The Determination of Minimum Flows for Sulphur Springs, Tampa

DRAFT
average daily withdrawals for two series of dates in Figure 2-24 A and B. The first
pumping episode involved pumping over a four-day period during January and February
2001 (Figure 2-24A). Initiation of pumping on January 31 resulted in an almost
immediate drop in water levels, with the rate of decline slowing after the initial fall. The
response of specific conductance was not so immediate, as a steady rate of increase
that totaled about 400 µmhos/cm began approximately one day after pumping began.
Cessation of pumping on February 3 resulted in a quick rebound in water levels and a
slower decline in specific conductance.
The time series shown in Figure 2-24B is for an 8-day pumping episode during June
2001. Again, water levels showed an immediate drop in response to pumping, followed
by a slower, but continued, decline. Compared to the four-day winter pumping episode,
in which water levels fell to 5.9 ft, (Figure 2-24A), water levels fell to 3.9 ft. after eight
days of pumping in June (Figure 2-24B). Specific conductance again showed about a
one-day lag, with a steady increase that totaled about 1,100 µmhos/cm over the
following seven-day period. Again, cessation of pumping resulted in a rebound in water
levels and a decline in specific conductance.
In order to gain some insight into the effects of withdrawals from the spring, it is helpful
to filter out the seasonal changes that occur in both the flow and specific conductance
values for the spring. Plots of daily specific conductance values vs. day of the year are
overlain for the years 2001 and 2002 in Figure 2-25. The curves for these years are
very similar, reflecting the periodicity of a seasonal component. Some of the unusually
large spikes in specific conductance correspond to those times that the spring pool was
manually lowered.
In addition to the plots for the two individual years, there is also shown a LOESS
(Locally Estimated Scatter plot Smoothing) regression line of the data for the year 2001.
This year was chosen simply because there were no significant manual pool lowerings
during that year, although there were a noticeable number of withdrawals. The LOESS
regression line bears a similarity to the seasonal pattern of average monthly values
shown in Figure 2-15. It is reasonable to assume that this regression model can
represent the seasonal component of specific conductance. If this assumption is valid,
then by examining the residuals, or the difference between the actual data and the
values predicted by the model, one could examine how other factors may influence
conductance in addition to the seasonal component.
The residual conductance values for 2001 are plotted in Figure 2-26 along with daily
pumpage records from the springs. The pumping events have been shifted by one day.
That is, today's specific conductance is plotted against the previous day’s withdrawal
event. This takes into account the slight lag time of the response of the water quality to
the withdrawal event. During the first half of the year when most of the withdrawal
events occurred, the number of spikes in specific conductance is noticeably higher
when compared to the latter half of the year. Equally obvious is the direct
correspondence between the pumping events and the specific conductance spikes.
2 - 27