Geyser Activity from Data Loggers
by Ralph Taylor
of you may know, I have had a research permit for the past four years to
monitor geyser activity using electronic data loggers. For those years I also assisted Tim
Thompson, the Geothermal Technician, by downloading and redeploying the
National Park Service data loggers late in the season after Tim left the
Park. This summer one of my jobs as an
NPS volunteer for Nancy Hinman was the deployment and management of the NPS
data loggers in addition to my own loggers.
At the start of the summer we had 16 loggers between my loggers and the
NPS loggers. Over the course of this
summer we acquired several more; the total is now 29 loggers, 25 of which are
currently in use (the other four just arrived).
a little introduction to data logger terminology is in order. A data logger is a device that
records some kind of information (in our case, temperature) at user-specified
intervals. The logger itself uses a
single chip microcomputer and a memory chip to record the data as
required. The sensor for the loggers we
use is a thermistor, an electrical resistor with the property that the
resistance to electrical current varies with temperature. The logger measures the temperature by
measuring the resistance of the thermistor and uses a calibration chart or
equation to determine the actual temperature.
The logger is deployed by placing the thermistor in a location
where the runoff water from an eruption or overflow passes over the
sensor. The logger itself is located in
a sheltered location nearby. The NPS
requires that data loggers on front country features be placed out of sight, so
we typically bury the logger in sinter sand, cover it with rocks or logs, or
otherwise attempt (usually with fair success) to keep it out of sight.
logger must be launched to start its task of data collection. To launch the logger, the logger is
connected to a computer using a serial port cable. Software provided by the logger manufacturer communicates with
the logger and allows the user to specify the intervals at which the
temperature is measured, whether to log temperature in degrees Celsius or
Fahrenheit, and to specify a title for the data set that is collected. Once launched, the logger samples the
temperature at the specified intervals and stores the information in its
appropriate intervals the data in the logger must be collected. This is downloading, and is done by
connecting the logger to a computer using the serial cable and performing the download
using the software provided by the manufacturer. Some of the loggers we are using can be downloaded and relaunched
using a pocket-sized shuttle that transfers the data from the logger to
the shuttle memory and relaunches the logger.
The shuttle must then be connected to a computer to recover the data.
data collected by the download process is in a binary file in a format defined
by the logger manufacturer. The logger
software that we use can graph the temperature as a function of time, allowing
zoom and pan to examine any portion of the data closely. The temperature data can also be exported as
a text file containing date, time, and temperature data.
geysers the temperature graph is fairly easy to interpret. For example, the graph shown at the right is
from Grand Geyser using the Hobo Pro data logger that GOSA purchased for the
Old Faithful Interpreters. The graph
clearly shows two eruptions of Grand Geyser (the two tall narrow peaks), followed
by Vent and Turban afterplay. Next, there
is a period of no overflow, then a shorter peak (about 40°C)
caused by the Vent and Turban restart.
Finally, the long period of inactivity follows, and then the Turban
cycles are visible once overflow resumes.
zoom capability allows one to look at just a portion of the graph, for example,
the actual eruption start, and to determine the time accurately. The graph shown at the right is the first
eruption from the previous graph with the actual temperature points
marked. From the graph it is clear that
the time of the sharp rise in temperature was 11:08. The actual start time of the eruption must be compared with the
time from the logger to determine time lag from when the eruption starts until
the hot water reaches the sensor (for Grand, the sensor is under the boardwalk,
and water takes well over a minute to reach the sensor). Electronic times are recorded in the OFVC
logbook with an “E” after the time to show that the times are derived from a
data logger. These times are not
corrected for lag, and are also affected by the sample time interval. For Grand, we generally sample at 30 second
intervals, so the electronically detected start time may be an additional 30
seconds late. These lag and sample time
errors affect the actual start time detected, but do not affect the calculation
of intervals and averages.
of eruption times by examining the graph is time consuming and is impractical
with almost 30 loggers deployed, some on geysers that erupt more than 30 times
a day like Plume. Over the years that I
have been using loggers, I have written software to detect the eruptions
automatically, and this allows the information from each week’s downloads to be
analyzed and summarized in about one (long) day. The summary statistics are the maximum, minimum, mean (average),
and median interval for the past week, month, and season. The median value is the value that is the
middle value in a sorted list of the intervals. The median is often more indicative of the “typical” interval
than the arithmetic average, since one or two exceptionally long or short
intervals can shift the mean, but the median is not so affected. The summary table is kept under the glass in
the “ice cream window” at the Visitor Center and is posted on the GOSA website.
