Key areas are sites that are representative of what the condition or trend of a larger area such as a pasture, allotment, herd management area, or watershed (BLM 1996). As such, key areas reflect the impact of management actions across the entire area.
Key areas are selected based on management objectives with an understanding of how different management and natural processes interact to impact the environment (i.e, create patterns on the ground) (BLM 1996). Key areas should be placed within a single ecological site or plant community (i.e., not in transition zones or ecotones), and in areas that are likely to show a response to management actions. The number of key areas needed depends on the size and variability of the area and the management objective.
For instance, in arid rangeland in the western U.S., livestock are often limited in their distribution to within a relatively short distance to water (generally less than 2 miles), with the degree of use of land by livestock decreasing as you move away from the water source. Beyond approximately 2 miles, there is little grazing impact because the animals rarely get out there, so there is little need to monitor this area. Areas close to the water source get lots of use, are also not representative of the impact of grazing, and generally are not monitored as key areas (although the condition and trend of these areas may be monitored separately as critical areas). In the simple example illustrated in Figure 1, key areas might be located in a doughnut-shaped buffer around the water source (see figure below).
Figure 1: An example of a key-area approach to selecting monitoring locations. The square represents the entire monitoring area of interest (e.g., an allotment) and the different color subdivisions stand for different ecological sites. The light blue point is a water feature, and the dashed line is a road. The appropriate area for monitoring grazing impacts (hatched ring) is typically beyond the high-impact zone around the water feature but within the area actually used by livestock. In this case, two key areas (crosses) were selected to represent the two ecological sites.
In this example, the two key areas are assumed to be representative of the impacts that grazing is having on the portion of the total area that is being used, and as long as reasonable precautions were taken in the selection of the key areas to avoid bias and the site selection process was well documented, these sites could serve as a defensible platform for monitoring grazing on the allotment.
It is important to note, however, the extent to which inferences can be drawn from these key areas. Even though significant consideration might have gone into selecting the key areas, unless they were selected randomly from a large set of possible areas, the extension of key area conditions to the larger area represents a subjective assessment. It is subjective because there is no way to quantify the bias (or lack thereof) of the key areas relative to the condition of the entire area. In a strict statistical sense, the inference space of these areas is restricted to the key areas themselves and no valid, objective inference can be drawn to other areas that were not sampled. If statistically rigorous inferences are not required and if all stakeholders agree on the location of the key areas, then the subjective nature of extending key area conditions to larger landscapes may be acceptable.
An alternative strategy to key areas for monitoring the allotment could be to draw a set of random samples within each of the ecological sites (i.e., a stratified random sample, Figure 2).
Figure 2: An example of a set of stratified random points (blue dots) selected for the example system in figure 1. In this example, three random locations were selected within each ecological site.
Considering the grazing example, a stratified random sample may not be as efficient as the key area approach because it involves sampling more locations, but it offers several important advantages:
Because the sample locations could be located anywhere in the entire area, the stratified random sample preserves the inference space as the full extent of the allotment. In other words, conditions measured at the random sample locations can be statistically extended to the larger area of interest.
As long as the sample locations were selected using a randomized procedure, estimates derived from the random sample locations are statistically unbiased.
Because they were selected at random from all possible locations within the allotment, the random sample locations can be used to address several different assessment and monitoring objectives. In contrast, because the key areas were selected to evaluate the impacts of grazing, they can only be reliably used for that purpose (and only while the conditions under which they were originally selected remain).
This final point is an important limitation of key area monitoring to being able to respond effectively and efficiently to different monitoring needs. Consider the following extension to the grazing monitoring example: a new natural gas well is located in the northeast corner of the allotment, the existing road is widened and improved, and a new access road to the well is installed (Figure 3).
Figure 3. The same system and sample locations as in Figure 2 above, but with the addition of a new natural gas well and improved roads.
The gas well and the improved roads both have influence over conditions in a portion of the allotment. These areas may overlap the influence zones of grazing. The random sample locations are capable of recording any changes in condition associated with the new well because they are stratified by ecological site and located in the allotment without regard to any particular management activity. On the other hand, the key areas, because they were placed to monitor grazing management, do not provide a good indication of representative conditions with regard to impacts of the gas well. Additionally, the influences of the gas well and improved roads may affect the key areas and make them no longer very representative of grazing impacts. This kind of effect could lead to determination of adverse grazing effects when in fact, the change was due to some other management activity!
To sum up, some important considerations when selecting sampling locations are:
Key areas are (ideally) sensitive to changes in management related to a specific monitoring objective, but only if the system changes with respect to that objective. If the system changes due to some other factor, the key areas may not pick it up or may attribute the change to the wrong management activity
Changes in the system not related to the original monitoring objective can affect whether or not key areas are representative of conditions related to their original management activity.
Randomly-located monitoring locations may not be quite as efficient as key areas for detecting change relative to a particular management activity (although this depends a lot on things like how variable the system is, how strong the management impacts are, etc.), but they are much more robust to changes in the original management conditions of the area.
Because random locations are not placed with respect to any particular management activity, they can be used to monitor changes due to multiple impacts.