spatial distribution of the populations in thelandscape, and (3) considering cost and time efficiency. Thisevaluation helped to quantify the advantages and disadvantagesof the stratified random sampling method comparedto other methods in a large wildland area. For detailedmethods and results see Rew et al. (2006).In short, the stratified random sampling method providesone of the most reliable samples of the true spatial fielddistribution. Grid and random point sampling providedresults that were similar to each other and to the resultsfrom the stratified random sampling method; however,stratified random sampling was more time efficient than thegrid and random point sampling because less ground had tobe covered to sample the same area. Therefore, if the objectiveis to find out which NIS are present and where theyoccur in the management area, stratified random sampling isthe preferred method. It can be used on most landscapes, orareas of any size where a survey rather than an inventory isrequired. This method provides information on the environmentsin which species are more likely or less likely to befound. Even without the probability maps, this informationcan help prioritize further surveys and monitoring efforts.With regard to producing maps, more data are likely toimprove the accuracy of the maps, but at the very least themaps provide visualization of the entire management areaand also help with prioritization and targeting of particularareas and species for monitoring and management.The method could also be used for rare plants or taxaother than plants, except that different stratification factorswould be necessary.Although the stratified random sampling method canseem complicated, it is basically building on currentknowledge, but collecting data in such a way that we canlearn more about the current and likely future distributionof target NIS.Collecting data in areas where NIS are less likely to befound is often seen as a disadvantage and a waste of resources,but absence as well as presence data are importantif we, land managers, want to understand more about NISdistribution over the whole management area.To produce the probability models, technical knowledgeof statistics and GIS are required, but the more basicanalyses can be performed by anyone with knowledge ofExcel or other spreadsheet programs. We are developing astep-by-step guide to initial data analysis and the modelingprocedures. The data dictionary we used in the field, theExcel macros, and ArcView GIS extensions which our officehas written are available to anyone. Contact either author ofthis chapter for more information (see List of Contributors).ReferencesHirzel, A. and A. Guisan. 2002. Which is the optimal sampling strategy forhabitat suitability modelling? Ecol. Modelling 157:331-341.Maggini, R., A. Guisan, and D. Cherix. 2002. A stratified approach formodelling the distribution of a threatened ant species in the SwissNational Park. Biodivers. Conserv. 11:2117-2141.[NAWMA] North American Weed Management Association. 2003. NorthAmerican Invasive Plant Mapping Standards. http://www.nawma.org/.Rew L. J., B. D. Maxwell, and R. Aspinall. 2005. Predicting the occurrenceof nonindigenous species using environmental and remotely sensed data.Weed Sci. 53:236-241.Rew L. J., B. D. Maxwell, R. J. Aspinall, and F. L. Dougher. 2006. Searchingfor a needle in a haystack: evaluating survey methods for sessile species.Biol. Invasions 8(3):523-539.55Inventory and Survey Methods for Nonindigenous Plant Species