Some Results from Summer
we have covered all the boring details of how these gadgets work, let’s look at
some results. For this article we will
discuss only the results for the 2000 summer season; the results from previous
years will have to wait for a future article.
are monitors on 19 front-country geysers at this writing. Some have been in place since May, others
were acquired during the summer and have only been in place a month or so. This article covers only the highlights; a
more in-depth examination is more suited for a GOSA Transactions
had an interest in Geyser Hill for many years, and my research permit is
primarily aimed at monitoring activity there.
There are loggers on Aurum, Lion, Little Cub, Depression, Plume, Plate,
and Boardwalk Geysers.
have come to expect, dry weather resulted in long and erratic intervals for
this geyser. The logger was first
deployed in late June after the wet weather ended. There were only a handful of intervals in the usual “short
interval” range of 2:30 to 2:45, (six under 2:35, and only 24 between 2:35 and
3:00. During the previous few years,
these short intervals have been much more common, and recurred for a few days
following heavy rain. This year, even
after rainfall, the sub-2:45 intervals did not occur. There were ten intervals over eight hours, one reaching 11:58 on
4 August. For the three months for
which data is available there was a slight trend to increased intervals, with a
few periods of a 2-3 days of shorter intervals following wet weather.
an autumn snow starting on 21 September, with about a foot of accumulation at
Old Faithful (and with most of the roads closed for a day). Starting at midnight on the 22nd,
about the time the snow started to accumulate, Aurum’s intervals dropped to
between 2 hours 31 minutes and four hours.
The short intervals only continued until the snow melted; long intervals
resumed at noon on 24 September.
logger on Lion has a most unfriendly environment, as the logger is immersed in
very hot water on every eruption. I had
one logger destroyed here, but the current logger, in a waterproof case, has
survived for both 1999 and 2000 so far.
This summer’s activity was fairly uniform, with only a few really long
intervals between series—two intervals in late July were well over 12 hours,
but the great majority of initial eruption to initial eruption intervals were
between six and ten hours, with the median series interval being 7:29. The intervals between series gradually
decreased from about 8:30 early in the summer to the present average of around
change this summer from the behavior in the past few years was the total
absence of long series. The longest
series recorded was five eruptions, and there were only three of these, all of
which included a minor eruption. Fully
58% of the series consisted of two eruptions.
As has been the case for the three years that I have monitored Lion,
there is only a weak relationship between the number of eruptions in a series
and the following interval.
Cub is a good example of where a data logger can reveal information about a
geyser that is often seen but not often recorded. The logbook has large numbers of “in eruption” and “near start”
notations, so there are few intervals available from that source, and almost
none during the night hours. With the
data logger it is possible to record intervals and durations with pretty good
accuracy (the end of the eruption is difficult to determine exactly from the
year, Little Cub’s intervals have been fairly steady at about 1 hour 22 minutes
all summer, until late September. Early
in the season, until about the first of August, there were occasional short
intervals, some under one hour, which tended to occur in the afternoon between
12:00 and 18:00. This trend to a
diurnal pattern early in the season has been apparent in 1998 and 1999, to a
greater extent than in 2000. After
early August, there was a cyclic variation of intervals with a period for the
whole cycle of about one week, plus or minus a day or two. The periods of longer intervals appeared to
coincide (within a day or so) with SMax as indicated by eruptions of Little
8 September there was a cycle with noticeably longer intervals, and especially
with an absence of short intervals. The
intervals returned to normal 1:25 or so until 19 September when there was an
interval of 1:51 recorded, followed by two days (again, the snow days…) when
the intervals rose to about 1:40. The
graph at right shows the Little Cub intervals for September. Note the large peak (with one interval at
almost three hours). It is not certain
that there was no eruption at that time, but there is no trace of it on the temperature
show that Little Cub has had its short intervals (below 1:15) between 12:00 and
18:00, and its long intervals (over 1:35) between 14:00 and 22:00. These are a few interesting anomalies for
research minded geyser gazers to ponder.
the summer of 2000,Depression Geyser was very regular, with a median interval
of 5:59 and a standard deviation of just 27 minutes. The great majority of intervals fell between 5:30 and 6:45. As usual, the long intervals were associated
with strong west winds. There appears
to be a cyclic variation of intervals with about a three-day period, but the
trend is not strong. Recorded durations
were all in the four-minute range.
has been steady at about 40 minutes for the whole season. The daily range in intervals tends to be
within a 12-minute spread. There was a
trend to shorter intervals during the last two weeks of June (monitoring
started on 16 June), then the intervals stabilized in the 38-minute daily
median range. There is a noticeable but
small (about 3 minute) variation in the intervals on a three to five day cycle;
the minimum intervals tend to occur around the time of Little Squirt
eruptions. A few instances of long
intervals associated with heavy rain (apparently due to cold rainwater entering
the system) occurred. There were no
Giantess eruptions, and no other geysers appear to have affected Plume’s
intervals. A pronounced dip in intervals
to less than 40 minutes around 8 September did not appear to correlate with any
other monitored geyser activity. The
intervals rose over the following week to about 44 minutes, and then returned
to 41 minutes or so late in the week. A
similar dip in intervals occurred on 20-21 September, followed by a gradual
increase in intervals to 43 minutes by the 23rd, with some intervals
as long as 50 minutes by the 25th.
