EXPERTS, ASSESSMENTS, AND MODELS - Dams and Development in China - page 107

cial networks in response to dam-induced displacement, we would need to collect baseline
data from before construction on the dam was initiated and then collect similar data after
completion of the dam and displacement. The costs of such an undertaking—in time, en-
ergy, and financial resources—would be enormous. We opted instead to use a cross-sec-
tional study design that allows us to compare multiple dam sites at a single snapshot in
time, an approach that, as I described in the Lancang case study ( chapter 3 ) , does not lend
itself to measuring diachronic change.
We also encountered the related problem of cumulative impacts. When several large
dams are built on the same river or even on tributaries of the same river system, the im-
pacts—on water quality, threatened species, and human communities—accumulate. But
EIAs and other scientific studies are generally conducted for single projects, not for mul-
tipleprojectsinthesamebasin.Tocomplicatematters,noonereallyknowsabouttheshape
of the impact curve: Do the effects suddenly ramp up with the addition of one more dam,
crossing a kind of threshold that causes irreversible perturbations in the system, or do the
effects accumulate more gradually? Does the first dam built in a basin cause significant
impacts, while the effects of subsequent dams are only marginal? 9 As it turns out, these
crucial questions and many more just like them remain unanswered.
Data Visualization
Because our goal was to turn a set of complex data into a series of patterns that decision
makers can easily understand and interpret, we needed an effective visual framework. Our
team initially preferred a visualization technique called a “spider diagram,” which plotted
both magnitude and salience on a circular graph resembling a spider web, with each of the
twenty-one impacts (seven each in the biophysical, socioeconomic, and geopolitical sys-
tems, as noted earlier) plotted along the outside circumference of the web and the salience
values plotted along the radius. However, when we conducted experiments in several large,
undergraduate university courses to test the effectiveness of this visualization strategy, we
found that most people had trouble interpreting such diagrams. Users with experience in
graphic design, abstract reasoning, or mathematics tended to interpret these outputs with
greater ease and accuracy, whereas people lacking such a background struggled to make
sense of the data conveyed in the diagrams—an unacceptable state of affairs if our goal
was to speak to policy makers and other constituent groups without doctoral-level training.
Bar charts—with magnitude dictating the height of the bar and salience displayed as color
shading—proved to be the most suitable data-visualization technique, one that transcended
academic training and cultural background. Model users can input data for magnitude and
salience either numerically or by clicking and sliding a set of bars within the computer in-
terface.