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

Table 1. Data on weather stations used in this study

Approximate Distance Record
Weather Station Elevation from Coring .sites Period
Durango 1991 m 92 km 1885-1974
Pagosa Springs 2005 m 42 km 1941-1974
,Vallecito Dam 2331 m 29 km 1943-1974

Climatic Correlation Analysis

The possibility of retrieving climatic information
from growth layers depends largely on site factors.
Years of greater than normal precipitation are more
difficult to evaluate since growing conditions are
not necessarily enhanced by excess precipitation.

Because of the interrelationships between temperature
and precipitation, their effects are not always
easily separable. In the San Juan area, a comparison
of temperature and precipitation records of several
weather stations indicates an inverse relationship:
warmer periods are associated with low precipitation,
colder periods with high precipitation (Barry and
Bradley (this vol'. p. 43). In the Southwest, where
moisture is the limiting factor (except at the upper
timberline), the growth layer thickness is generally
inverse to temperature (Fritts 1972). The relative
importance of the temperature and precipitation varies
at different times of the year.

Detecting the presence of precipitation in excess of
normal requirements for the life processes of trees
through growth layer analysis is difficult. In the
San Juan Mountains, thick growth layers represent
high precipitation. If the climatic regime is
reflected in growth layer data prior to cloud seeding,
the altered climatic regime due to cloud seeding
should be reflected in the growth layers formed during
the seeded years. Preliminary evaluation of the
regional growth response indicates that the annual
increment of wood is greater with increased pre-
cipitation. Thus, the mean annual growth layer
width should increase with an increase in annual
available moisture. An attempt was made to predict
the growth layer thickness during the seeded years
1971 through 1973. Three seeded years do not
constitute a statistically valid sample to conclusively
state that such a growth layer increase is correlated
with increased precipitation. The proposed method
does merit further consideration.

-Trend Analysis: Preliminary Evaluation

Trend analysis is a method of nonparametric correlation
for continuous time series (Glock 1942). It requires
neither the elimination of secular trends, such as the
age curve in trees, nor the estimation of a mean. The
analysis gives a measure of the degree of parallel
fluctuation between pairs of variables (e.g. tree
growth and precipitation, tree growth and temperature)
in successive time intervals (Glock 1950). As in
other types of correlation, a high coefficient suggests
but does not prove relationship. The trend coefficient
is useful for preliminary work involving the compari-
sOn of many sets of data.

The trend method yields a coefficient, t, which
measures the amount of covariation between two
variables. The coefficient is given by

t

N
E (Xi - Xi_I) (Yi - Yi-1)
i=2

~ I(Xi - Xi_I) (Yi - yi-1)1
i=2

where N = number of years in record, Xi = value of
first variable for ith year, and Yi = value of
second variable for ith year. Thus the trend co-
efficient takes into account not only the number of
cases of correspondence in the direction of variation,
but also the degree of that correspondence. The
trend coefficient ranges from +1.00 to -1.00 with
the extreme values indicating perfect parallel
variation and perfect inverse variation, respectively.

The influence of a few departures of high amplitude
upon the trend coefficient may be detected by a
trend index (i). This is the ratio of the average
departure of parallel trends to the average de-
parture of opposite trends, i.e.

i

E( + XY) / n+
E( - XY) / n

where E ( + XY) = sum of parallel trends, n + = number
of cases of parallel trends, E( -XY) = sum of
opposite trend values, and n = number of cases of
opposite trend. If i is greater than 1.00, the
amplitudes of departures of parallel trends are
greater than those opposite. If i is less than
1.00, the reverse is true. If the value of i de-
parts much from 1.00, then a few departures have an
undue influence in determining the value of the
trend coefficient. Because the trend coefficient
may lead to an erroneous conclusion as to the
existence of parallel variation throughout a series,
the value of i should always be determined. In
instances where i is far above or below 1.00 and
the ratio n+/n_ is nearly 1.00, the calculation of
the ratio T = t/i provides a better comparison of
the relationship between the variables than does t
by itself.

J

1
]

Many sets of climatic data were evaluated and com-
pared to make selections for the detailed den-
droclimatologic analysis. Weather records for six
recording stations were used in this evaluation.
The precipitation and temperature data for each
station were arrayed in sets incorporating 15
months as follows: (1) 3 months of the pre-
ceding growing season for each tree ring, (2) 6
months of the winter prior to the growing season for
each tree ring, and (3) 6 months during the grow-
ing season for each tree ring. The tree ring data
represented the three coring stations described
previously. The rapidity and ease of direct compari-
sons using the trend method allowed the grouping of

72