On the 23rd and 24th the intervals varied from 37
to 47 minutes. It is interesting to
note that the sudden changes in Plume’s intervals occurred at about the same time
as the sudden changes in Little Cub’s intervals.
Plate and Boardwalk
two geysers are closely related spatially, lying about 10 meters apart along a
fracture line leading toward Giantess.
They also exhibited closely related activity this summer. Plate Geyser had two oddities in its
otherwise regular intervals this summer.
First, there were occasional long intervals for the first month that I
had a logger on Plate. These occurred
at approximately weekly intervals.
During that time, Boardwalk was being seen infrequently. Logbook entries for Boardwalk corresponded
with long Plate intervals and with one short interval of just 12 minutes.
graph nearby shows the Plate Geyser activity for the summer. The intervals were consistent except for two
short intervals until late July, when long intervals became more common. For the part of June and most of July the
intervals hovered around 1:30, then began to climb to 2:00 where they remained,
varying cyclically at about weekly intervals.
This variation may have been related to SMax (shown by the dots
at the top of the graph). It appears
that the longer intervals tended to occur near SMax.
In August I acquired another
logger and deployed it on Boardwalk Geyser.
Once results from that logger were available it was clear that all of
the long intervals of Plate Geyser were coincident with Boardwalk eruptions,
and that the occasional very short intervals of Plate occurred when Boardwalk
either began or ended its eruptions during a Plate eruption. Not every instance of Plate and Boardwalk
concerted eruptions resulted in the odd double eruptions of Plate, but many
graph to the right is the temperature graph of both Plate and Boardwalk,
showing how the Boardwalk eruption coincides with the Plate short
interval. The top trace corresponds to
the right scale and represents Plate Geyser; the bottom trace is Boardwalk
Geyser. In this case, Boardwalk started
first at about 18:55, then Plate started at 19:06. When Boardwalk ended at 19:11, Plate paused for two minutes,
then restarted. Using the same
combination of the two curves it became clear that Boardwalk’s eruptions caused
Plate to have longer intervals following Boardwalk’s eruptions.
course of the summer, Plate’s intervals show a gradual but continuous increase
from about 1:30 in June to about 2:30 in late September. The Plate Geyser durations have also
increased, from about six minutes in June to about 6:15 in September. Boardwalk’s intervals have varied wildly,
ranging from 8 to 32 hours with no discernable pattern.
Other Upper Geyser Basin
are data loggers on Castle, Daisy, Grand, Grotto, and Oblong Geysers. The following sections briefly discuss the
Castle logger was deployed rather late in the season, so we have data from only
August and September. During that time,
Castle’s intervals have remained firmly at just over 12 hours excluding minor
eruptions. There were four intervals
between 13 and 14 hours, and one at 11:30, but the rest were between 11:40 and
12:30. There have been 17 minors
between 13 August and 26 September. The
interval from the minor to the subsequent major eruption ranged from 3:12 to
10:09 with the mean interval being 6:33.
The mean interval between the first two major eruptions following the
minor was 12:58, or just about one hour longer than the normal intervals. We hope to maintain a data logger on Castle
for the whole winter, so any seasonal change should be captured.
Geyser has been quite steady at about two hours. There have been occasional intervals of as much as five hours (on
17 May) and numerous intervals between 2:25 and 2:30, mostly before 20
June. There were also several long
intervals (2:30 to 3:15) on 4 and 5 August.
The only recent long interval was 3:22 on 17 September at 10:19. On the whole, Daisy has been quite regular,
rarely falling outside the prediction window.
Grand Geyser, on the other
hand, has been difficult to predict this summer. It had a very sudden change in interval from about 14 hours to
about 9 hours on 30 July, then just as suddenly switched back to about 14 hours
on 28 August. The intervals declined
again, but this time gradually over a week’s time starting on 14 September, but
then reverted to 14 hours on 26 September.
