Geography Reference
In-Depth Information
entire species at 15,800, and other published reports had suggested that white-lipped deer
were rare and declining, and that most lived in Sichuan and eastern Tibet with only a
few in Qinghai. Alas, the county agricultural topic did not provide enough information to
clarify why their estimate was so high (stating only that “extrapolations were conducted
to the entire county based on twelve sample areas”). I began to suspect something fishy
in selection of the “sample areas” when, the following year, I observed a group of over
200 white-lipped deer traveling across a slope only a few kilometers from the govern-
ment-run guard station. That day, on that slope, the density of white-lipped deer was
quite high. But in and of itself, this did not mean that it was equally high throughout the
county. I have little doubt that the authors of the county agricultural yearbook saw many
white-lipped deer, but I am almost certain that they “sampled” in just those places where
they knew they would likely find what they were looking for. 5
Thereafter, I began to query authors of wildlife studies that claimed to have conducted
“random sampling” whenever I had the chance. Twice, I had authors inform me that what
was published as representative ( daibiaoxing ) or random sampling ( caiqu suiji bushe
yangdian ) was, in fact, very much biased. 6 The author of a study that estimated musk
deer density in Sichuan (indirectly, based on fecal-pellet counts) wrote to me that he and
his students were “very lucky to receive the help of local people in locating sites where
musk deer were active.” A similar story was related to me by the co-author of a study
estimating density of argali in Qinghai, where campsites were “recommended by local
people.” It requires no additional mathematics to imagine how inaccurate an extrapola-
tion will be when based on samples obtained only where animals are suspected to be
congregated in the first place.
Distance sampling to estimate the density of wild animals is a valid technique only
to the extent that its underlying assumptions are not grossly violated in its field applica-
tion. Exposition of its fancy-looking equations and presentation of confidence intervals
in results impart a sense of scientific rigor to a paper, but only those who have done the
field work can ultimately judge whether the data were gathered in a way that justifies
the implied sheen of scientific precision. If the field conditions simply did not allow for
collecting the type of data assumed by theory, if biologists have had no choice but to
make a hash of the critical assumptions, no equation or computer-generated output can
rescue the effort. Equations can be presented and calculations can be made, but too large
a chasm between the field data and the elegant model dooms the resulting numbers to
meaninglessness. That unfortunate fact can be pretty frustrating for someone who has
spent considerable time and withstood considerable hardship to obtain the data, and it is
compounded for someone who wishes to be seen as a scientist.
Biometricians have been adding continued sophistication to the mathematics of distance
sampling for the last quarter-century, but have made only scant progress in allowing for
valid inference when fundamental assumptions cannot be met. Generally, the more so-
phisticated the model, the greater the need for large sample size or auxiliary data. Often,
Chinese wildlife scientists appreciate the sophisticated mathematics but have no way to
obtain the extensive data needed to support them. Hemmed in by authoritarianism that
requires them to be experts and a belief that complexities can be simplified on the one
Search WWH ::




Custom Search