WMOD00593

• Contract/Permit #: 14-06-D-7052
• Title: Ecological Impacts of Snowpack Augmentation in the San Juan Mountains, Colorado
• Date: 3/1/1976
• State: CO
• Weather Modification - Doc Type: Report

the progress made in ecological modelling in the past
five years. Cooper et al. (1974, p. 23) recommend
extensive use of simulation models in projects akin
to SJEP and we support that recommendation. How-
ever, it should be remembered that simulation models
are no better than the data that go into them,
either as estimates of parameters or as validation
checks. Further, ecological models should not be
expected to approach the precision of hydrologic
simulations, if only because of the complexity of
the systems they treat.

Other Considerations

Three other research strategy topics are briefly
mentioned here. They are the use of standardized
procedures, the extrapolation of research results,
and the statistical interpretation of those results.

-Standardized Procedures

In the preparation of environmental impact state-
ments, the standardized matrix procedure of Leopold
et al. (1971) has received much attention in the
past five years. It offers a way of evaluating the
relative magnitudes of environmental impacts by
matching biological organisms and communities to a
list of possible impacts. Each cell of the matrix
represents an expected effect which is then scored
by the ~.gnitude of impact and/or the significance
of the organism. The most susceptible organisms
can then be identified by a summation of the scores.

Cooper et al. (1974) point out that this approach is
particularly useful in a small, well-defined project
but is less easily applied to multidisciplinary
research where a wide variety of interests and
judgements are involved. A matrix evaluation was
applied in a limited way as part of the definition
of research needs during Phase I of SJEP. A matrix
was designed for the project and some ecosystem
components which later work showed to be susceptible
to the effects of snowfall augmentation, were
successfully identified in it. The procedure was
not applied to the entire project but, in those
parts where it was used, our experience shows it to
have been useful.

-Extrapolation of Results

There are three reasons for extrapolating the
research results presented elsewhere in this report.
First, it may be necessary in the future to predict
potential impacts of cloud seeding in other mountain
ranges from SJEP results. Second, the characteris-
tics of simulation models will often need to be
estimated from observations made in areas other than
that to which the model is applied. Third, there
is the need to estimate impacts over an entire
mountain range from results derived from small study
areas within it. The last of these has been of
direct concern in this project; the first two refer
to the potential applications of the project results.

The need for extrapolation is a strong reason for
including error estimates with all substantive
predictions of environmental impact and we have
attempted to do this in Chapter IV of this report.
These error estimates are, of course, based on data
derived from study areas, but it is probably safe to
assume that the errors involved in extrapolation
will be at least as wide as those in the original
analyses. In any case, it is essential that these
estimates be made.

'1.

The studies made as part of SJEP have been conducted
on spatial and temporal scales varying from broad-
scale mapping of the entire San Juan Mountains to
detailed plot studies of individual plants. We have
found it difficult to draw these levels of inquiry
together in a way which would allow an extrapolation
of detailed results to impact evaluation for the
entire target area of cloud seeding. The problem
has not been treated explicitly in this report and
is one that is clearly worth further study.

One solution is to treat the problem of research site
selection as a statistical one. This requires that
study areas be selected so as to be representative
of the target area for cloud seeding. This was
attempted in SJEP but without complete success. As
a result, the areas chosen for detailed field study
in the alpine tundra ecosystem were later found to
be atypical in some characteristics (Caine,
Appendix A, this vol. p. 19~. If sufficient
information were available at the start of a project
like this one, it would be possible to identify
representative sites statistically. In SJEP, the
necessary information did not become available until
late in the project's life, through the work of
Andrews (this vol. p. 87) and Krebs (this vol.
p. 81). This should not be interpreted as a
criticism of SJEP since the overview projects were
intended to give a post-facto evaluation only.

An alternative solution is that which has been taken
to rationalize the procedures used in SJEP. Since
most work in the project has been concerned with
ecological processes, it is probably sufficient if
typical species and communities, especially ones
that are well known or have been studied elsewhere,
are included in the study, whether the area in which
they are studied is typical or not. In fact, we
feel that it is even more important that species,
communities, and situations that are susceptible
to environmental change be included. In this kind
of decision, the experience of an ecologist should
prove more efficient than the best of experimental
designs.

-Statistical Interpretation

In reporting statistically tested conclusions, the
SJEP studies generally follow accepted scientific
convention in requiring rejection of a null
hypothesis (H ) with some earlier defined risk of
error (common~y, p = 0.05), The null hypotheses
used in the individual reports of Chapter IV are
normally those of "no difference between experimental
and control situations" or "no correlation between
snowpack and response variables." This approach to
scientific analysis is unlikely to be criticized,
except for the decisions to use prior probabilities,
and the choice of critical rejection levels. These
questions have been discussed frequently (e.g.
Plutchik 1968, p. 112) and need not be considered
here.

I

i

However, when environmental impacts must be
estimated, it is important to remember that'acceptance
of a null hypothesis does not imply its proof. There
is a tendency to interpret "no proven effect," i.e.
acceptance of H , as if it were "proven no effect,"
i.e. proof of HO. The latter statement would, of
course, requireOthat errors in decision making be
estimated on the other tail of the probability
function used in statistical testing (i.e. p = 0.95,
rather than p = 0.05). In an evaluation of the
ecological impacts of snowpack augmentation, this

30

'?