Note also in the graph that there was a noticeable shift from 12 hours
to 14 hours at the end of June. There
were a few technical problems with the logger that caused us to miss several
eruptions; the gaps were filled in by data provided by Mary Beth Schwarz and
logger on Grotto Geyser showed that there have been marathon eruptions about
every other day. On one occasion, on 7
and 8 September, there were consecutive “marathon” eruptions of 14:32 and 13:55
duration respectively. Apart from this
one occasion, “marathon” eruptions were separated by four or five “normal”
Geyser has had an unusual summer. The
data logger has been deployed since 13 July.
During that time, Oblong has been erupting in series, with from one to
five eruptions in a series. For most of
the summer the series were two eruptions long, with the second eruption
following the first by about three hours.
After 22 July, the intervals between series ranged between 16 and 52
hours, much to the dismay of Oblong fans.
The graph shows the scatter of the intervals clearly.
interesting trend was the appearance of longer series starting on 17
September. There was a series of five
eruptions on 25 September, the last series for which data is available at this
writing. Could the increase in series
length be somehow related to the emergence of Giant from its period of
Lower Geyser Basin
loggers are deployed on Fountain Geyser, Great Fountain Geyser, and Silex
Spring at present. The loggers on
Fountain Geyser and Silex Spring were deployed after the initial eruptions of
Silex Spring once some loggers became available. I will not discuss the Silex Spring results here since another
article covers that subject.
some initial technical difficulties, we have a data logger on Fountain
Geyser. The data only covers the period
from 19 August to late September. In
brief, Fountain’s intervals varied widely, ranging from 4:00 to 11:30. There was a definite increase in intervals
in late August (between 27 and 31 August) from just over seven hours to about
eight hours. The median intervals have
remained around eight hours through September, and the shortest intervals have
been about seven hours.
Great Fountain Geyser
July, Great Fountain maintained median intervals around 10:30 with numerous
intervals over 13 hours. During August
and early September the long intervals jumped to as much as 16:30 on several
occasions, and for much of this time a 90% prediction window was plus or minus
three hours. On 21 September, Great
Fountain went into “wild phase” eruptions for at least five days. This wild phase was preceded by two
eruptions only 4:18 apart, occurring at 06:41 and 10:59. NPS personnel or geyser gazers witnessed
neither eruption as far as I can tell, so we do not really know what these
eruptions were like. Interestingly, the
wild phase that started on 21 February also began with two eruptions about four
and Lone Pine Geysers were monitored for most of the summer. Data loggers have also been deployed on some
less well-known geysers this summer, but I am unable to discuss these because
Geyser has been monitored for many years.
In the past two years Echinus has stopped its regular intervals of under
an hour and has become a relatively unpredictable geyser. The data logger showed an interesting
pattern of eruption intervals, as shown in the graph below.
The tall spikes on the graph
represent long intervals, some over five hours. A detailed look at the temperature record revealed that these
long intervals were actually a “normal” 1:45 interval that had several temperature
peaks that correspond to periods of overflow and boiling. These “false start” events each result in a
delay of just about an hour. The really
long intervals had four or five of these overflow events.
period between 15 July and 25 July was the period immediately following a
basin-wide Disturbance. Note that
Echinus became quite regular for these two weeks, and then resumed its erratic
behavior. By about 5 August, Echinus
had settled into a pattern of a day or so of regular 1:45 intervals, followed
by three or four days of erratic, long intervals interspersed with shorter
intervals. The periodicity of the long
intervals is an intriguing puzzle at present.
Lone Pine Geyser
Grigg and I have been monitoring Lone Pine Geyser for several seasons now. Generally, the intervals average 16 hours,
with some long and short intervals (over an hour longer or shorter) early in
the summer, then periods of stable intervals.
This summer from 4 July until late September the intervals were all
between 15:30 and 17:00. Late in
September one very long (22-hour) interval occurred on 22 September. Perhaps coincidentally, this was during the
12-inch snowstorm. A period of a week
in mid-September and another period in late September are unrecorded since a
coyote decided to munch the wire leading to the thermistor. I replaced the thermistor and wire, and
carefully buried the logger and wire under more rocks and tree branches. Unfortunately the coyote was more persistent
than most and managed to destroy a second sensor. Such are the tribulations of electronic geyser monitoring. Thermistors make expensive snacks for
this not-so-brief summary of our electronic monitoring has been
interesting. Nancy Hinman, the
Yellowstone Geothermal Geologist is looking into the monitoring program, so we
can expect a more structured approach to monitoring of the thermal features in
the future. One result of the large
number of data loggers is a growing mass of data files, and one of the ongoing
projects is cataloging these files with the view to making them accessible to
researchers. One of my projects is
analyzing the temperature records and producing the analysis that I have
briefly discussed here. We hope to make
some or all of the analysis available on the GOSA web site in the future.