How you choose where to place your plots, points, or transects (i.e., your samples) is important. In most situations you cannot count every animal living within your study area. Rather, sets of points or plots are chosen to represent a cross-section of the study area. How and where you sample will determine how well your monitoring program represents the study area. Below are some of the major factors that must be considered when setting up a sampling scheme.

• Too few samples over too large an area. If your study area is very large and the number of places within that area that you sample are few, it is reasonable to surmise that trends calculated from your sampled locations may not reflect trends for the overall area. Unfortunately, it is difficult to clearly define when you have too few samples from a geographic perspective. However, in most cases pilot data or data from similar studies usually indicate that sample sizes for monitoring programs need to be large for most species to accommodate the variability inherent in population counts.
• Pre-determining where to sample. Under most circumstances an individual who picks their own sampling locations without some sort of random or systematic sampling scheme involved ends up putting samples in sites that are not representative of the study area. Rather they are influenced by that person's thoughts of what places are attractive and what places should be avoided. This clearly limits the usefulness of your monitoring program, as any analyses of such data can only represent the chosen plots. The solution is easy: Select your sampling points with a random or systematic scheme. Special habitats, environments, or circumstances can be incorporated into such schemes as illustrated below.
• Excluding certain areas from sampling. There are many situations where you may decide that for matters of safety or logistics you must restrict samples to trails, roads, or along bodies of water. Other times you may decide to randomly choose your sites, but need to exclude some areas that fall in difficult to access spots, such as the middle of a swamp or on steep slopes. These are legitimate decisions, but be aware that when you ultimately write up your reports or papers you must state that these trends represent only those areas that had some possibility of being sampled in the first place and do not completely represent the study area.
• Stratifying on Vegetation or Habitat. There may be times in which you will want to stratify your sampling locations. That is, to increase or decrease the density of sampling points in certain areas in your landscape. This is often done when parts of a refuge or park are particularly interesting, but would have too few samples to calculate reliable trends using the overall sampling frame. Be aware, however, that once you create a stratified sampling scheme it is very complicated and sometimes impossible to make changes later to the boundaries you have chosen without screwing everything up. Consequently, the boundaries used to create stratification schemes should be related to permanent physical features, such as stream corridors or soil types, that are unlikely to change in subsequent years and boundaries based on vegetation, habitat, or current land use should be avoided. Vegetation can change over time, but your sampling scheme does not have the same flexibility.
• Choose New Points Each Year or Use the Same? Use the same if you want to detect changes over time. Using the same points each year eliminates the variation in counts associated with location. Secondarily it makes logistics easier. Be aware, however, that in larger datasets you will likely need to statistically account for the fact that you use the same points by choosing tests that incorporate “repeated measures.”
• Map out all Your Points Before Going Out Into the Field. By mapping your points on maps or a GIS system you will be better able to visualize the consequences of your choice of points and your sample sizes. Mapping your points will quickly illustrate any errors that were made in choosing them. If you wait until after the first year’s data are in, or, worse, wait many years you may have to start the entire process over again, and possibly making the early data UNUSABLE.
• Not all Points Need Be Visited Every Year. In many circumstances it is completely legitimate to not run your surveys every year. It will depend on your goals as to whether you need datapoints every year. There are also a number of sampling frameworks that permit sampling to occur in multiple stages or rotate the number and location of samplings points across a number of years. These often are more efficient designs in that you will get more precise estimates of change for less field time, but, the penalty is they also require custom analysis programs and great understanding of analysis techniques to properly analyze.

It Can Be Done. Well, after that list you may have some trepidations about where you will be putting you sampling points. Good. An inappropriate sampling frame is perhaps the most common mistake made in creating a monitoring program.Rather than recreate a number of good sources of information here about how to choose points we will direct you to a few web sites. The best place to start would be the National Park Service’s Inventory and Monitoring Program’s information on sampling frames.

Authors' note: We believe that the National Park Service site has an excellent overview of setting up sampling frames (http://science.nature.nps.gov/im/monitor/). We suspect, that other such web sites may exist as well. If you know of such a site or have a bibliography of key resources please let us know by emailing sam_droege@usgs.gov. Thank you.

Next question: HOW to monitor? (choosing a monitoring approach, counting technique and setting sample sizes)