Adaptive Sampling Methods - Introduction to Ecological Sampling - page 40

Restricting the final sample size by a stopping rule has been discussed by
Salehi and Seber (2002), Brown and Manly (1998), Lo et al. (1997), and Su and
Quinn (2003). Another approach is an inverse sampling design, in which sur-
veying stops once a set number of nonzero units has been selected (Seber and
Salehi, 2004; Christman and Lan, 2001). More recently, Gattone and Di Battista
(2011) have suggested a data-driven stopping rule that can be used when there
is limited information about the rarity and clustering of the population.
There have been many applications of adaptive cluster sampling to a range
of environmental situations. Each offers an insight into the design and appli-
cation of adaptive cluster sampling. Some recent examples are the use of
adaptive cluster sampling for surveys of freshwater mussels (Smith et al . ,
2003, 2011; Hornbach et al . , 2010; Outeiro et al . , 2008); rockfish (Hanselman
et al . , 2003); rare fish (Davis and Smith, 2011); a highly heterogeneous popu-
lation of yellow perch (Yu et al . , 2012); fish eggs (Smith et al., 2004; Lo et al.,
1997); larval sea lampreys (Sullivan et al . , 2008); larval walleye pollock (Mier
and Picquelle, 2008); subtidal macroalgae (Goldberg et al . , 2007); sediment
load in rivers (Arabkhedri et al., 2010); forests (Talvitie et al . , 2006; Acharya
et al., 2000); deforestation rates (Magnussen et al . , 2005); shrubs (Barrios et
al., 2011); understory plants (Abrahamson et al., 2011); herbaceous plants
(Philippi, 2005); annual plants (Morrison et al., 2008); bark beetle infestations
(Coggins et al., 2011); giant panda habitat use (Bearer et al., 2008); marsupials
(Smith et al . , 2004); waterfowl (Smith et al., 1995); and herpetofauna (Noon et
al., 2006); and in hydroacoustic surveys (Conners and Schwager, 2002).
3.3 Other Adaptive Sampling Designs
Adaptive sampling can be applied to stratified and two-stage sampling in
the same way that the adaptive component is added to simple random sam-
pling in adaptive cluster sampling. In stratified and two-stage sampling, the
population is partitioned into distinct subareas, called strata in stratified
sampling and primary units in two-phase sampling. The two designs, strati-
fied sampling and two-phase sampling, are similar in how the survey is con-
ducted in this respect. With stratified sampling, all of the strata are surveyed;
in two-phase sampling, only some of the primary units are surveyed.
The idea behind adaptive allocation sampling is that strata or primary
units that contain clusters of the rare population are preferentially allocated
extra survey effort. Ideally, if the locations and sizes of all the clusters were
known, strata or primary unit boundaries could be marked around each clus-
ter. Additional effort would then be allocated to the strata or primary units
that contained clusters. More often, the locations and sizes of clusters are not
known, and the study area is partitioned into strata and primary units in a
way that minimizes the within-primary unit or the within-stratum variance.