119
Reliability. Reliability is largely a product of effective and representative sampling, but
there are additional properties that a reliable survey protocol should have. It should be
reliable in the sense that if applied, it will produce interpretable results. At the finest
scale (traditionally the scent station but, in our design, the transect) if a lynx is present in
the area, the probability of detection should be as constant as possible. This allows the
proportion of occurrences to infer use. At a larger scale, we want to reliably state that,
given a certain level of effort, we will have detections if lynx are present, and therefore a
lack of detections indicates a lack of lynx.
At the fine scale, placing scent stations 100 m apart and perpendicular to the major slope
produces a structure that will be encountered by lynx moving through the country and
removes small-scale differences associated with station placement. At a broader scale,
the protocol requires placing no fewer than 25 transects at a density of 1 transect per
every 2 miles for a period of 2-4 weeks to ensure that an area is adequately sampled.
Details
Broad decisions concerning where to sample
Decisions as to where to sample are based primarily on the interest of the managers. If
grids (25+ transects) were placed randomly within a major cover type, the grids
themselves would be a representative sample of the cover type. A manager may,
however, need information about lynx in a specific area, and can place grids
preferentially. In broken habitat, such as forested areas separated by low elevation
prairie, dry forest types or deciduous forests not thought to be lynx habitat, or lands
which have been converted to agriculture, the sampling does not need to conform to a
rectangular grid. All that is required is that the placement within the lynx habitat be at a
density of about 1 transect per every 2 miles. An easy way to accomplish this is to put a
large 2x2 mile grid across the landscape and use only those points which fall into habitat
as the sample. In all cases the grid should start at a random location. Do not move the
grid to get the highest number of points in habitat. One approach that may work well is
simply to use section boundaries as the grid. If these boundaries are not associated with
vegetation changes, then they can be thought of as random. If, however, there are
specific features that are generally associated with section boundaries, such as changes in
